ChillHouse (English) - Alexey Moiseenkov https://darkolorin.com/essays.html English translations of posts from ChillHouse. en Weekend thoughts. https://darkolorin.com/posts/1348.html https://darkolorin.com/posts/1348.html Sun, 01 Nov 2026 20:23:12 GMT <p>Weekend thoughts. Quite a while ago I came to the opinion that modern tech is very different compared to what it used to be. Let me explain.</p> <p>Most of the market and people in tech right now live in the past. Wow, you can make yourself a dashboard, some little calorie-tracking software, or a planner. Sometimes these dreams go beyond personal software and the “like it used to be” game starts. Like, I’ll make it, people need it. But it seems that time is gone.</p> <p>I like the following analogy. In the middle of the last decade, good cameras showed up in phones so you could shoot vids, all sorts of effects and whatever. That made it possible for anyone to shoot a decent clip, because now you don’t need a video camera. Absolutely everyone is “their own director.” The exact same thing is gonna happen with software. You’ve gotta become a blogger. Be a persona. Know how to pull people toward you. And the previous generation of entrepreneurs absolutely can’t do that.</p> <p>All the accelerators, VC, and the whole тусa are tuned to evaluate product, track records, and whatever. But as we know, that’s not something a blogger necessarily needs at all. And “growth,” which everyone used to evaluate too, now has to be evaluated differently. Now every other person under a YC launch writes about how they replicated the whole product with one prompt. And that’s exactly the same effect as with bloggers. Posting a similar video is easy as hell. But only a few go viral.</p> <p>Welcome to the attention economy instead of the product-and-growth economy.</p> <p>And serious software will stay right where it is. Models, research, integration, hardware, and so on. Where the barrier to entry will grow by hundreds of times. That’s how it is.</p> <p><a href="https://t.me/chillhousetech/1348?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A lot of stuff has piled up that I’d like to write about. https://darkolorin.com/posts/1347.html https://darkolorin.com/posts/1347.html Fri, 30 Oct 2026 15:36:42 GMT <p>A lot of stuff has piled up that I’d like to write about. There’s s1 Anthropic in there, and various other lending crises of the compute business. But I’ll start, I guess, with math. <br> Note: I’m not a scientist and even less a mathematician. But I spent half my life in math and was pretty deeply into it, I even used to go to the Steklov Institute at the Russian Academy of Sciences. Well, those who know, will get it.</p> <p>So. OpenAI dropped a seriously strong result last week. This isn’t some little trifles. This is something on the order of a Fields Medal (the Nobel analogue but for math). Very cool stuff. Sure, it still has to go through review. But still.</p> <p>So I think a very important time has come in mathematical disciplines. Moving in certain domains has become easy, and what’s left is to understand where to move. I think a big chunk of math was so far removed from the applied part that it developed as theory with no clear angle on how to use it. I think in places like that AI will help like hell. The scientist just needs to set the vector and it’ll find it, connect it, and prove it. Proof for the sake of proofs will start to vanish, and that was a huge layer of math.</p> <p>But I think the real juice isn’t in those areas. It’s in that very theoretical physics, or in areas at the junction of several sciences or directions. The most interesting thing is where AI will find something a human, due to the sheer size of the field, simply couldn’t pick out and line up. How much new stuff you can learn. And here, without “human in the loop,” I think there’s still no getting around it for now. It still doesn’t really go, building new concepts without being pointed. But it does brute force like a master.</p> <p>So it’ll become more interesting to go into science. And all sorts of purely-theoretical, three-level nudges will disappear. Because it’ll be way more important to be able to connect it to experiment and reality.</p> <p>But time will tell.</p> <p>Links:<br> <a href="https://openai.com/index/building-scientific-understanding-with-academic-communities/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/building-scientific-understanding-with-academic-communities/</a></p> <p><a href="https://t.me/chillhousetech/1347?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Dear readers, we’re actively collecting use cases for RL for computer use, for… https://darkolorin.com/posts/1346.html https://darkolorin.com/posts/1346.html Thu, 29 Oct 2026 23:13:46 GMT <p>&quot;&quot;&quot;<br> Dear readers, we’re actively collecting use cases for RL for computer use, for a local model, and so it runs fast, and I’d be super grateful if you take a minute and dictate yours. For starters we want to focus specifically on using the apps and capabilities that are already on your computer (basically built into the OS). The most active/coolest use case and its author will get 1k USD from us for any neural net as credits. We’ll count it in a couple weeks, I think. <br> You can dictate in either Russian or English, translation is built in. There’s a voting thing too <br> 🙂</p> <p>Share it with friends and acquaintances. Thanks!<br> Link<br> &quot;&quot;&quot;</p> <p>Links:<br> <a href="https://intents.trymirai.com/?utm_source=chillhouse&amp;utm_medium=post&amp;utm_campaign=tg_my" rel="noopener">https://intents.trymirai.com/?utm_source=chillhouse&amp;utm_medium=post&amp;utm_campaign=tg_my</a></p> <p><a href="https://t.me/chillhousetech/1346?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> How much do you spend on AI for coding per month? https://darkolorin.com/posts/1345.html https://darkolorin.com/posts/1345.html Fri, 23 Oct 2026 17:59:52 GMT <p>How much do you spend on AI for coding per month? (Or for other stuff, but that’ll be less)</p> <ul> <li>0, I write by hand </li> <li>$20 but I still write by hand quite a bit </li> <li>~$100, I barely touch anything by hand </li> <li>$200+ </li> <li>$1k+</li> </ul> <p><a href="https://t.me/chillhousetech/1345?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, you can mess around in the browser now Still in an experimental… https://darkolorin.com/posts/1344.html https://darkolorin.com/posts/1344.html Sat, 17 Oct 2026 14:57:56 GMT <p>By the way, you can mess around in the browser now <br> <a href="https://uzu.run/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://uzu.run/</a> <br> Still in an experimental form, but soon it’ll make it to the main branch in a proper state. Overall it’s behaving pretty decently. Also, should we bring in jev-like models? Feels like especially in the browser it could be kinda fun.</p> <p><a href="https://t.me/chillhousetech/1344?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A pretty interesting picture is shaping up https://darkolorin.com/posts/1343.html https://darkolorin.com/posts/1343.html Thu, 08 Oct 2026 22:26:18 GMT <img src="https://cdn4.telesco.pe/file/esINpqUE3XKFHlr_SCh5IxYfh8Fob1s067CAKta-l1gnYb1T_xwjLaxFXPtBkVjtHmOQit5_qfxQ_EbYVM7RdykKZ3KvT32l6zc53VR11V_SdPyebfXZwTeqzKPJZ1hkNc1z8c5WOU5y_GPmdvqcQpFRxVcS113OU9jHeyfGD_0jFA5uNW-vm0E0YLrsFUS3M4IS3iBUEABV2pUOrPcoIbO-4-GnaUhMpHTOPQ4aooh083854hWOzsUsYAx__70Or7f8xB3ggTaC1vEiVNJ0eIZmz8SA1nVkvoYv45IgM3STz4hmIxQHvd6_GoETcVvcOqcFo3K0maVYhfzJlr4bVA.jpg" alt="" loading="lazy"> <p>A pretty interesting picture is shaping up</p> <p><a href="https://t.me/chillhousetech/1343?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Great PR for the IPO, after all there are very smart people sitting there models https://darkolorin.com/posts/1342.html https://darkolorin.com/posts/1342.html Thu, 08 Oct 2026 18:23:37 GMT <img src="https://cdn4.telesco.pe/file/a3FzV37LzHfVfYY-Nm7UcA48SjFPu0cgmNAgpd6CWdkHGMpFqoy30vDKBL1NSZa3zRORLXvWLONEfRYK4Lj543WHVbH7QSzI5P4r7zusfrnL2DlUP-QASZJzLEADsy679IrhdZgD4frY7S8CdY5Oiygxa9P6vQtAsLcEMREvpA4EPwma5Q2OBPIjEW35Sa_OQ74OhNmTlLsnUaXt66Hbj2J1LdNBT849cai7yEefnziof4yrXFbGQb_qQ9_-FHuW3ILmExA57j2mwx5Gk0JBzisJB1vEBsKm2Doq1D0Qsv031M9ricP1CfcePcV8n7fDKaQPzsD-ErMgPRZgDI1nvQ.jpg" alt="" loading="lazy"> <p>Great PR for the IPO, after all there are very smart<br> people sitting there<br> models</p> <p><a href="https://t.me/chillhousetech/1342?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Hot take: working on a pretrain team in top labs is the most boring thing… https://darkolorin.com/posts/1341.html https://darkolorin.com/posts/1341.html Fri, 02 Oct 2026 22:14:42 GMT <p>Hot take: working on a pretrain team in top labs is the most boring thing imaginable. Everyone’s there just for the money. Prove me wrong.</p> <p><a href="https://t.me/chillhousetech/1341?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Turns out google can do something. https://darkolorin.com/posts/1340.html https://darkolorin.com/posts/1340.html Wed, 30 Sep 2026 20:38:13 GMT <img src="https://cdn4.telesco.pe/file/Hw4Bt4nZaZYAWUuJVCvd7i9iu6nhsKGm93MmdvvK1G3ViR_ezJcAvTwWxA5SAomBsleuZQi60vzJV6yf8X1c3cQsF4ayU-aKtTBun95YVkgpR511rsTGni0lkLtjP2U8IrQHNAO8NwQj_NA6ObtnQYrFh57q1W14KiMCfiNuvVmgX0EkNzp6u5NoMJ38cZmHU5jKUjk9KQh6c_kxE_KUKrt4mNqeoJD2h4govNokFu6bu0qBFtTUL_5R9ZxYGxQsAAZDT4mVessuhdU-KrSlF3Q7Z_rGBA5yPVLKaQa2-doHG325SMrdXJKY0lWIye2I3GOxHEW2PHS0hitMeCNVxQ.jpg" alt="" loading="lazy"> <p>Turns out google<br> can do<br> something. Or it’s just a table with little numbers.</p> <p>Links:<br> <a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/</a></p> <p><a href="https://t.me/chillhousetech/1340?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In my work email every day there are like 50 emails with the subject: we've… https://darkolorin.com/posts/1339.html https://darkolorin.com/posts/1339.html Mon, 28 Sep 2026 09:56:05 GMT <p>In my work email every day there are like 50 emails with the subject: we&#x27;ve built new ai empowered gtm tool. After that it’s usually something like: cold outreach automation, market research, and most often of course SEO/GEO or whatever that thing is about rankings inside chat bots. Hundreds of emails, most of them written through AI, replies — zero.</p> <p>Almost everyone builds the email on the same principle: <br> • a hook in the subject (like some dude wrote Ferrari, turned out that’s the name of their tool) <br> • inside, some kind of reworked description of what the “agent” found about your company (usually it can’t really find anything, with your own eyes it’ll be three times faster to understand) <br> • a standard template about how we’ll find you a bajillion deals, deliver a bajillion results</p> <p>Sometimes I open <br> these guys’ sites <br> — 100% made by Claude</p> <p>Links:<br> <a href="https://www.enablement.ch/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.enablement.ch/</a></p> <p><a href="https://t.me/chillhousetech/1339?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Launched here a small experiment, a little model, basically a very solid… https://darkolorin.com/posts/1338.html https://darkolorin.com/posts/1338.html <p>Launched here <br> a small experiment, a little model<br> , basically a very solid compressed and pretty peppy one, gentlemen are testing it themselves (they even ported it to the required engine)<br> you can check what comes out in the video or read about the experiments from the guys <br> <a href="https://t.me/fuckup_files/1049?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/fuckup_files/1049</a></p> <p>Links:<br> <a href="https://huggingface.co/trymirai/Qwen3.8-27B-S-experimental?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://huggingface.co/trymirai/Qwen3.8-27B-S-experimental</a></p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1338?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> On the topic of apps and the AI boom. https://darkolorin.com/posts/1336.html https://darkolorin.com/posts/1336.html Fri, 18 Sep 2026 14:22:56 GMT <img src="https://cdn4.telesco.pe/file/BYAmE_3PCzTtxPMjMVkBtlii2A-gMycIAT2B4RivtiQxPnkfQrwng4JvGX_gauiztTV45CRoBNCF0Kp6HfwAWgC7Vr1nU5GlLNZSl53eIEKdddpWcxtcg5rfdp41105PiQPYDCPcwH7IY0QZKXCZwk3DvQpMqQNc7irgCViFMM1jBC3dWjtysDDtjk3xM2wxLkM6Huf011wP7SxU52s5pgTqyEuQsF2PRdZH-dBm1so3PDvsx3_BIq0mj_vjW0OoUg0noqQGvucrJvMpzsottbd8EmSoYb-Xsk6f8GLeQBSlWIKa4Qp9BS4AzCda1zgGm2kHRACXWbuo7z86zyAAnA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/Gd6qNB_Ry2OZr46U0v9z8PVKeuhu45nYbspoe68aYdk4Bao_ZVO42SMei5Az5b4ZBxm2LUvGGkO2SXBtWMUlZi0F7V5e-mWs9PsRWOvG-BFAvYVvOyCAvWAWuhEjlO5ckvEmeqFornXMPd5pWhc4STDH1v2xGHYQBapYk4vZVHHVCtDySv-ttdbcmNcOVZWJVa0h1tp3vQailDiKbv18vJKJpe0V4aP8GtIxzsCb3SI5dJY3O5triQ1ryTbjer71SJ_7pHqPjEIDvRoBe5y6Dxs5LOxSxzBVKvlOTPQ0beojKTlYUpPrJCL0kZ3KIi-scz-1-EUV5qnMSctJi2tRcw.jpg" alt="" loading="lazy"> <p>On the topic of apps and the AI boom. Basically there are tons of apps, but nobody uses them. Personal software is coming out.</p> <p>Links:<br> <a href="https://open.substack.com/pub/a16z/p/charts-of-the-week-so-many-apps-so?r=3p9xs&amp;utm_medium=ios" rel="noopener">https://open.substack.com/pub/a16z/p/charts-of-the-week-so-many-apps-so?r=3p9xs&amp;utm_medium=ios</a></p> <p><a href="https://t.me/chillhousetech/1336?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> On the topic of AI oversight https://darkolorin.com/posts/1335.html https://darkolorin.com/posts/1335.html Tue, 15 Sep 2026 14:36:30 GMT <img src="https://cdn4.telesco.pe/file/WPT7H8Vv70JYtmq3wWUDGPlfPbH070ZqCAep6ddpei3Mjr0sC_bfC5t1QsrhdWdGqSyycj0QPZgJ3CN4N6CZl4sceTWfLllldpf6mErvhr2REUMp_0vorQVbH7EzpI-H1e4GbbvkV8DZAdPKAser7ZeY7UhCGq-wcKtfv-INhgyu_GidecgukTBVcHE9aO6t8LBJRG4_DOVaA9hreX62MtNLAIOGmX6aw2awzttfyCJ8ZxGvPIxabR6myW2J9fxctm21Cpsq6cZmg1dUZqGH-KrNVpxuWqFx8NnqfDl7zKL04uc3ZYxwUVi1pOzGyfBiYpfJFVymz91O8lFKicCGlQ.jpg" alt="" loading="lazy"> <p>On the topic of AI oversight</p> <p><a href="https://t.me/chillhousetech/1335?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ve got one question for all these labs and companies that are pumping the AI… https://darkolorin.com/posts/1334.html https://darkolorin.com/posts/1334.html Mon, 14 Sep 2026 16:38:55 GMT <p>I’ve got one question for all these labs and companies that are pumping the AI hype up to the heavens. Why do people leave these top omega badass teams — and not just people, but leaders? Like, if you see that your so-called lab is on the brink of a breakthrough and conquering the whole world. And you bounce to go build your own startup around post-train. WTF?</p> <p>I mean, sure, maybe some folks don’t like the people and the vibe. But there’s money and opportunities coming out of their ears in there. Layoffs? Lol. Doubt it. But maybe it’s about equity.</p> <p><a href="https://t.me/chillhousetech/1334?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> New essay. https://darkolorin.com/posts/1333.html https://darkolorin.com/posts/1333.html Sat, 12 Sep 2026 15:11:41 GMT <img src="https://cdn4.telesco.pe/file/ePItZBy1vFuYpdVmrmR_97Qnd4sMaL3qdO5OHSn-gyHZOGqZkcco6sC0r_nJpE-vQ40WpRcS-MtTessnZ5lIvUiJs3VLVV3dvLjv7frPZvF7TqAu7gUGBLov-ebHsJsjqbbDqO29-S9jNanjfCti95hT25TD4h6a4ev_GmPlAPdga0TVH_v48XWYzdmHph0xhzxuT5GSnVYLzFKgRyyCBZh_WtGbKaaOFtC4yl8lnq3Hm3ubtO0tkesAcJ9STWn09KbqOWVhRy3nlJf8MP5srYGIk95S9ZEHCh4DvQqEhz_GxaQ1Xw7JUHQK2hR8w0uylh7MtE1s-nVlExkkuocEzA.jpg" alt="" loading="lazy"> <p>New essay.<br> Dario, head of Anthropic. The main point is in the screenshot. The rest is worth reading. Next it’s worth thinking about competition and an IPO.<br> And<br> Sam’s take<br> goes in the same direction. Probably someone else is supposed to think that way too, right?</p> <p>Links:<br> <a href="https://darioamodei.com/post/we-must-pace-the-frontier?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://darioamodei.com/post/we-must-pace-the-frontier</a><br> <a href="https://x.com/sama/status/2098811563415150910?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/sama/status/2098811563415150910?s=46</a></p> <p><a href="https://t.me/chillhousetech/1333?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Wanted to throw in some more memes. https://darkolorin.com/posts/1332.html https://darkolorin.com/posts/1332.html Thu, 10 Sep 2026 20:24:53 GMT <img src="https://cdn4.telesco.pe/file/fiGIcttDicRWjJydQ9Wpo-9Ab6YXV-x4t6SOikMvhvbLlRWuYnYlgRqa8fULBq-iCtTJgOk1Juncjx1XrPJSvGxzXN99xgpc1HGuckSUcxWyExjPtxne7HdipvpW3ovjPgnTJDu1dhw22jKjtOXRD9AVBK-CiJBrSfsc7hHWyEBCbL7T9cXjqaXZSy3AuwIQ92FK-L_RZLEapPdOPVqwEWXi8of7ZKMAYASPzXR6oZmcix5nhCMmbKsVzZxBM72y5FI5wIJTg9HGaUd7uwMZIxe5QC3A05FS9AWrHXA7rKo3MOxRD2_3jtvU_lOO5lQe3eToZ5B81RMV3jmV2gNvVg.jpg" alt="" loading="lazy"> <p>Wanted to throw in some more memes. OpenAI <br> halted<br> new sign-ups for the 200-dollar plan. A few moments before</p> <p>Links:<br> <a href="https://x.com/thsottiaux/status/2098113585683808624?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/thsottiaux/status/2098113585683808624?s=46</a></p> <p><a href="https://t.me/chillhousetech/1332?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> To everyone especially freaking out, I’ll remind you of a few details about… https://darkolorin.com/posts/1331.html https://darkolorin.com/posts/1331.html Thu, 10 Sep 2026 11:49:09 GMT <img src="https://cdn4.telesco.pe/file/S2KQaJIqP2mrhoVP7w8iIzHsmvSyr37_3kmdZiVbnQcCeOB3UJiFFUbLsArRyPzRQczUQIV1ra7xNKEoDkW45Q0w6plTPbOklQxhNEeF5lIYm_XkWRdw8FVTnwZSlq4iSYFOiA7uaaz7f0Q4d0t8MvVYJIYDTtNThGV46XcOSWsTcOcs2a_eRyR3tXKm1RFOrtzzXC7--n37Jf0FV787fKcmtv49Zn1RARkp3vdLrYHUj36r1foX_atfXKpkAtt3_or7i88goF2qIC3qDkN3pRDumPgEJW5yr091fsqSAxrgJoKJu7p0GrxcbCjbflja4YEoA2-zTWYKp5mZ1jsJxA.jpg" alt="" loading="lazy"> <p>To everyone especially freaking out, I’ll remind you of a few details about the latest events:<br> 1/ the HuggingFace “hack” by agents is, damn, not “some model sitting there thinking how it could break some service.” It was a very specific set of instructions from exploitgym that directly told it what to do. Agents themselves aren’t breaking anything yet (on their own initiative). What is this nonsense?<br> 2/ AI solves a millennium-old problem all by itself! That’s it, we’re all done for. No. It didn’t solve it by itself. Some guys took a massive chunk of existing research and trained on traces and then brute-forced it intelligently (same as with other hypotheses). No model is solving anything on its own yet. Try it yourself.<br> 3/ but the funniest part<br> is this post by some 27-year-old dude who worked at OpenAI and then at Anthropic and started pushing this line: AI will kill us all. You’re more likely to get killed by something like a car on the road than by AI. If these risks bother you in general, I recommend not leaving your house starting now.</p> <p>Links:<br> <a href="https://x.com/parkerthayer/status/2097759699626328575?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/parkerthayer/status/2097759699626328575?s=46</a></p> <p><a href="https://t.me/chillhousetech/1331?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The future is here https://darkolorin.com/posts/1330.html https://darkolorin.com/posts/1330.html Wed, 09 Sep 2026 19:51:43 GMT <img src="https://cdn4.telesco.pe/file/YNg7WDcSWlMYOC6fNHUh1pMGgBbJicWRmw38ifyLkRzdoyBGCMFRpvBIak-j1cNpHomjQ9ASQdq5C_gsuHihrvpa2blYVIPLmstpScpe15dA30bPorNqqoSk2Rb5dpLKr_rn1q1kNNLckzec1GmROFRzjWNAlX63cvyDa-SYhE8E5waAbwDHbHGkdGGAm4-mfS0SjhdzdCEorT80YnsV6uBGLJA025ONKgvkH9TQEFaZ_WspBjdo1Laj2gVgz8m3u2qX1ltnXJoqqLuhteSTlubPl-nC8GA4_hAChlHcBdMhE7Ekexeb9ypN-Vwiz5g45mk1863dq88kRG3D9-kB7g.jpg" alt="" loading="lazy"> <p>The future is here</p> <p><a href="https://t.me/chillhousetech/1330?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well that’s a solid result. https://darkolorin.com/posts/1329.html https://darkolorin.com/posts/1329.html Tue, 08 Sep 2026 17:25:23 GMT <p>Well that’s a solid result. <br> Solving Navier–Stokes. <br> . <br> There’s some drama along the way like the guys figured out which direction to dig, and OpenAI peeked into their Codex logs and ran it on their own cluster with a beefier model :) <br> A bit more details, 300 billion tokens out. 10k agents. Now that’s brute force, I get it. <br> Still, top! Grats!</p> <p>Links:<br> <a href="https://openai.com/index/navier-stokes-solution/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/navier-stokes-solution/</a><br> <a href="https://x.com/willdepue/status/2097377325817974908?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/willdepue/status/2097377325817974908?s=46</a></p> <p><a href="https://t.me/chillhousetech/1329?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Major release from our small lab. https://darkolorin.com/posts/1328.html https://darkolorin.com/posts/1328.html Thu, 03 Sep 2026 15:45:00 GMT <img src="https://cdn4.telesco.pe/file/fegDcSrWz3YIB-wDSjrjW00XYI4rUGTMDsSuKDqYt7jY7TS36Y28zPqifbW_gTOgSBEdfwgDT0qHKr5igl-dTCCqZEh_9P3SDGSGP8CzC28sySkjn1p_xzBDDzeIx5PRiMeT0f6eRwyueE8271CSO59Lj4uXoGLRa8znvZafxlH6WGjUnQy5moKslaSoEzFTAdy1_jLT7YTWNe5jwarfR2x5RCXpNoNE3n-gHJPjlwqIHxnumkX6AgK_2arwuUTWZF53o1eYIaCqStwrEZnBWddVEFy8AuCwTY9Dg7MyppUa4JTquEVxEuAoL6PF7hW8-AX09rbLZMuMg0hCafOUtg.jpg" alt="" loading="lazy"> <p>Major release from our small lab. Support for speculative decoding in our inference engine. <br> . <br> To start: Qwen3.6 27B, support for Qwen3.8 27B and Muse Glimmer is on the way. <br> We massively outperform every possible stack for Apple Silicon on performance. Because we just went and did everything as co-design: kernels, quantization, and speculation. All from scratch, all ours. <br> You can try it already today <br> M checkpoint <br> - smaller <br> L checkpoint <br> (lossless) <br> All the models we support are here, we’ll be happy if you try something else too. <br> Benches are here. You can dig into what and how was measured and how much better it is. <br> More details about the release — blog post</p> <p>Links:<br> <a href="https://github.com/trymirai/uzu?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/trymirai/uzu</a><br> <a href="https://trymirai.com/local-models/alibaba-qwen3-6-27b-mirai-mirai-m-4?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/local-models/alibaba-qwen3-6-27b-mirai-mirai-m-4</a><br> <a href="https://trymirai.com/local-models/alibaba-qwen3-5-9b-mirai-mirai-l-8?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/local-models/alibaba-qwen3-5-9b-mirai-mirai-l-8</a><br> <a href="https://trymirai.com/local-models?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/local-models</a><br> <a href="https://trymirai.com/metrics?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/metrics</a><br> <a href="https://trymirai.com/blog/speculative-decoding-in-uzu?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/blog/speculative-decoding-in-uzu</a></p> <p><a href="https://t.me/chillhousetech/1328?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Typical cycle of an AI native company. https://darkolorin.com/posts/1327.html https://darkolorin.com/posts/1327.html Tue, 01 Sep 2026 21:43:27 GMT <p>Typical cycle of an AI native company. You build a harness for some task or an aggregator of APIs + prompts. You scale with paid traffic (90% die here because it’s actually hard). Then, once you’re big, it gets hard to ship features because you depend entirely and completely on the model (well, maybe it will, maybe it won’t). And you try like Harvey to make your own or at least post-train. That’s when you realize you’re worth billions, but you can barely ship features anymore and the labs are breathing down your neck. And where do you find people for all these companies who can do model training and evals and all that all that, and also in a month. And nobody knows. <br> Anyway, that’s how it is for a lot of people right now;</p> <p><a href="https://t.me/chillhousetech/1327?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well that’s not bad https://darkolorin.com/posts/1326.html https://darkolorin.com/posts/1326.html Tue, 01 Sep 2026 18:05:54 GMT <img src="https://cdn4.telesco.pe/file/uygr8Q_JNJYSb8Cb_gX4nWKibG59UwaLo7EBfszEjit8qJBMSxCVgEzSR8RAArFrO6hUhq5v7sLk91tTat_WsVQyItlTrJF4OQ35NKddgy5eQqkQVPMu-kqTZZT-YXcoBhxEXJrGe2ze82J-sLqhxhu1J9B9w8n1t96Vg-sNFfChXvrutYOTCfujr7pGLRKV-BYBCV1Y5SO1R8ttUyGe_r8PtEo_O2YCj2ui5MLPZa1hygktVWoOgKgAi9Z87SE499SFjZKN6HFp8NFGhUtbd-5ErYCUv5WNIjEmJb3z1I5Snn1DXE71wJ0_ELTOt_vg-IuarnJDW1l6GCQ0VyqewQ.jpg" alt="" loading="lazy"> <p>Well that’s <br> not bad</p> <p>Links:<br> <a href="https://platform.claude.com/docs/en/models/fable-5-1/overview?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://platform.claude.com/docs/en/models/fable-5-1/overview</a></p> <p><a href="https://t.me/chillhousetech/1326?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The “coolest” business ideas of this year. https://darkolorin.com/posts/1325.html https://darkolorin.com/posts/1325.html Mon, 24 Aug 2026 09:22:24 GMT <p>The “coolest” business ideas of this year. Like, “solid little themes” to try, with some small comments:<br> - all sorts of vertical RL envs, verifiers, other crap to make RL easier. Important thing: before you build, make sure you’ve got contacts in labs so you can sell them all this. Otherwise it’s a good business valued at 50–100M. Gets bought out by big companies and labs a lot.<br> - infra for robots, I don’t know much, but seems to land fine, I’m talking software. Same logic as the previous point. Preferably build it knowing the customer.<br> - heavy optimizations at the data center level etc. All kinds of super narrowly tuned stuff for inference or training models (where every run costs millions). Like, you know how to do training for a specific architecture properly or squeeze out higher MFU.<br> - pure research, yeah, because even if you don’t really do jack shit, they’ll buy you as a team. The whole point is to assemble a solid lineup. The rest isn’t that important. Betting on neo-labs at funds is popular. Because the money returns are almost guaranteed.<br> - application layer. But here it’s important to understand what’s relevant is more the UX part, not the business part. Labs and big companies have a need to implement certain features (look at OpenAI’s acquisitions)<br> Pretty much all of this is hard to validate without deep domain knowledge. Absolutely not relevant outside certain geographies and without talent onboard. That’s how it is.</p> <p><a href="https://t.me/chillhousetech/1325?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The worst thing a startup founder can link is that faux-leather design vibe by… https://darkolorin.com/posts/1324.html https://darkolorin.com/posts/1324.html Thu, 20 Aug 2026 23:39:02 GMT <p>The worst thing a startup founder can link is that faux-leather design vibe by Claude code design. Doesn’t matter if it’s a deck, a site, some tiny landing page. First thing you want to do is close it.</p> <p>Don’t know about you, but I can tell from the first scroll.</p> <p>Let me explain:<br> - nobody read the content, almost certainly — it’s abstract phrases, ambiguous wording, and so on<br> - waaaaay too much stuff, and in tiny font<br> - there’s almost never any thought-through narrative<br> - but the most important thing is: the industry is allergic to this<br> - oh right, and there are people who use<br> this kind of thing<br> as a browser plugin</p> <p>Links:<br> <a href="https://www.pangram.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.pangram.com/</a></p> <p><a href="https://t.me/chillhousetech/1324?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A new trend in venture capital https://darkolorin.com/posts/1323.html https://darkolorin.com/posts/1323.html Wed, 19 Aug 2026 11:40:04 GMT <img src="https://cdn4.telesco.pe/file/rlS5U_r4smbFO_IeiFnPYfoICwlOwE99nIJBeIbMybzHTO3UsjFGL_Ag3LKa306I2dEEBETLE6GdmqayokO6cqPWQMm5-Nhd0vAr1xGa77JxNdAEGgGnhQczmV3tkrHQe3KXf4EKEUjQ21mLE8tFgUTlKMQuD-0NRw187hj4QrB6eKOqXL9eFNakKJa8FiaFXogLt1l18A4f3nijMNXTi7B5t_71virUMRpvdIcE8LoQJA_U1fdFDdWr4p0dbxpVJGRvOn46dWXVCeMPhUel9jEDaXFYlBLBrvCEqhSpuamu7lY8aDs_pApYyLrKSN-rpkCjGVCo92QTpotdMcFeow.jpg" alt="" loading="lazy"> <p>A new trend in venture capital</p> <p><a href="https://t.me/chillhousetech/1323?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The first big casualty of the big labs and the Chinese models. https://darkolorin.com/posts/1322.html https://darkolorin.com/posts/1322.html Fri, 14 Aug 2026 11:38:14 GMT <p>The first big casualty of the big labs and the Chinese models. <br> Mistral <br> . <br> In reality their strategy is: we’ll build a data center for the local govs and other folks, and become like SAP/Palantir — consulting, rolling stuff out. And their own models, well, if you need them you can take them, here’s the API.</p> <p>Links:<br> <a href="https://venturebeat.com/infrastructure/mistral-ai-wants-to-build-1-gigawatt-of-european-compute-by-2030-and-lock-in-customers-now?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://venturebeat.com/infrastructure/mistral-ai-wants-to-build-1-gigawatt-of-european-compute-by-2030-and-lock-in-customers-now</a></p> <p><a href="https://t.me/chillhousetech/1322?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In short, a rage bait post. https://darkolorin.com/posts/1321.html https://darkolorin.com/posts/1321.html Tue, 11 Aug 2026 22:24:11 GMT <p>In short, a rage bait post. About business, entrepreneurs, venture, and the “grown-up tech industry.”<br> TLDR: venture as it exists now is over, there’ll only be big rounds, entrepreneurs from the previous cycle (like Instagram cofounders) have lost their value, their experience isn’t needed.</p> <p>Now in order. We’re standing at the threshold of a tech cycle. The entry threshold, obviously. AI is the kind of thing that cuts off all the “mid,” the “eh, kinda,” and everything below that in quality. The tech swaps two important things in priority: the ability to make and the ability to verify.</p> <p>Before, the biggest entry barrier was: assemble a team, find money to pay them, motivate them, etc. The flow of people into the industry was growing. There were tons of hypotheses, but to test them you needed time and resources. The hypotheses were super simple. Like oh let’s sell this, or tag photos on a map, tag clients in a spreadsheet and track progress. Everything was open (the internet was less regulated). That’s why entrepreneurs were famous for their soft skills. It wasn’t that important to know the problem and solve it, it was important to find people and iterate. That was valued above all.</p> <p>Today I see the picture differently. The barrier to testing a hypothesis has dropped to basically nothing. Simple things you can test on your knee. Look at the vibe-coding knockoffs. And then it turned out that solving the problem in the real world costs hundreds of millions, while your new SaaS is, God willing, two revenues. And for me the following flipped. Before it was important to find who with and set it up. Now it’s important to understand very precisely what to do, and ideally to formulate it concretely. And that notorious “knowledge” in the age of AI companies turned out to be very technical and very hard.</p> <p>The first companies like Perplexity &amp; Harvey made it in. And then reading the descriptions of companies raising 200m right away, it’s not even clear. And it’s not because they’re making bullshit (clowns lolol), it’s because you don’t understand what the essence of the problem they’re solving even is.</p> <p>After the crisis among entrepreneurs there will be a capital allocation crisis. Because now you need to understand “what” to give money to, not who to give money to. Which really breaks the investors’ world. And soft skills move to the background, because assembling a team for a clear problem and knowing what’s inside is easier than for “Tinder for dogs.” Which means you need investors who actually get it. And there will be few of those, because it’s easier for them to build their own company.</p> <p>You can already see all this now: top research people are leaving Google, faces in investment funds are changing, because the previous leaders just don’t understand what’s happening. Just straight up, basically, don’t grok the meaning of what’s going on.</p> <p>This stage exists in a lot of industries. Where there basically isn’t venture. Where there have become way fewer high-risk ideas with potential. Because the ideas got too complex. That’s what’s coming for us in tech too. And yeah, I wouldn’t invest money in a track record from past years, I’d rather throw it at students and good ML professors. That’s how it is.</p> <p>Maybe I’ll write a second part too. But I’m predicting a huge number of “great entrepreneurs” who, in the AI era, won’t get anything to work. At all.</p> <p><a href="https://t.me/chillhousetech/1321?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Observations from the last few weeks. https://darkolorin.com/posts/1320.html https://darkolorin.com/posts/1320.html Mon, 03 Aug 2026 19:45:25 GMT <p>Observations from the last few weeks. <br> 1/ there are two camps on how to reach AGI — we’ve got rsi (recursive self improvement) and world models. The first is represented by Anthropic &amp; OpenAI. The second by Google DeepMind. And because the DeepMind leader doesn’t believe OpenAI is pulling it off, you supposedly need to do both, and that’s why everything is so slow. Basically a corporate delay. But whose bet will hit is unclear to literally anyone. <br> 2/ people are returning to China — the ones working at top places in the US. Main idea: they believe they’ll be able, like the US back in the day, to pull off the “Chinese dream,” namely get valuation multiples for their companies that are several times higher. In other words we’re in for Chinese dominance in AI. I’d say that’s more than likely. Given the amount of effort and faith. And super talents. <br> 3/ a crisis is brewing in math. Many PhDs are, to put it mildly, going to get devalued. And I think there’ll be a big restructuring of the whole science world in general. Because AI will be able to literally discover and solve 50% of current problems. And there still aren’t any “totally new” meanings in sight. And I think the problem of finding something new is the same as the problem of writing text that resonates. Because in relativity theory what matters is not so much the theory as the idea. After that you can, so to speak, run the rest of the way somehow. <br> 4/ hiring is hell. A regular researcher at Anthropic (and other top labs) gets 10m+ over 4 years. Life is pain. <br> 5/ a lot of companies have barely realized that scaling depends on their ability to train their own models. Harvey, Cursor (couldn’t and gave up to do it somewhere else) and others <br> 6/ turns out quantum computing is developing really fast and there’s a lot of interesting stuff there especially in the area of quantum neural nets (lolkek), should I write about it? <br> What’s your news?</p> <p><a href="https://t.me/chillhousetech/1320?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Haven’t done this in a while. https://darkolorin.com/posts/1319.html https://darkolorin.com/posts/1319.html Sun, 02 Aug 2026 10:50:11 GMT <p>Haven’t done this in a while. Used to be on the channel once, back in the beginning of time. Anyway, AMA or ask me anything (ask whatever). Drop it in the comments. I’ll try to answer when I can. <br> Sunday stuff.</p> <p><a href="https://t.me/chillhousetech/1319?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> On networking. https://darkolorin.com/posts/1318.html https://darkolorin.com/posts/1318.html Mon, 27 Jul 2026 11:10:02 GMT <p>On networking. Overall it’s all simple. If they don’t want to talk to you, it means you’re doing something that isn’t interesting to the other person. Explaining, persuading, getting offended is just useless. <br> In the Valley it’s easy and quick to notice. And in general that’s normal. The second way to find people is to have a track record people are interested to hear about (that’s why corporate work at a good company is a great старт). <br> That’s it, fuck. There are no other ways. Warm intros don’t work if you’re doing “something nobody cares about” or you don’t have a track record. Admitting that to yourself is, of course, hard.</p> <p><a href="https://t.me/chillhousetech/1318?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ve got this one kinda contradictory argument. https://darkolorin.com/posts/1317.html https://darkolorin.com/posts/1317.html Tue, 21 Jul 2026 21:12:34 GMT <p>I’ve got this one kinda contradictory argument. Basically about entrepreneurs and business. Pretty much every “expert” for the last ten years has been selling: taste/eye, soft skills, broad horizons, and other crap. And it was always like “you don’t have to know the details, the main thing is the ability to see where to run.” Well AI, to me, is gonna flip that whole topic. It’s impossible now and in the future to know where to run if you’re not super deep in the subject. Like, understanding matters a lot. Because everything else gets commoditized through AI-ish stuff.</p> <p>Sure, people will say there’ll be plenty left where the nonverbal component matters a lot. Fine. I’m talking only about tech. Look at the quality of vibe-coding products from “big-name” founders. They were handed the best tools. And what do we see? A large chunk of it is something nobody needs. Because their strength was “ability to manage teams,” sales, and so on. Not the quality of the solution/product. And to make quality you have to actually know your shit. Now some schoolkid who knows his shit will sell faster because his solution is better and cheaper. And it’ll be possible to prove and show that. Because the tools are the same for everyone now.</p> <p>In other words, the ease of production is gonna massively devalue a huge category of “businessmen from the past,” and they won’t be able to adapt. Because learning new stuff is extremely hard, especially in an industry this fast. Prove me wrong.</p> <p><a href="https://t.me/chillhousetech/1317?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Kimi k3 dropped, and now a new Qwen 3.8 max is dropping too, which will also… https://darkolorin.com/posts/1316.html https://darkolorin.com/posts/1316.html Mon, 20 Jul 2026 09:36:20 GMT <p>Kimi k3 dropped, and now a new Qwen 3.8 max is dropping too, which will also be aka a “fable class model.” The internet will be flooded with reviews like, look, the model can do everything the big labs can. Let’s try looking at it from a slightly different angle. Because the releases are very important, but the point is different.</p> <p>Over many years we’ve seen the Chinese catching up to Western models. Mostly by buying up traces and data “kinda like distillate.” You have to understand that distillation by itself isn’t just dumb copying, it’s like for RL it solves the cold start. So yeah it matters to “steal” data, but it doesn’t solve much on its own. Next, the guys are really good at architecture and training on old GPUs and clusters. That’s cool. Plus open source (here the OpenAI policy director, btw he’s literally a politician, started writing on Twi and spectacularly shit the bed, I haven’t seen this level of nonsense from people of that rank on the internet in ages, like the dude has never been online even once, and he writes that he’s been on the net since childhood, go read it yourself because it’s interesting). But it’s important to understand that the main effect of these releases is public information that a lot of tasks can be pushed to a very high level and quality just by doing good RL. See frontend arena. There Kimi k3 wrecks everyone and everything including sol &amp; fable. Meaning you don’t have to take some arch-expensive new top model and pay a ton of money if you only need to solve a certain type of tasks. In other words startups like thinking machines &amp; engram.ai are doing exactly that. Like come here, dear enterprise, and we’ll do everything for you. And now every business in the world has understood this. Healthy competition, finally!</p> <p>Such is life.</p> <p>Links from the original post:</p> <ul> <li><a href="https://x.com/deanwball?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/deanwball?s=21</a></li> <li><a href="http://engram.ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://engram.ai/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1316?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well then, here’s our first release to the archive. https://darkolorin.com/posts/1315.html https://darkolorin.com/posts/1315.html Thu, 09 Jul 2026 07:49:47 GMT <p>Well then, here’s our first release to the archive. <br> We achieve a 4.37-fold speedup over autoregressive decoding, and outperform a highly optimized DFlash baseline by 24.7% <br> This is speculative decoding. SOTA (well, on qwen3.6 for sure). We beat everything known in public (well, dflash). A super important topic for fast on-device inference, especially when you actually understand the task. <br> GitHub <br> HF <br> Support in our inference engine will roll in soon enough</p> <p>Links from the original post:</p> <ul> <li><a href="https://arxiv.org/abs/2607.06763?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://arxiv.org/abs/2607.06763</a></li> <li><a href="https://github.com/trymirai/sglang?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/trymirai/sglang</a></li> <li><a href="https://huggingface.co/trymirai/weaver?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://huggingface.co/trymirai/weaver</a></li> </ul> <p><a href="https://t.me/chillhousetech/1315?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Lilian Weng, the cofounder of thinking machines, always has interesting posts. https://darkolorin.com/posts/1314.html https://darkolorin.com/posts/1314.html Tue, 07 Jul 2026 07:59:22 GMT <p>Lilian Weng, the cofounder of thinking machines, always has interesting posts. <br> This time it’s about how to build harnesses, or the wiring/scaffolding/environment for running models to solve specific tasks. <br> I’d say a really important point that’s very often missing in “duct-taped” harnesses is this thing:</p> <p>The design should be deliberately simple and generic to enable generalization, likely with reference to existing software engineering practices to benefit from prertaining knowlege.</p> <p>And my take is that we’ll be redesigning modern OSes around an AI that’s always running in the background (i.e., inference should basically be spinning forever in the background).</p> <p>Links from the original post:</p> <ul> <li><a href="https://lilianweng.github.io/posts/2026-07-04-harness/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://lilianweng.github.io/posts/2026-07-04-harness/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1314?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Notes from the field. https://darkolorin.com/posts/1313.html https://darkolorin.com/posts/1313.html Thu, 02 Jul 2026 12:52:14 GMT <p>Notes from the field. We use coding agents for all sorts of tasks. But specifically end2end for experiments, to see what’s what (like rewriting a test app from Electron to Rust and stuff like that).<br> So yeah, I’ve got the impression that fable from Anthropic is the first model where you already have to think about what to sic it on. Otherwise it’s reaaaally expensive. Like, just vibing for yourself with this thing — not gonna work. And we arrive at a simple thought. To understand where to point it you need a person who actually knows and can do. And in my head a very clear picture forms of why labs hire by the ton.<br> Hard skills demand has spiked hard in the industry like never before. Demand for understanding, skills, intuition is super high. Because the rest the model will cover if needed. But if there are no skills and no understanding — a ton of money will go up in flames. We’ll see tons of rollouts that fail not because the model isn’t effective, but because the human in the loop couldn’t. And if earlier you needed years, probation periods for that conclusion, now you can figure it out in a month and that’s it.</p> <p><a href="https://t.me/chillhousetech/1313?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh, Fable they’re bringing it back https://darkolorin.com/posts/1312.html https://darkolorin.com/posts/1312.html Tue, 30 Jun 2026 23:57:54 GMT <p>Oh, Fable <br> they’re bringing it back</p> <p>Links from the original post:</p> <ul> <li><a href="https://x.com/anthropicai/status/2072106151890809341?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/anthropicai/status/2072106151890809341?s=46</a></li> </ul> <p><a href="https://t.me/chillhousetech/1312?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Fresh trend in the Valley. https://darkolorin.com/posts/1311.html https://darkolorin.com/posts/1311.html Sun, 28 Jun 2026 17:24:33 GMT <p>Fresh trend in the Valley. Suddenly (wonder why) a lot of people realized that token costs for Anthropic and OpenAI are too high and maybe you can somehow cut them down. And the geniuses got the idea to crank out a ton of startups doing post-training of models on company data etc. They even raised 200m (of course they’ll write it up nicely about everything and anything that’s supposedly going to be there “as a product”).</p> <p>Overall the idea isn’t bad for Enterprise. With a good team you can replay it in pretty much any market or something close to that. Curious how the market behaves in general ahead of the AIs’ IPO.</p> <p>Links from the original post:</p> <ul> <li><a href="https://engram.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://engram.com/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1311?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Remember how you asked about decentralized inference and all that. https://darkolorin.com/posts/1310.html https://darkolorin.com/posts/1310.html Sat, 27 Jun 2026 17:21:22 GMT <p>Remember how you asked about decentralized inference and all that. So, food for thought:<br> - this is what a normal<br> stack<br> looks like, what do you think, is it реально to make players install this, so to speak?<br> - what’s the fundamental difference from regular cloud infra with OSS models? (OpenRouter?)<br> - oh right, inference on its own is supposed to come bundled with tons of other features<br> - if the models are small enough that some regular dude can host them, that means you can run them locally, meaning for free<br> - and yeah, who the hell provides the SLA, support, or anything important for a business? Right — nobody.<br> And let’s never use anything that has the words “utility token” in it, life’s taught us something, hasn’t it.<br> But I guess for some people it might be fun to play around with.</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.baseten.co/resources/guide/the-baseten-inference-stack/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.baseten.co/resources/guide/the-baseten-inference-stack/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1310?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Tsk-tsk-tsk https://darkolorin.com/posts/1309.html https://darkolorin.com/posts/1309.html Thu, 25 Jun 2026 23:23:44 GMT <img src="https://cdn4.telesco.pe/file/jzj7uRVYXlbXOK_dqzvmNEpLj7zD7M7U5RDALBxn3IH8cWk5z2f858mTfMwOB0cL7sl-Dccy4yLrJTOdZ9Urc7akywyPkyyCd9hiJmPGfVrmrxQ5-fdR5ih1Gi0XnqvZfPqnj085zDTUc8P1yALJ3VKB47r1m03gg-EDbV18aX5n2qi7OmjJqBzUnedhWsFeyGzNDkT-0mS-X-RpWsLt6hAG3QVipD5u_b6Huyy_kyeVF4DilRcYQJoxK5bsdaRupNmmGzTg2oeFrk9Rpt7xHw6d2gna5tmW5W0rYLVruFSq8CV2TQ02L84tV-KubTkWWVHF9bO_-JLeFQfcngXEsQ.jpg" alt="" loading="lazy"> <p>Tsk-tsk-tsk</p> <p><a href="https://t.me/chillhousetech/1309?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A few observations. https://darkolorin.com/posts/1308.html https://darkolorin.com/posts/1308.html Wed, 24 Jun 2026 10:29:33 GMT <p>A few observations. About “job hunting.” I regularly stumble on Russian-language posts like “I’m awesome, but yeah, such is life, looking for a job.” I’ll try to toss in my two cents from the hiring side:</p> <ul> <li>if someone is sooo awesome, like insanely wow, and the comments are all these metaphors like “you’re wow,” then why the hell are you, sorry, looking for a job? There have to be reasons, and if there are uncontrollable circumstances, that’s exactly where I’d start</li> <li>all these weird, purely-ours sayings: my superpower, what I do best is see “profit/systemness/collect meanings and turn them into releases AND so on” — what is that even about? Never understood all these abstract descriptions. Imagine a footballer telling his agent “I see the team’s philosophy and I can figure out how to steel my resolve and play.” Well damn, everyone’s like that. Give me something I can actually touch</li> <li>the first thing I’ll do is go grab references, if the role is above line-level, and the description is exactly about that, and then I don’t give a crap about all your “searching for meaning and ability to assemble systems”</li> <li>life happens in different ways, sure, but you still need to bring your awesomeness into something measurable — either via references or via metrics</li> <li>please cut out your vibe-coding craft projects, they show your total lack of understanding of the market more often than the opposite</li> </ul> <p>And I’ve never once seen posts on the US market, where supposedly the competition is tougher, in the style of “I’m super awesome, but give me money, I’m looking for a job.” In my head those two things just don’t fit together very well. And probably that’s why they don’t do it here. Edited: for roles above line-level.</p> <p>My recommendations, after looking through a ton of resumes and reading posts and actually searching for people for my team:</p> <ul> <li>write honestly what you did and what you got done, don’t decorate it, if your position isn’t line-level</li> <li>focus, obviously, on what matters in the application</li> <li>remove all the superpowers, whining, life situation, and the long story — it kills any desire to talk (we’re strangers, after all)</li> <li>showing desire matters most</li> <li>specify who can recommend you</li> </ul> <p><a href="https://t.me/chillhousetech/1308?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Someone tossed me the idea that I could record mini-podcasts as voice… https://darkolorin.com/posts/1307.html https://darkolorin.com/posts/1307.html Tue, 23 Jun 2026 17:10:50 GMT <p>&quot;&quot;&quot;<br> Someone tossed me the idea that I could record mini-podcasts as voice messages, like 2–4 minutes. Works as a format?</p> <p>- Yes<br> - Kinda meh, I don’t like voice messages<br> - Don’t care<br> - Yes, but plz not often<br> - 🥳<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/1307?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There was a lot of work. https://darkolorin.com/posts/1306.html https://darkolorin.com/posts/1306.html Thu, 18 Jun 2026 17:24:27 GMT <img src="https://cdn4.telesco.pe/file/Q3qvUNxDJS9jdTzCRO2ql9TlXtV1dw47h0dwcSUeAPCJoCvNfFpSEKJ75WBVheSuw7Jz__TLlOKBYk0ZZicYBLD_8FXa3uiRxvQwh4iRIe8S3kVJjNKrr87LtE6kpik3DaKt8-AOU1c9I71JaDd8zCuvanrcP8K_p3jrVT0A9_M1O3GYCxstvHpmQtWcJFo1eHCJlo8m1o8KB3Kuyd90ZJ1cH3L25iSepF75oMQF8sFjyK9j9Cyu4-B9tteMqX2yKTRneSC3EF5xjhqSIRde6ehabpzY-2b_oD7j1BiTkeVt6x9tkSUcN5ksiZECUXYQYYKlB52koyPjfNKtkIYnDA.jpg" alt="" loading="lazy"> <p>There was a lot of work. But then some news rolled in too:<br> - fable got banned by Anthropic, there’s some kind of crap there honestly, and politics<br> - they bought Intercom for 3.6 bil<br> - they finally bought Cursor for stock worth 60 bil (so is this the most expensive startup acquisition? Whatever)<br> And meanwhile there’s a good pic about organic launches without marketing and pouring money in and traffic and all that stuff</p> <p><a href="https://t.me/chillhousetech/1306?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Btw we shipped our first research result. https://darkolorin.com/posts/1305.html https://darkolorin.com/posts/1305.html Mon, 08 Jun 2026 18:48:08 GMT <p>Btw we shipped our first research result. <br> Rolled out the first quantized versions of Qwen 3.5 models. <br> Thanks to joint optimization of quantization + the inference engine, we managed to noticeably beat Unsloth/llama.cpp and MLX on the key “quality vs speed” tradeoff. On Apple devices from iPhone to beefy Macs, the models push 40–60% more tokens per second with no quality loss. <br> Post <br> hf collection</p> <p>Links from the original post:</p> <ul> <li><a href="https://trymirai.com/blog/quantization?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/blog/quantization</a></li> <li><a href="https://huggingface.co/collections/trymirai/qwen35-release?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://huggingface.co/collections/trymirai/qwen35-release</a></li> </ul> <p><a href="https://t.me/chillhousetech/1305?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Noticed this thing with “successful investors and entrepreneurs.” They very… https://darkolorin.com/posts/1304.html https://darkolorin.com/posts/1304.html Sun, 07 Jun 2026 11:05:41 GMT <p>Noticed this thing with “successful investors and entrepreneurs.” They very often dish out advice like: don’t cling to mistakes for too long, learn to move past them; or don’t try to execute a hundred ideas at once, invest in one thing and keep focus. You know, those standard rails. And then usually I hop on a call with that person—whose one of the latest posts is exactly that—and ask how things are. And he’s doing a hundred ideas at once, he doesn’t give a damn about people, the idea is what matters, and so on. Or in other words, he completely contradicts what he writes and what he advises everyone else. Seen that?</p> <p><a href="https://t.me/chillhousetech/1304?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Note to Claude users. https://darkolorin.com/posts/1303.html https://darkolorin.com/posts/1303.html Thu, 04 Jun 2026 21:36:34 GMT <p>Note to Claude users. But overall a really measured article came out. Extremely unlikely that Dario even looked at it. <br> Without that judgment, Claude is a capable assistant, but not a system that could drive AI progress on its own. <br> It is genuinely unclear whether today’s training methods and architectures could unlock that capacity.</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.anthropic.com/institute/recursive-self-improvement?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/institute/recursive-self-improvement</a></li> </ul> <p><a href="https://t.me/chillhousetech/1303?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ll toss something into the stash about frontier labs. https://darkolorin.com/posts/1302.html https://darkolorin.com/posts/1302.html Wed, 03 Jun 2026 12:04:21 GMT <p>I’ll toss something into the stash about frontier labs. Online people love to yap about Open Weight Models. Like they’re about to catch up any minute now and are always breathing down their necks. What I think after talking to different folks, and what problems I see:</p> <ul> <li>the talent problem: how do you hire a team to build a small or open model if a solid research talent at Anthropic is pulling $10M a year (real numbers), and you still need compute and all that stuff </li> <li>say you’re ready to invest in training people for a year or two, but then you’ll be catching up forever — how do you solve that? well you need to raise a hundred mil or two right away and chill, otherwise I don’t really get what the competitive advantage in the offer is </li> <li>but say you do lure people with upside, then there’s this opinion going around that “frontier models are 90% of the market in money”, where do you get data for training and RL from, i.e. literally everyone only buys “the best” (they’re ready to pay a market premium for “a model that’s a bit better” — pay 20% more or even 2x) </li> <li>a условный Opus is 10x more expensive than Chinese models and everyone buys it </li> <li>in this situation competition gets even harder because they start covering up the distillation process with the fact that there are closed models for industries and even special deploy teams that integrate those models in closed mode (see Mythos, and OpenAI’s math models). Or in other words: distilling a model into Claude Code will be easy, but the one that runs in a bank — no way.</li> </ul> <p>All this looks like a typical oligopoly and I think there won’t be room for OSS over time except for certain verticals or solutions. I think the Chinese make everything open just to have at least some data and attract people. I don’t see any other real reasons.</p> <p>Oh and decentralized inference is bullshit.</p> <p><a href="https://t.me/chillhousetech/1302?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> When you confidently pitch your idea to a seasoned VC https://darkolorin.com/posts/1301.html https://darkolorin.com/posts/1301.html Sat, 30 May 2026 13:14:41 GMT <img src="https://cdn4.telesco.pe/file/DBOQME7wa_SVHbNhhgNnc9c0rEVzMHGO0P14erwCBRGEN3BJ5f07rZMYseZue9Cz-JPH5OsVaY7HqsIGEeLBffkDPcRxc-Ye8cnpdv0NaTl0Y_YrH-gOXZYvEySWVng_dKNVzHCiAG6MhK3alDEpcLTfr-2o5FYDG3k7MqapWAYBP5qSPR56BcB81fSahrd5ZZ1fd01cwvLmtjf6IWbkDQW6iN7-ggXaBIc0IQ4gFmHocg1B1myc0_7m2avWSlgSnOVqH03bs697bmOKc_WPdzfI9psIkEF2LtFhFsGSWb92QV3WTTuscfVUuoLpw6GnAvBk32AWhblEadq5-3UNRg.jpg" alt="" loading="lazy"> <p>When you confidently pitch your idea to a seasoned VC</p> <p><a href="https://t.me/chillhousetech/1301?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Ok ok, sold https://darkolorin.com/posts/1300.html https://darkolorin.com/posts/1300.html Thu, 28 May 2026 17:07:20 GMT <img src="https://cdn4.telesco.pe/file/Rm5_Rx7cOCKq1Y9iPQmpe7FSVs7331fG1L8QdcBQSloTneVmBlaotFbgcC_02LmjUjRWnDgDXRhEr2cXLpMTQa4XpcUKk_FjEq4na6Hs79uWsdLo6bivE1hweNOyJwiVNUvH00MgvH8hfT_GgaZZ2umUVyUxT6kN2AAnJIDg2NjoqLNy_BPnwM5HLWjJ4wPp-lojXI_4f_VlhzpXk9wlpYqKY2CmdWzMM80MXDccCRpBevFyMqTEngDUouJbPk04Q-DaDuZQs8HAqwhlfM5umM-rCMGhNQC5PQfY6XI4u-kDJQAN2PjuFARUBJ7bvOO6NKyX-P26eq8FVKMBibfHkg.jpg" alt="" loading="lazy"> <p>Ok ok, sold</p> <p><a href="https://t.me/chillhousetech/1300?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Found a great overview about solving Erdős problems. https://darkolorin.com/posts/1299.html https://darkolorin.com/posts/1299.html Wed, 27 May 2026 20:42:24 GMT <p>Found <br> a great overview <br> about solving Erdős problems. Highly recommend. And yeah, I’ve already said that overall the <br> channel <br> is top-tier (even if the lifestyle posts aren’t my thing). <br> I mean, right now AI is doing what people are unlikely to do because of their human nature: systematically combing the field. The fact that in that field it turns up unsolved Erdős problems is more a fact about the state of human mathematics than about AI models. But still, a fact.</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.sergeynikolenko.ru/blog/gpt-and-unit-distances-erdos-hypothesis?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.sergeynikolenko.ru/blog/gpt-and-unit-distances-erdos-hypothesis</a></li> <li><a href="https://t.me/sinecor?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/sinecor</a></li> </ul> <p><a href="https://t.me/chillhousetech/1299?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Have you also started noticing even when the text/email/design/content is… https://darkolorin.com/posts/1298.html https://darkolorin.com/posts/1298.html Wed, 27 May 2026 13:59:06 GMT <p>Have you also started noticing even when the text/email/design/content is mostly made through AI?</p> <ul> <li>Yeah, but not right away, it took time</li> <li>Yeah, you can tell immediately</li> <li>Nah</li> <li>I don’t give a damn how it’s made</li> </ul> <p><a href="https://t.me/chillhousetech/1298?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> All the ads have been hanging there for way too long. https://darkolorin.com/posts/1297.html https://darkolorin.com/posts/1297.html Tue, 26 May 2026 14:24:13 GMT <p>All the ads have been hanging there for way too long. I even forgot the last post was like that. <br> Let’s bring back the read-but-not-fully-understood column. <br> 1/ stumbled on this in the course of vibe-coding, so to speak. I don’t really practice it myself, but it’s funny how people who dump hundreds of hours into it later think “wait, why though.” The guys have a funny observation. <br> A big article, but I recommend it to those who are into this. <br> 2/ dvarkesh has started looking like a student and is calling in professors; this time it’s about chip design <br> 3/ in any confusing discussion I recommend this article <br> 4/ a tiny bit about games and about game balance, because I’m throwing matches and getting pissed <br> 5/ we shipped a new site btw Trymirai.com — give me criticism <br> 6/ about product management, recommend (in the age of AI)</p> <p>Links from the original post:</p> <ul> <li><a href="https://every.to/p/after-automation?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://every.to/p/after-automation</a></li> <li><a href="https://youtu.be/oIk3R-sMX5o?si=UK12voUpKjVma2Jn&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/oIk3R-sMX5o?si=UK12voUpKjVma2Jn</a></li> <li><a href="https://www.lesswrong.com/posts/WQFioaudEH8R7fyhm/local-validity-as-a-key-to-sanity-and-civilization?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.lesswrong.com/posts/WQFioaudEH8R7fyhm/local-validity-as-a-key-to-sanity-and-civilization</a></li> <li><a href="https://youtu.be/uCPQGncOx-Q?si=_cZj3lGm_9or35xq&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/uCPQGncOx-Q?si=_cZj3lGm_9or35xq</a></li> <li><a href="http://Trymirai.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://Trymirai.com/</a></li> <li><a href="https://x.com/tdrobbo/status/2053836764997959898?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/tdrobbo/status/2053836764997959898?s=46</a></li> </ul> <p><a href="https://t.me/chillhousetech/1297?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Promo time. https://darkolorin.com/posts/1296.html https://darkolorin.com/posts/1296.html Fri, 22 May 2026 08:35:10 GMT <p>Promo time. I got asked recently, like, even with your attitude to vibe coding, what would you recommend. Obviously: go and use your own brain before doing anything, but if you really gotta, then my good friends are making a course on the topic. Quick info below:<br> Vibecon course (they showed me the syllabus, it’ll do, since we’re not expecting a linear algebra or probability theory course there).<br> What I liked:<br> • 3 weeks, 5 sessions, not dragged out<br> • solid setup: cursor + claude code + next.js + wispr flow<br> • they rebuild the program for every cohort (because everything gets outdated literally in a month, and that’s true), and then students get access to all the updates</p> <p>Bottom line. If you already know how to learn on your own with ChatGPT, this isn’t for you. But if for some reason you still aren’t using Claude Code, it’s a good chance to start<br> 🗓<br> Start: May 26<br> 🔗</p> <p>Site with details<br> They’ve also got a short test to see how “AI-pumped” you are right now. If you’re curious to measure yourself —<br> a bot with a quiz</p> <p>Links from the original post:</p> <ul> <li><a href="https://vibecon.cc/r/50?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://vibecon.cc/r/50</a></li> <li><a href="https://t.me/ai_native_test_bot?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/ai_native_test_bot</a></li> </ul> <p><a href="https://t.me/chillhousetech/1296?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A thought crept in. https://darkolorin.com/posts/1295.html https://darkolorin.com/posts/1295.html Thu, 21 May 2026 12:37:17 GMT <p>A thought crept in. Doesn’t it seem to you that a harness is technical debt, since there are systems there that rely on specific prompts. Like, in practice we’re hoping that a phrase worded in a certain way will give a better result. But that changes with every model version.</p> <p>So even if the model changed (even if they suddenly slipped a quantized one into your API or switched the inference version) everything can drift apart on you? Turns out prompts—and the more of them you have—are a kind of technical debt, and a hefty one at that. Maintaining something you don’t understand is a so-so task. In the sense that without knowing what the model does well and what data they poured in, making prompts is gonna be rough.</p> <p>Overall my take: if the model was trained properly on certain tasks, the prompt’s influence won’t be that big. But on the other hand Cloudflare reports after experiments on Mythos that, like, harness is our everything. Where the truth is isn’t clear, but what’s definitely clear to me is: building “forever” harnesses without your own model is shooting yourself in the foot.</p> <p><a href="https://t.me/chillhousetech/1295?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A few bits of news and comments: 1/ OpenAI went and solved one of Erdős’s… https://darkolorin.com/posts/1294.html https://darkolorin.com/posts/1294.html Thu, 21 May 2026 00:31:13 GMT <p>A few bits of news and comments:<br> 1/ OpenAI <br> went and <br> solved one of Erdős’s problems. This is straight-up good. Very good. I haven’t had time to dig into the solution yet, hope someone posts a normal overview (not from the people who were giving quotes for the OpenAI blog). But overall, a lot of folks were saying they’d go after Erdős problems specifically. Since the task is basically combinatorial, and I wrote at some point that in this kind of task where you have to make complex deep hypotheses about structures, there will be a lot of breakthroughs. Hell yes! A real assistant. Wonder how many tasks like that there are. <br> 2/ On math: I *highly* recommend this article on Substack. Also, if enough people subscribe, maybe I’ll start mirroring longreads there. The article is about how math creates new ways of knowing, and you won’t discover them by brute force. Or that proofs aren’t the goal of math. Very long text—if you can’t be bothered, read it with an LLM. <br> 3/ Anthropic pays xAI a billion a month for compute. Nice. <br> 4/ Karpathy moved to the pretrain team over there, where they pay Elon a billion for compute. Obviously for the money. <br> 5/ they say Claude mythos is coming soon</p> <p>Links from the original post:</p> <ul> <li><a href="https://openai.com/index/model-disproves-discrete-geometry-conjecture/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/model-disproves-discrete-geometry-conjecture/</a></li> <li><a href="https://open.substack.com/pub/davidbessis/p/the-fall-of-the-theorem-economy?r=3p9xs&amp;utm_medium=ios" rel="noopener">https://open.substack.com/pub/davidbessis/p/the-fall-of-the-theorem-economy?r=3p9xs&amp;utm_medium=ios</a></li> <li><a href="https://open.substack.com/pub/moretosay?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://open.substack.com/pub/moretosay</a></li> </ul> <p><a href="https://t.me/chillhousetech/1294?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Seems like the most useful announcement at Google’s conf is, of course, the… https://darkolorin.com/posts/1293.html https://darkolorin.com/posts/1293.html Tue, 19 May 2026 21:19:12 GMT <img src="https://cdn4.telesco.pe/file/rdrRRc-xD7KZMDQ0Mib1GSzMHNWlNtKEkmPEN_YT6frW_5U_gRsAK4Xn9CSHEX6gqhis4DsFe6Dhm5ZFTTUrqkVZWN4u1Z7hiUl5CnnNZFPJzOMe8SGhOS278nykOP2NhCiV3CtEOEW4kckcyLWwwvtp4KGTFqdQwVLOz7Pf3cC5Pp76kx80vTgg8H8oKrYXHb-vY-OxC1JukxYTTVCq7Khwz3Esg1VELQ8SL5sjXA6VY5lbN6eqxJC7kgUrX_BXVtWhvBWs-P9QQnf1jyoZ4hLKrZNuJ2lVOJXCMc5XDXaNxEWeZ5rFP569WXU-GCan7SMkaYubxMYV0_kfUWkK8A.jpg" alt="" loading="lazy"> <p>Seems like the most useful announcement at Google’s conf is, of course, the realization that there aren’t enough GPUs for everyone. Like, for everyone.</p> <p><a href="https://t.me/chillhousetech/1293?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Elon didn’t win the lawsuit against Sam. https://darkolorin.com/posts/1292.html https://darkolorin.com/posts/1292.html Mon, 18 May 2026 18:00:04 GMT <p>Elon didn’t win the lawsuit against Sam. Well anyway, I don’t think the goal was to win in the first place. But if it was, then it’s clownery, pure and simple.</p> <p><a href="https://t.me/chillhousetech/1292?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Observations. https://darkolorin.com/posts/1291.html https://darkolorin.com/posts/1291.html Mon, 18 May 2026 11:58:54 GMT <p>Observations.<br> Every week I get about 20 projects/products/teams coming in and telling me what they did/plan to do and so on. What I usually notice:<br> 1/ Decks, all the materials, blurbs, emails are made through AI. You can straight up see that people didn’t read what they themselves made, decks are a hundred-fucking pages, vibe-coded frontends—you know, with that Claude Design aftertaste (that’s when you ask it to make it pretty please). And they all expect that someone will dig into this and study it and, what’s more, give them money?<br> 2/ Most projects don’t know evals even exist. An agent to “plug in some thing” or we’ll improve your agent for “some thing.” But how to understand that the proposed approach is better than the basic “just feed everything to ChatGPT,” no clue. At least go look at benchmarks and scoring systems or something.<br> 3/ Well even if there are zero evals and benches, then you should be paying for that. Otherwise it’s really hard to understand what to do and how to evaluate any of it.<br> 4/ A ton of “slop,” meaning I spend 2–3 minutes and see factual/numerical errors in the materials. Like they write market size and two numbers multiplied give the wrong result. And there’s just a ton of that. No desire to read and talk to a person who didn’t even check what they wrote.<br> 5/ Everyone asks for millions, and I don’t understand why. Everything that was done was done with a $200/month subscription. Why do you need millions?<br> 6/ People don’t know the basics. Yeah yeah, the difference between a harness and a model, what metrics mean what, don’t know about recall etc. in search tasks. What is there even to talk about then?<br> It adds up to a strong feeling that they don’t even write the texts and answers to questions themselves. And that’s sad, and that’s exactly why it’s gotten harder to do anything if you don’t think for yourself, but if you’ve got your own head on your shoulders then now is the best time, of course!</p> <p><a href="https://t.me/chillhousetech/1291?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> It seems to me the key mistake a lot of people make — both founders and… https://darkolorin.com/posts/1290.html https://darkolorin.com/posts/1290.html Sun, 17 May 2026 17:44:56 GMT <p>It seems to me the key mistake a lot of people make — both founders and funds/investors — is this positioning like: achieving something means getting into YC, speedrunning, taking money from a top fund. <br> That’s a distortion. And it seems to me people focus on that more than on what they, damn it, are actually doing. <br> Because the main thing is to take money from top guys (who of course were almost all top at some point ages ago, when getting money from them was harder, because they had little money).</p> <p><a href="https://t.me/chillhousetech/1290?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If you saw somewhere some info, clickbait, an interview about how to make your… https://darkolorin.com/posts/1289.html https://darkolorin.com/posts/1289.html Thu, 14 May 2026 09:55:15 GMT <p>If you saw somewhere some info, clickbait, an interview about how to make your first cash in venture. With a project or an idea. Then please check these three points and don’t listen to air stirrers.</p> <p>They’ll give you money for your idea or team or product if:<br> 1/ you have a track record, meaning previous results that are known in the industry<br> 2/ you have metrics (measurable traction) at the moment you’re raising money and they’re too good to pass up (there’s also the option to show a wow effect — tech — but that’s rare)<br> 3/ you have a network of rich or badass people who trust you and will give you money</p> <p>That’s it. There are no other ways to raise cash. Decks, descriptions, advisor advice and all that — that’s already the next step.</p> <p>Next, all else being equal (obviously we remove track record): they’ll give money to someone young under 30, better 25, because it’s easier for them to risk their lifestyle and time.</p> <p>Next, the easiest places to raise money:<br> 1/ in the US<br> 2/ Europe/England<br> 3/ everything else<br> 4/ China/Korea if you know the language</p> <p>Doing a startup is one of the least likely ways to earn money for a living. You’ll more likely kill your nerves. But at the same time, probably the most interesting activity, full of action and opportunities.</p> <p><a href="https://t.me/chillhousetech/1289?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well finally it’s clear what they gave two yards to thinking machines for. https://darkolorin.com/posts/1288.html https://darkolorin.com/posts/1288.html Mon, 11 May 2026 20:52:42 GMT <p>Well finally it’s clear<br> what<br> they gave two yards to thinking machines for.<br> Interactive models. In short, a little model that listens to you in real time and immediately, in the background, figures out what needs to be done. It changes its behavior while you’re talking to it.<br> Nice that at least someone is getting away from this pathetic chat interface. Finally.<br> Video</p> <p>Links from the original post:</p> <ul> <li><a href="https://thinkingmachines.ai/blog/interaction-models/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://thinkingmachines.ai/blog/interaction-models/</a></li> <li><a href="https://youtu.be/A12AVongNN4?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/A12AVongNN4</a></li> </ul> <p><a href="https://t.me/chillhousetech/1288?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Posts from the people I sometimes read on Telegram, not an ad: 1/ some random… https://darkolorin.com/posts/1287.html https://darkolorin.com/posts/1287.html Thu, 07 May 2026 18:21:33 GMT <p>Posts from the people I sometimes read on Telegram, not an ad:<br> 1/<br> some random dude’s channel<br> who explains math analysis, ran into it on TikTok, feeling young again<br> 2/<br> Slava<br> just posts whatever, stuff I barely ever see elsewhere. Been знакомы for a hundred years.<br> 3/ and one more<br> Slava<br> and we’ve also known each other forever, about AI<br> 4/ at one point I seriously wanted to go into a quant hedge fund, but it passed me by, still it’s interesting to read what the cool guys from there say<br> what’s what and how<br> 5/ and in the CEOshnaya I got banned because like I supposedly don’t worship OpenAI kekw<br> And what do you read that’s not popular?</p> <p>Links from the original post:</p> <ul> <li><a href="https://t.me/qpxpq_math?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/qpxpq_math</a></li> <li><a href="https://t.me/kuteevschina?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/kuteevschina</a></li> <li><a href="https://t.me/slavasmirnov_ch?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/slavasmirnov_ch</a></li> <li><a href="https://t.me/rybasharp?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/rybasharp</a></li> </ul> <p><a href="https://t.me/chillhousetech/1287?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A list of so-called neolabs, where the focus isn’t revenue, but research or… https://darkolorin.com/posts/1286.html https://darkolorin.com/posts/1286.html Thu, 07 May 2026 13:34:47 GMT <p>A list of so-called neolabs, where the focus isn’t revenue, but research or something like that. Valuation at around a yard and it’s not clear what’s next. Curious to dig into.</p> <p>Stolen<br> from an investor at Menlo</p> <p>Links from the original post:</p> <ul> <li><a href="https://x.com/deedydas/status/2052266037979320752?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/deedydas/status/2052266037979320752?s=46</a></li> </ul> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1286?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> AI. https://darkolorin.com/posts/1285.html https://darkolorin.com/posts/1285.html Wed, 06 May 2026 10:51:34 GMT <p>AI. My two cents on the macro. <br> For the last half year I’ve been pretty deep in the тусовка, sometimes even around the people who decide stuff and build the industry — including funds, and big model makers. So, a few observations. <br> tldr: the cycle is wrapping up. Better not build startups right now.</p> <p>What we’ve got:<br> 1/ infra. Here it’s simple: the ones printing money are building the most expensive place where that money will work. Overall the hardware re-leasing model is super low-margin. But when you’ve got ads behind you (where margins are insane), you can invest. Capacity is wildly short, honestly. Even finding a rack is hard (like, getting a b200 node for a year).</p> <p>2/ most venture money is poured into the big labs. The logic is simple. Training models is very expensive, like very expensive. Mistakes, accordingly, are also very expensive. We pour money where the chance of a mistake is minimal. None of the big labs do frontier research as their main focus. They scale what already works. Top up RL, data, and so on. Plus every lab is trying to get as close to the user as possible (Anthropic is going prosumer, like Microsoft once did — to me that’s the best comparison).</p> <p>3/ researcher comp can hit $10–50M a year, because not many people have seen how to train a 1–10T model. That knowledge costs a lot. Because mistakes cost a lot — see llama4. Hence VCs hope maybe they’ll make something useful, but also they’re the only ones whose chance of making something is above zero. Hence $100M seed rounds. And yeah, if anything you can sell them back to the places that make models (since hardly anyone can do it).</p> <p>4/ given the size of the checks, you get a situation where an average startup needs to show insane outcomes, because why give $10M and get back $20M when the numbers over there are 10x higher. Moreover, AI startups have this cute property: most companies have very low revenue retention. Meaning the typical cycle is roughly: a customer takes all AI services to try (all competitors as a bundle) and picks 1 or even zero, deciding to build their own. Your revenue in posts grows, but everyone understands that in a year you’re fucked. This is a real situation. Notice how lovable and co went quiet. Yeah, because the growth changed and it’s not enough anymore. And investors initially gave you a crazy valuation.</p> <p>4/ and here the only solution is brewing. Grow revenue on paper ultra-fast and try to unload previous-round investors onto new ones — or better, onto retail — i.e., IPO. And that’s a dangerous game. And for those raising at $2–10–50B valuations there isn’t much choice. And at those stages it’s already harder to sell so everyone’s happy.</p> <p>5/ given that big lab revenue is heavily subsidized, an IPO will unload investors, but the market will look at it all and say “weeeeell not that great a business,” which will create a cooldown. And where do they pull money from, right — the riskiest segment, i.e. early-stage venture. And the odds that consolidation kicks off only went up. The big will get even bigger, and the small won’t have a chance (see pts 3–4).</p> <p>So there’s this meme going around the Valley that Ilya Sutskever’s SSI is a hedge fund. Because he knows exactly what’s going on on the ground, and he’ll definitely be able to make money on all of this. Such is life.</p> <p><a href="https://t.me/chillhousetech/1285?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, but seriously. https://darkolorin.com/posts/1284.html https://darkolorin.com/posts/1284.html Sun, 03 May 2026 22:47:27 GMT <img src="https://cdn4.telesco.pe/file/up9Gl_jAfY177kS2hd3_1Cdya33yn6Em2c6nQQEbbMpFbPS4GTFSYlIG47-rqoQSarqDbQhcq3TZjhG0LyHsu4bjJ_BcPRK03E5oVjhs6Q91SdPUybBu77joSuPXgmBmM0dEA2Saku0z0BYidB7p_xpln_U9nztDbfU0-Q9OiTBYhSSQR6KWzwe4AXEkRJuMz9rPkXHj2_D1ePXpKVti1naaSr-_eoo7yOytnqt9BXdIk1wvrT8__V3EQcyuvJlOFJ4-fbtq26wks4QewRzPnlSr1AiVPXWmqrNCnznVbyk8KVr_2sbd8oWdK6BUwc2NvtETeLQCr2j_3mrB6zplcA.jpg" alt="" loading="lazy"> <p>Well, but seriously. I think that, overall, between the AI &amp; software industries you can basically put a conditional “gets absorbed by” sign. All software is AI. A tech startup today just can’t not use ai. That’s how it is. That’s the tldr after talking with big funds and companies.</p> <p><a href="https://t.me/chillhousetech/1284?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Anthropic $44B ARR https://darkolorin.com/posts/1283.html https://darkolorin.com/posts/1283.html Fri, 01 May 2026 14:41:14 GMT <p>Anthropic <br> $44B ARR</p> <p>Links from the original post:</p> <ul> <li><a href="https://newsletter.semianalysis.com/p/ai-value-capture-the-shift-to-model?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://newsletter.semianalysis.com/p/ai-value-capture-the-shift-to-model</a></li> </ul> <p><a href="https://t.me/chillhousetech/1283?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m always kinda amazed by the dudes who write “like, through our service a… https://darkolorin.com/posts/1282.html https://darkolorin.com/posts/1282.html Tue, 28 Apr 2026 15:57:04 GMT <p>I’m always kinda amazed by the dudes who write “like, through our service a gazillion websites or apps or decks got created,” so that means we’re riding high. <br> TechCrunch writes that the number of apps in the store is growing at an unprecedented pace. <br> And I’d be asking a different question — how many of those apps have users over the long haul. And will anyone keep using all this going forward.</p> <p>Links from the original post:</p> <ul> <li><a href="https://techcrunch.com/2026/04/18/the-app-store-is-booming-again-and-ai-may-be-why/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://techcrunch.com/2026/04/18/the-app-store-is-booming-again-and-ai-may-be-why/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1282?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Digest of “promising” startups YC is looking for this summer. https://darkolorin.com/posts/1281.html https://darkolorin.com/posts/1281.html Tue, 28 Apr 2026 01:44:26 GMT <p>Digest of “promising” startups YC is looking for this summer. Though btw we talked to a partner of theirs and he was kinda like “yeah we’re looking for young ambitious people and generally that’s it.” Well besides, apparently they also want something from the list.</p> <p>The usual vertical AI solutions (doubt it’ll survive tbh, unless it’s strong enterprise) for personal healthcare, even agriculture, all those corporate brains (so that’s what the internal wiki is called now, apparently).<br> As usual, space data centers and supply chains.<br> Weird categories like “startups that sell to very big companies.”<br> Still, as a market snapshot it’s useful to read.</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.ycombinator.com/rfs?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.ycombinator.com/rfs</a></li> </ul> <p><a href="https://t.me/chillhousetech/1281?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m subscribed to Dwarkesh’s newsletter, who’s probably the top1 AI podcaster… https://darkolorin.com/posts/1280.html https://darkolorin.com/posts/1280.html Sat, 25 Apr 2026 23:41:20 GMT <p>I’m subscribed to Dwarkesh’s newsletter, who’s probably the top1 AI podcaster right now. So, he’s hiring a researcher for his team so that, like, he can not only ask questions on the podcast but also answer them (idk, sounds more like flexing, but still). He put up $20k in prize money for a hiring contest and posted questions like these (translated with ChatGPT with my edits for better digestion of the material)</p> <p>1/ A couple years ago there was this idea that AI progress could slow down as the field leans more and more on RL (reinforcement learning). First, because as the task horizon gets longer, the AI has to do work for many days before we can even tell whether it did it right, and if we’re still living in naive policy gradient land, then reward signal per FLOP goes down. And they already scaled from GPT-4 to o3, it’d be hard to keep going at the same pace. And yet AI progress is still fast and maybe even accelerating, if you believe the rumors about Spud or Mythos. What’s going on? What did the earlier intuition miss that led to longer timelines? (You can dispute the premise of the question itself.)</p> <p>2/ What’s the most plausible scenario where companies building foundation models actually start making money? If you treat each individual model as a company, its profit could pay back the training cost. But if you don’t immediately train a bigger and more expensive model, revenue stops coming after 3 months. So when does profit even begin?</p> <p>Maybe at some point scaling hits a plateau, but if frontier progress slows, then the combo of distillation and low switching costs (and cloud margins are to a large extent held up precisely by high switching costs) makes it really easy for open source to catch up to labs, eating their margin.</p> <p>So how do labs end up making money?</p> <p>3/ After OpenAI’s new funding round at an $852B valuation, the OpenAI Foundation’s stake is now worth $180B. Anthropic’s cofounders promised to donate 80% of their wealth.</p> <p>But people don’t have a clear understanding of how to productively use hundreds of billions (and soon trillions) of dollars to “make it so things go well with AI.”</p> <p>If you were running the OpenAI Foundation right now, what specifically would you do? And when?</p> <p>It’s not enough to just name an important cause, you have to explain how exactly to convert money into real impact. Describe a concrete strategy.</p> <p>4/ What should countries do that currently aren’t in the AI production chain (semiconductors, energy, frontier models, robotics) so they don’t end up completely bypassed by the dudes pushing AI forward?</p> <p>You can drop answers here (remote role is approved btw, not just SF)</p> <p>Interested in hearing my answers? But writing 1k words like he asks, I probably won’t, kek<br> 🙂</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.dwarkesh.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.dwarkesh.com/</a></li> <li><a href="https://airtable.com/app8aYOTzMkv9qeAJ/pagHhju8B5tgu4yXc/form?utm_source=substack&amp;utm_medium=email" rel="noopener">https://airtable.com/app8aYOTzMkv9qeAJ/pagHhju8B5tgu4yXc/form?utm_source=substack&amp;utm_medium=email</a></li> </ul> <p><a href="https://t.me/chillhousetech/1280?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> It came out funny. https://darkolorin.com/posts/1279.html https://darkolorin.com/posts/1279.html Fri, 24 Apr 2026 09:00:15 GMT <p>It came out funny. A lot of people thought about how AI will democratize capabilities and how easy it’s become to do something. But if it can do it, then it’s also easy. So the question that turned out to be more important: is it even worth it?</p> <p>And that question is way harder. Before, only a few could just write decent software as a team. Now the class of problems has grown, and it turns out the world of IT (or, in the Western way, tech) with AI’s arrival has actually moved into an adult phase. Basically like auto manufacturing in the 60s. Now it’s clear how to make it, but should you, and how do you figure out whether you should.</p> <p>Stratification has grown even more. I think a similar situation will happen with content.</p> <p><a href="https://t.me/chillhousetech/1279?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well let’s try https://darkolorin.com/posts/1278.html https://darkolorin.com/posts/1278.html Thu, 23 Apr 2026 18:05:08 GMT <img src="https://cdn4.telesco.pe/file/dWSdxJDohzE1mmmENytYiawTCz-vf063F5vT_jZ_CINbGDmw2xmwOSd1wPDm1o0llyCjMawqlVlyXdVTGj6sEamtSj58Z1Mrqm_s1chsWlIqxW1Ukfh6vjymFA1Lop6QVpyqwlJMv9FXpehAlR_zk58rVBRvHXlfkaTXT0RgSgIZvrwLaARTwunbbI-6a202219pzkcjn73_UBODQFHysEgz-1oWqwNQRXrrTAzyUFGbZgcLjBmVSxttf0gopx1MIheJmTfuLNTPUN1gC5POCXmferhITLWFOolumMOAcCDgNfHNMpwH6bji_w46UhSwUDPakIF4c5X8QUp3MvbWOA.jpg" alt="" loading="lazy"> <p>Well <br> let’s try</p> <p>Links from the original post:</p> <ul> <li><a href="https://openai.com/index/introducing-gpt-5-5/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/introducing-gpt-5-5/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1278?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Quarterly poll. https://darkolorin.com/posts/1277.html https://darkolorin.com/posts/1277.html Thu, 23 Apr 2026 11:18:10 GMT <p>&quot;&quot;&quot;<br> Quarterly poll. Do you vibe-code? And if yes, give the average number of hours a day, like, roughly</p> <p>- Yes, but I don’t know how many<br> - Yes, every day 1–2 hours<br> - Yes, every day more than 2 hours<br> - Yes, every day more than 4–5 hours<br> - Yes, every day more than 6–7 hours<br> - Yes, but like a couple times a week for an hour or two<br> - No<br> - 🥳<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/1277?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> AGI is coming for us. https://darkolorin.com/posts/1276.html https://darkolorin.com/posts/1276.html Wed, 22 Apr 2026 14:43:39 GMT <img src="https://cdn4.telesco.pe/file/ebEVMBJ0ocA8Ekfeyqk32OsgZvRUyXm6pOTxGv86EjWGvXTrD6ux1fKw2uR8xM48zmWBKeZsaZLMNJqUNEBbUoF62ueThdelpqmdKaUx4SXcFbxH27woF9DBpjt7cV-OSZwet0UrcBPsanBFBnHGE6EVwq75I5kT9gu2Hc0g6HkbHnk1A3XoZv5fopHuW068AUtCuQeHgXn7NSGu7M2T-pk7Xu2ZHajfDRQk4AtyJyIJRSVKbYR0gT8_w9jC4zzo_fhregN6r8HvTp69STvJzgFg-CPyOrZnJ66rdcnBxAo2etTQmOk7ObOX3vL35zyringMD6cUD914GkFZ4o6NMw.jpg" alt="" loading="lazy"> <p>AGI is coming for us. With every iteration of working with a regular document, your document degrades as text quality-wise. <br> Paper</p> <p>Links from the original post:</p> <ul> <li><a href="https://arxiv.org/abs/2604.15597?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://arxiv.org/abs/2604.15597</a></li> </ul> <p><a href="https://t.me/chillhousetech/1276?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Cursor for 60 yards? https://darkolorin.com/posts/1275.html https://darkolorin.com/posts/1275.html Tue, 21 Apr 2026 22:46:49 GMT <p>Cursor for 60 yards?<br> Damn damn<br> &quot;&quot;&quot;</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.nytimes.com/2026/04/21/business/spacex-cursor-deal.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.nytimes.com/2026/04/21/business/spacex-cursor-deal.html</a></li> </ul> <p><a href="https://t.me/chillhousetech/1275?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, a new era at Apple — new CEO John Ternus (ex head of hardware) https://darkolorin.com/posts/1274.html https://darkolorin.com/posts/1274.html Mon, 20 Apr 2026 20:44:10 GMT <p>So, a new era at Apple — new CEO John Ternus (ex head of hardware)</p> <p><a href="https://t.me/chillhousetech/1274?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And where’s mythos then? https://darkolorin.com/posts/1273.html https://darkolorin.com/posts/1273.html Thu, 16 Apr 2026 14:37:30 GMT <img src="https://cdn4.telesco.pe/file/JLEnw0DogVFFuMU2WfPckRLbXFHd-_W1fZ2HIAdxfH5OVJLYlDVkgP1turGC5YvaBdMNq3fp_38tZa4JbrNsp9pegSlYLEMo0luAl9xu_xG5kpDYG7rzn8LRdARRxOgKbWsLJGufoQwVk7LDIgK3H_UV4auu1khxHbGDP5YFtTtY1evkngPl4Z_9WS43FL7_XRrt7jzr5iijPuzZVM-8NF9xffNB-Epuf6d__cTqw72E1ig7VVfbznM9SAHg4G3xUI92Aef9cS0XxJ-XCegfes7k5wTBgNFrtf5y1jDtW-JiSseRA1F3UE6YS5qU2W-tYmmhr9lD05t74BiGu8B_LQ.jpg" alt="" loading="lazy"> <p>And where’s mythos then? Well, and I hope they won’t cut down 4.7.</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.anthropic.com/news/claude-opus-4-7?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/news/claude-opus-4-7</a></li> </ul> <p><a href="https://t.me/chillhousetech/1273?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Came back to angel investing the other day, in a scout role for a16z by the way. https://darkolorin.com/posts/1272.html https://darkolorin.com/posts/1272.html Thu, 16 Apr 2026 10:54:15 GMT <img src="https://cdn4.telesco.pe/file/Uon-OWs1vUYCVzmzI5WApDqrzPscQFAMalOas5n28uieltJEp9j1ULkXdv-ya3-o9KIqdauZWSmyCZJSs9C66OTzmVg-q72JvrfeTKJX1QvVHemy6345pX7lyNZEhWlcIH6kFJLMytEXUARlSr8top91zv3Mq1FAW1s2jm2_fv8Mm226e6mUs41zNtQUkSC1rt7YpFX0gO95pKrFV_KMcOeyvx9OqQ23YSIGeZ8HHU0dkQNErqCEZwRP1fNnQRUHz41Xn_Tz9SGc7hgBCJu6USghnvUHiNKqvQ0DOLWLVEoeYscYaw2JaGkprUSQnJlj_ScDQ5PQiCaiAk35tmAy1g.jpg" alt="" loading="lazy"> <p>Came back to angel investing the other day, in a scout role for <br> a16z <br> by the way. <br> If you’re doing something cool in:<br> - ai frontier research <br> - ai infra <br> - ai application layer <br> I’ll be happy to take a look at the deck or hear the pitch. And if it’s all top, then write a check and intro you to the fund’s partners :)</p> <p>Links from the original post:</p> <ul> <li><a href="https://a16z.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://a16z.com/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1272?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There’s no bubble, what are you talking about. https://darkolorin.com/posts/1271.html https://darkolorin.com/posts/1271.html Wed, 15 Apr 2026 17:47:08 GMT <img src="https://cdn4.telesco.pe/file/ZWOgO-aJUOaH_uQA_VAbcH-eS7Z4ZFGWTjrKmeI_IgxIM22AalV705hSLEN_yz7GDKJOtfEirByPMMDE28LFl1kvhqvX6yCBSKT5jVRYvSUsjf28mzOCwlwaaNCBevRL64nHak2s4xqivHxg42OUYMW82-NpabAATIXEvv4UeVTMPimoLIN6iAxfD8WIojW3pQpLPWLmIRda6tyWDoXzZ8wYFovcsqAaWl38RW7J7dpKZHndBAY1YGWosPNH0eVaYO3ezKZhccD0evXE2jwo7c2DrAghNDrRmG1-hEDOc9wBs_YIEmuL_8tDIS7TvbuUW8-U0xrbcOcAFNAMpeBgfg.jpg" alt="" loading="lazy"> <p>There’s no bubble, what are you talking about. Used to make shoes, now we’re selling video cards <br> 😏</p> <p><a href="https://t.me/chillhousetech/1271?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Out of the interesting stuff. https://darkolorin.com/posts/1270.html https://darkolorin.com/posts/1270.html Wed, 08 Apr 2026 19:54:58 GMT <img src="https://cdn4.telesco.pe/file/IJBoI2HG3Quj7cHZgXtt02-OKwb4isdHKs1JpGwPlyBoPcS3LpbmPGu5KI3F9BCnLyHMKF6xAbRI-dzlkOjFTPWCphjVNqF2SF-C3w6G3SLyt9PUc-BhMmW90awTXWor1vxF0erBoHeXwTYU_W3kyqGieFq8u4CZc56uXZzD_KoWlSsHrFqQyyvKAQLsdGQ2so5yKVxFQc3mR27wUYPBsb1K4ucaCU5hGYyPGHGRNiyRG9XTson5obRNTH61lUphQogN1kcB3SP6rkJD70Y73pbgxN56Knel8Gl9z4Pd-YIosNhkveGt9gADMZWcOMVZ6jnXGIBa59CxDcfmiGcIig.jpg" alt="" loading="lazy"> <p>Out of the interesting stuff. AI in general is really massively affecting how you and I start doing different tasks, and not just with the help of AI itself. <br> From the mid-1950s to 2016, people played Go at roughly the same level. But as soon as a model showed up that beat the strongest player, the moves that started appearing turned out to be more diverse. In other words, active development of the game and the sport kicked off. And a lot of the ideas weren’t the ones the model used, but totally different ones. <br> Not a new paper, but really damn cool. Basically, AI gives more room for absolutely new ideas about how you can do certain tasks. Love it. <br> The chart shows the quality of decisions made in Go.</p> <p>Links from the original post:</p> <ul> <li><a href="https://arxiv.org/pdf/2311.11388" rel="noopener">https://arxiv.org/pdf/2311.11388</a></li> </ul> <p><a href="https://t.me/chillhousetech/1270?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meta finally cooked. https://darkolorin.com/posts/1269.html https://darkolorin.com/posts/1269.html Wed, 08 Apr 2026 16:38:16 GMT <img src="https://cdn4.telesco.pe/file/X-o491SzRNVGNd5nf0VbEgE67ZtNRGS3eIXxVBEzQy11glgVM4yzNsebQBh8ljwh65RSuOwlsNfHvTXs6HA-atDLaRZzReLCjsZiamoD_qQ3yE9RWBFEk7ZetADvzAn-l2GHRePygcwlqZAhHxQvvn9zqIlEtuoanWJoGYzqLshRrn9cfOH7-eaA0TurbZy1MPZ8rQKgisDhhkB6jGfaup-HfulTcnQIktkhfcZVubdNKrZyQnagUgQOAuDONNI82MlJmtmaZhJyy5RSoQrgHXrg6QOTzl2aY0YziM5iq7lr14EM_m58r2zwDINA3V4kKAVbNgXGHtC_P1fhCM51rw.jpg" alt="" loading="lazy"> <p>Meta finally<br> cooked.<br> Of course not about coding. And not top everywhere. But it’s something already.</p> <p>Links from the original post:</p> <ul> <li><a href="https://ai.meta.com/blog/introducing-muse-spark-msl/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ai.meta.com/blog/introducing-muse-spark-msl/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1269?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A card for the new version of Anthropic’s model just dropped. https://darkolorin.com/posts/1268.html https://darkolorin.com/posts/1268.html Tue, 07 Apr 2026 19:50:53 GMT <p>A card for the new version of Anthropic’s model just dropped. <br> What I’d want to underline as a trend and a few basic things:<br> - new models will get better and better across a larger number of use cases within one domain (they keep pouring in data and RL), they poured in stuff around vuln hunting and found a lot of dangerous things (it’s important to understand that solving tasks like that is a consequence of RL — we teach a specific approach and then the model can do its job better)<br> - this is horizontal scaling. Meaning models improve less for long-horizon tasks, but do small tasks better (within the context). Pretty much all the benches they post are tasks in a 128–256k token window.<br> - I’m expecting an ever larger class of well-verifiable tasks to get solved<br> - given this kind of RL progress, we’ll need a shitload of tokens, and even a 40–60% margin on inference is peanuts, you need 99%+ (ideally many hundreds of X in cost reduction)<br> - the new model seems to eat fewer tokens judging by some benches, which is good, compression methods will most likely be baked right into post-train</p> <p>Links from the original post:</p> <ul> <li><a href="https://www-cdn.anthropic.com/53566bf5440a10affd749724787c8913a2ae0841.pdf" rel="noopener">https://www-cdn.anthropic.com/53566bf5440a10affd749724787c8913a2ae0841.pdf</a></li> </ul> <p><a href="https://t.me/chillhousetech/1268?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I think the main problem in the AI business right now is driving down the cost… https://darkolorin.com/posts/1267.html https://darkolorin.com/posts/1267.html Tue, 07 Apr 2026 17:25:58 GMT <p>I think the main problem in the AI business right now is driving down the cost of inference and training. Everything else is secondary. Claude code, and as a result open claw and all the other coding agents, are heavily subsidized — I think the real cost is dozens of times higher.</p> <p><a href="https://t.me/chillhousetech/1267?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Back after the trip. https://darkolorin.com/posts/1265.html https://darkolorin.com/posts/1265.html Sun, 05 Apr 2026 14:32:30 GMT <img src="https://cdn4.telesco.pe/file/Hlg2zivCi9vBSqgyBI5pFbAYneIyNm1RL6FhcUxfRV0wdVjg8mMW8fPAbMVWLShH8ZKeCVDDDGhqhqwEl82lz7G08W6sYkmYEVlEQ9A12ejfS4GjthaLEhER8gncnoJa0Gn6szAofysNEcmE_J3wA0uVU_IHoepOpKliS75_U5eI8owTQy0EfIyXrU06Dx33wcIz8qeRqN2LMtdXlKCT6CWWVudxIZroC6IkhBnn_M6UrK7dlRGwJczDXI46IvZWxnFNFP2sGNI5w7eHBC3HvUae6ApRedltw4UgkyUES7s5NU9lfSNDxHEnxljIDjeJSPxFT2OBUPyBxazHTSEelQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/Ww2gn7o_32bnzCbbfqMYd15ablaQVNGrwWEPXUf-zYjnkneidsnX3ct1KiI_9H9vMPwMTFAKKAB8c-YmgQ0yZfqwH-o0bgfQsoO9fgglzL1jRTXkyDMQxR_65PDBHx-GiNeNnmMCh9LCOSFaU28yehLau8JnzbNs2puoMpqe1slRdnalGMwviGclbegHOV5Q-HunPJG6sD-dPYs5mP4QQ9HpvQhTHoyQFxl_ZrytXERw9oSJZmRJ1vISc0tVKBAucH_HrrHlv18waGg8QK_W_A1PYkB7dTFD1C9YeX5fmr_wrAr9Hi3oBn1k0GOLvBg1ek8NwetreZ1sJyPZLeuRhg.jpg" alt="" loading="lazy"> <p>Back after the trip. If anyone’s interested, for pics from Bolivia from 5k up drop some kind of reaction. Otherwise, just Sunday stuff. The best Apple ad.</p> <p><a href="https://t.me/chillhousetech/1265?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Everything you can vibe-code in an hour or two will lose its value in the next… https://darkolorin.com/posts/1264.html https://darkolorin.com/posts/1264.html Sat, 04 Apr 2026 01:07:04 GMT <p>Everything you can vibe-code in an hour or two will lose its value in the next couple years? Because it’ll get generated on the fly in an assistant by itself or some other way.</p> <ul> <li>Yeah, anything that gets coded fast and one-shotted will lose value on its own</li> <li>Not quite, distribution will still decide and who gives a damn how it’s made</li> <li>Not quite, the value will be with the project’s author</li> <li>Of course not! It’ll just have 1–2 users</li> <li>🥳</li> </ul> <p><a href="https://t.me/chillhousetech/1264?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> From the wonderful discussions https://darkolorin.com/posts/1263.html https://darkolorin.com/posts/1263.html Fri, 27 Mar 2026 20:11:32 GMT <img src="https://cdn4.telesco.pe/file/WWv5lHYTS-YPpJ7DKTYtxr8TojMZiKnB_zKbq2izpPabObJnP0zRp6drKCXnEgokmc1l1dNIsHaHnLIYyOrqFgUQVZ9y7x17OAVP7zgmojIyBnTFk0oEJ_9tMtfGxJoIMctoqg-dDHpQJqAac3LXa7TUo9e-i1xh9dFsIwJCUP9LBFWgniVmkeVePmoDlkFYY-DC3QRv2puWQhcaSSbKnW_BG3760rqORmwsfh-0icIs9BX6v984lNk2zcGcRxPkDr2WXBmgIV41JiQeX5EsmN80SZD5C70rOpRnwcGmj5CAX7o1iiI3yvRAcMWXn8vcnalqzHO2CnYvMcnUD-0QMA.jpg" alt="" loading="lazy"> <p>From the wonderful discussions</p> <p><a href="https://t.me/chillhousetech/1263?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ll throw in my two cents about this mass obsession with AI “things that, of… https://darkolorin.com/posts/1262.html https://darkolorin.com/posts/1262.html Wed, 25 Mar 2026 23:54:22 GMT <img src="https://cdn4.telesco.pe/file/ibjbpturfkYa8TmLOVG2f1Wj5HSLcpoxYUO8LHjy3gM6qMu955EwrEv5X4Fsc1Bi7VR98lcEDIfYX9y6X4Gm9Zac0ylBsiE51KvdnWceEKIXbhd5-fMdHxba5V6oVao4TES2OeeD5VOqPs7IvrRgEKdiMtGIwRv6zf_GrRLnyQ0x1gR4vCr3bzqD8lJRGE79HePwZDTno1qkHrPuOksXcG-OhfLMajB_K_Z_UXa6F3zD1f4R8COy00T6OV_LJ1w32XKi_DWUNgzbu__Ew12X3sRGrzC8JHTyw9K9SSyngHBUJZbP3IkoiMUOQFIH3-wBVeNK6fpWwntWlHvFk57XLQ.jpg" alt="" loading="lazy"> <p>I’ll throw in my two cents about this mass obsession with AI “things that, of course, make me 100x cooler and I’ll finally make money.” <br> 1/ as of today AI tools are a straight extension of their owner—if you lack knowledge, you’ll lack it with AI too. And the point isn’t that you can’t get a site you couldn’t make before, the question is what kind of site and what’s going to be in it. So yeah, the essence of projects is a direct continuation of the owner. And if earlier your startup didn’t take off with a team of 10 devs, it won’t take off now with a team of 10 agents either, and it’s not even clear where you’ll waste more nerves. <br> 2/ AI right now—it’s unclear how to integrate it to get measurable benefit. Like yeah, we notice that experiments got faster, or parameter tuning. But that’s all personal efficiency, and how to increase the efficiency of the whole company is unclear. Remember that meme: one developer will do a task in 2 hours, and two developers will also do it in 2 hours. Same thing with AI + a human. And the problem is even worse when it’s a bunch of people and agents. <br> 3/ automation == optimization. And you optimize when scaling up by multiples has become really hard. And if you suddenly started optimizing a project with $10k/month revenue using AI—you’re clearly digging in the wrong place. <br> 4/ AI is still a tool you need to steer to the goal with your hands, which is, sure, a vibe, but we’re waiting for the moment when you won’t have to do that and you can trust a task end-to-end without manual control. I think in dev this will partly happen, or already has (only for certain tasks). <br> 5/ ARC AGI 3 dropped here, and among other funny stuff they added replays of how AI plays games. Pretty hilarious. Wonder how fast they’ll tune harnesses for this bench too. Because the whole point of benches is just to beat the opponent since a lot of people couldn’t care less about real tasks.</p> <p>Links from the original post:</p> <ul> <li><a href="https://x.com/arcprize/status/2036860087948763526?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/arcprize/status/2036860087948763526?s=46</a></li> </ul> <p><a href="https://t.me/chillhousetech/1262?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I was thinking that, like, OpenAI clearly doesn’t get how to make it on… https://darkolorin.com/posts/1261.html https://darkolorin.com/posts/1261.html Thu, 19 Mar 2026 22:11:59 GMT <p>I was thinking that, like, OpenAI clearly doesn’t get how to make it on consumer without coding (well, that’s what they say publicly), lovable went the same way into analytics and whatever else, and I think this year everyone’s gonna try everything and this is the beginning of the end for the AI application layer. Or it’s finally time to come up with something bigger than a chat.</p> <p>Like, you at OpenAI hired all the top guys on the market, tried a browser, and an app store, and ads, and whatever the hell else, even your own TikTok. And it’s all kinda pale. Not like a total failure, but it’s not enough to become a business. I’ll explain. The main use case of ChatGPT is: ask and get an answer. That used to be a subtype of search. But what if all your competitors are shipping features with already-built distribution (Google etc), and yeah, in the moment you won, but habit won and people saw those same features in the products they already use. What do you do?</p> <p>The browser is still Chrome, search is still Google, and so on. If today I can type the same question into the search bar as in ChatGPT and get comparable quality. Why do I need ChatGPT? Why all these pro/omega-pro versions. And if they do that across all verticals. Everyone will train a decent model in their domain sooner or later. So what’s the exit for ChatGPT? AGI?</p> <p>Or do we keep believing in OpenAI and that their models are unreachable in quality?</p> <p><a href="https://t.me/chillhousetech/1261?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, looks like that AGI thesis is getting canceled, or they’ve just stopped… https://darkolorin.com/posts/1260.html https://darkolorin.com/posts/1260.html Tue, 17 Mar 2026 01:51:54 GMT <img src="https://cdn4.telesco.pe/file/drhoq4mu1UyJM_WT0aI7lEq2ml8Nu9BoWxAo5fP2hmXo50tynI4QANGu-XwwXwGp0nYvuFYotFBjqS300NGFRCxYScKEyne6grH7S6dAL7uhMAOju5ucv60OPwAVI0yOmM7vITKry5UlFH3QwEPHBZq967nId_4PCF1xki4_32-vcKAXfGD7nreoARAZ19y82vmeuLVGljPuMgTbBa-GZ0sx1Au26Z6nE-3n9v1UIDGLrZkdErABN1fYZHi7kVNVrOFYHdz-1b7zcv7VUf-7Ry8SgNqwiAMePNkTKAnbcPrTGHhoHDNw3ggakpeCc5eeosz8d6iTWO3reMk9u_qcFg.jpg" alt="" loading="lazy"> <p>Well, looks like that AGI thesis is getting canceled, or they’ve just stopped pushing it seriously. But still, OpenAI is heading into direct competition with Anthropic on coding and in business/enterprise. That’s extremely interesting and kinda suggests that by doing everything at once they let a bunch of competitors get way out in front (Google, for one). And even more interesting is the fact that, feature-wise for the app, it looks like they still don’t have any ideas. Even with the openclaw acquisition and the rest. <br> So where’s that legendary ChatGPT AppStore? Haha.</p> <p><a href="https://t.me/chillhousetech/1260?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “read it, but didn’t fully get it” column: - how quality dies in nets from… https://darkolorin.com/posts/1259.html https://darkolorin.com/posts/1259.html Sat, 14 Mar 2026 00:24:20 GMT <p>“read it, but didn’t fully get it” column:<br> - how quality dies in nets from context length and various tricks layered on top,<br> old work,<br> but now everyone’s gone and is mercilessly RL-bombing this bench (longmemeval). So I recommend taking a look, so you don’t always fall for a 1m context window<br> -<br> cool paper<br> from a real mathematician about AI’s “successes” in science, so to speak<br> -<br> a burner phone<br> as a concept of an anonymous chat with AI<br> - found out what the<br> Sharpe ratio<br> is<br> - still don’t understand why anyone even talks about<br> this<br> - Karpathy’s autoresearch seems to have flown into every corner, tried running it on our kernels, got +2% speed (not like it’s interesting)<br> -<br> recommend<br> about China and the US</p> <p>Links from the original post:</p> <ul> <li><a href="https://research.trychroma.com/context-rot?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://research.trychroma.com/context-rot</a></li> <li><a href="https://www.daniellitt.com/blog/2026/2/20/mathematics-in-the-library-of-babel?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.daniellitt.com/blog/2026/2/20/mathematics-in-the-library-of-babel</a></li> <li><a href="https://openanonymity.ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openanonymity.ai/</a></li> <li><a href="https://ru.wikipedia.org/wiki/%D0%9A%D0%BE%D1%8D%D1%84%D1%84%D0%B8%D1%86%D0%B8%D0%B5%D0%BD%D1%82_%D0%A8%D0%B0%D1%80%D0%BF%D0%B0?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ru.wikipedia.org/wiki/%D0%9A%D0%BE%D1%8D%D1%84%D1%84%D0%B8%D1%86%D0%B8%D0%B5%D0%BD%D1%82_%D0%A8%D0%B0%D1%80%D0%BF%D0%B0</a></li> <li><a href="https://news.ycombinator.com/item?id=47334694&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://news.ycombinator.com/item?id=47334694</a></li> <li><a href="https://danwang.co/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://danwang.co/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1259?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Back to the news. https://darkolorin.com/posts/1258.html https://darkolorin.com/posts/1258.html Sat, 14 Mar 2026 00:12:37 GMT <img src="https://cdn4.telesco.pe/file/Cb0JPQU6Sqvxm-Pgt-Ft8ulS3ahtWQWgH5_N0lMv8gf1KIv9UkYzIwT51kr_OR9ukxJg6r9J_yVGZoW5AM3QzablYkxaaWhEvmXGZlPJInrsDB9-aqeo5jeT4tuIbiTgXU5i7fUCutyqNqNOAslVKfYoZw6x1vRL2nRwt7tvVcHbM4GZI7yNIJnKbAUe6JJpvF2JXujA96ZL1lXJEGHm1DF0vF52Aye0-qQofbT5JmcFKc5P-ZMwo9oiCP8fH_kTp2JLoOkT0UQ0FsK55KnbZAUMsdjKDrv0ooY0y4vF9_xo1-vUI4pHVD-UE7cOsoFcShZFFyelBJaPepgFnZLyQg.jpg" alt="" loading="lazy"> <p>Back to the news. The trip to the Valley was productive. Barely had time to write at all, but this one was hard to ignore. More and more new models keep coming out!</p> <p><a href="https://t.me/chillhousetech/1258?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> We live in a unique time where just knowing what you’re doing is already a… https://darkolorin.com/posts/1257.html https://darkolorin.com/posts/1257.html Sat, 28 Feb 2026 18:41:18 GMT <img src="https://cdn4.telesco.pe/file/P6A1nbgq7OtljMMSwzyi135AinJCqAdoV0vOXd86fuBrKTEYEEo4JuQejoboKxxNUcvaNXbrXpqL9ltOO0rqMlrBP4-Kivf9_Gtql43f04Nqk2VNdGDjzsuMXpyoM5vwhTzDtF1g4WCsbW-LvGiqrkj9uQxSXXMVKbq_h_BwwCOAPRIfN3ex5icQU-Y1ZQSdg6kp2ZTiyaQDHWwA_MKntoTWhFrwVv-Cv2eHVdnUEyNTvxQ_c2tpichz5ezDVJALOB84igVmV90RJFVdJxqHi5Qf0R-QeLLgVbcOi64pBNNhLlOegdl1yLP7wyqkj2V_-gGl7JnOA0omC4hUmN46dw.jpg" alt="" loading="lazy"> <p>We live in a unique time where just knowing what you’re doing is already a skill.</p> <p><a href="https://t.me/chillhousetech/1257?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Jack Dorsey fires 4,000 people because of AI, meanwhile that same guy in… https://darkolorin.com/posts/1256.html https://darkolorin.com/posts/1256.html Fri, 27 Feb 2026 17:17:24 GMT <p>Jack Dorsey fires 4,000 people because of AI, meanwhile that same guy in September ’25 is throwing a party costing 68M bucks. <br> Yeah, AI is obviously a great excuse these days to cut people when you’ve got problems of a completely different kind.</p> <p>Links from the original post:</p> <ul> <li><a href="https://x.com/bulltheoryio/status/2027250361816486085?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/bulltheoryio/status/2027250361816486085?s=46</a></li> </ul> <p><a href="https://t.me/chillhousetech/1256?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Not a bad round at 110 yards https://darkolorin.com/posts/1255.html https://darkolorin.com/posts/1255.html Fri, 27 Feb 2026 14:27:47 GMT <p>Not a bad round at 110 yards <br> <a href="https://x.com/sama/status/2027386252555919386?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/sama/status/2027386252555919386?s=46</a></p> <p><a href="https://t.me/chillhousetech/1255?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If anyone missed it, a new doomsday report is out Citrini Research https://darkolorin.com/posts/1254.html https://darkolorin.com/posts/1254.html Fri, 27 Feb 2026 01:48:01 GMT <p>If anyone missed it, a new doomsday report is out <br> Citrini Research</p> <p>Citadel critique <br> (this is a critique of that scenario above)</p> <p>TLDR: some people think all AI agents will replace all of us and the collapse of economies is unavoidable. <br> The success and spread of AI as a technology may lead not to normal economic growth, but to a structural crisis. <br> Others are simpler and look at history and current data and say, like, rollout is slow like it always is. Nothing scary, and we’re underestimating everything that’ll get done and that we’ll end up consuming. That’s how it is.</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.citriniresearch.com/p/2028gic?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.citriniresearch.com/p/2028gic</a></li> <li><a href="https://www.citadelsecurities.com/news-and-insights/2026-global-intelligence-crisis/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.citadelsecurities.com/news-and-insights/2026-global-intelligence-crisis/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1254?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One of my theories why openclaw and the rest blew up so hard, besides the fact… https://darkolorin.com/posts/1253.html https://darkolorin.com/posts/1253.html Wed, 25 Feb 2026 19:49:16 GMT <p>One of my theories why openclaw and the rest blew up so hard, besides the fact that they launched good models at the start of the year, is that everyone got to feel like a manager. Everyone wanted to hire themselves an assistant and give them orders.</p> <p><a href="https://t.me/chillhousetech/1253?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Have you, with the arrival of Claude code etc, started spending less time on… https://darkolorin.com/posts/1252.html https://darkolorin.com/posts/1252.html Wed, 25 Feb 2026 17:32:13 GMT <p>Have you, with the arrival of Claude code etc, started spending less time on engineering tasks? Like, up to getting the result you need.</p> <ul> <li>Yes, everything got significantly faster</li> <li>Yes, but not like *that* much faster</li> <li>Pretty much the same, a lot goes into steering it and understanding how to do it</li> <li>No, I just got confused</li> <li>I’M NOT AN ENGINEER, just vibe coding</li> <li>🥳</li> </ul> <p><a href="https://t.me/chillhousetech/1252?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A pretty good overview came out of what big labs like doing these days… https://darkolorin.com/posts/1251.html https://darkolorin.com/posts/1251.html Tue, 24 Feb 2026 14:54:28 GMT <p>A pretty good overview came out of what big labs like doing these days: namely, looking at big tasks in coding and the execution horizon. <br> My take is that quality of execution is just as important as the horizon, and one without the other is pointless. <br> tldr: as expected, coding agents are good at what they were trained on, and can’t do anything new. No ideas and no solutions to current problems. <br> More details</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.modular.com/blog/the-claude-c-compiler-what-it-reveals-about-the-future-of-software?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.modular.com/blog/the-claude-c-compiler-what-it-reveals-about-the-future-of-software</a></li> </ul> <p><a href="https://t.me/chillhousetech/1251?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> New opening with us Also for inference we’re always open to new people https://darkolorin.com/posts/1250.html https://darkolorin.com/posts/1250.html Mon, 23 Feb 2026 15:23:00 GMT <p>New opening with us <br> <a href="https://trymirai.com/careers/machine-learning-engineer?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/careers/machine-learning-engineer</a> <br> Also for inference we’re always open to new people <br> <a href="https://trymirai.com/careers/inference-engineer?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/careers/inference-engineer</a></p> <p><a href="https://t.me/chillhousetech/1250?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> YC will help your startup, apply to YC, same with their partners https://darkolorin.com/posts/1249.html https://darkolorin.com/posts/1249.html <p>YC will help your startup, apply to YC, same with their partners</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1249?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Finally something actually sane for finding an investor and contacts. https://darkolorin.com/posts/1248.html https://darkolorin.com/posts/1248.html Sat, 21 Feb 2026 16:19:15 GMT <p>Finally something actually sane for finding an investor and contacts. No spreadsheets, digging around, comments under “raise” posts, and other crap. Just a list of funds and partners by categories and stages. Use it and be happy.<br> <a href="https://f4.fund/raise?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://f4.fund/raise</a></p> <p><a href="https://t.me/chillhousetech/1248?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> AI editing back in the 2000s https://darkolorin.com/posts/1247.html https://darkolorin.com/posts/1247.html <p>AI editing back in the 2000s</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1247?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Recently in the channel someone asked “so how are things with Mirai.” Time to… https://darkolorin.com/posts/1246.html https://darkolorin.com/posts/1246.html Thu, 19 Feb 2026 16:10:19 GMT <p>Recently in the channel someone asked “so how are things with Mirai.” Time to share. We raised a $10M USD Seed Round. <br> A bit more detail in the post.</p> <p>Links from the original post:</p> <ul> <li><a href="https://trymirai.com/blog/mirai-raises-10m-to-build-the-on-device-ai-capability-layer?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/blog/mirai-raises-10m-to-build-the-on-device-ai-capability-layer</a></li> </ul> <p><a href="https://t.me/chillhousetech/1246?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Noticed how many AI content-creation services there are right now, and all of… https://darkolorin.com/posts/1245.html https://darkolorin.com/posts/1245.html Thu, 19 Feb 2026 02:49:28 GMT <p>Noticed how many AI content-creation services there are right now, and all of them promise insane conversions, views, whatever. On Insta you just have to like a video or a reel made by a grid once (they’ve already started doing series there) and you immediately get the scale. So yeah. Simple question, or the saturation problem (the reason why infobiz and all that rarely works at scale): what happens if the entire target audience starts using the service? (All bloggers, everyone shooting reels, everyone editing images).</p> <p>The answer pretty quickly takes you to the next step in the strategy for building a product like this.</p> <p><a href="https://t.me/chillhousetech/1245?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> BBBRRRREAKING: Japanese company TOTO, the one that makes toilets, suddenly… https://darkolorin.com/posts/1242.html https://darkolorin.com/posts/1242.html Wed, 18 Feb 2026 20:45:33 GMT <img src="https://cdn4.telesco.pe/file/kzmcQVVvIARP3gNrc752rqh5at3aJrzfH37ImnWKeyY4LGgE1WBg9gtoY9c5K6_KAVBAkdOkMvctvdjbE3SHAbB1xGXy5uppFO84E_NAsDh4DAeXqcLS1IkPXYiH7_fjuJ1LYv92zblcq8WZeGUo_juNvBfw0MWJ_Ttc9LDmVgc_1b-WwZfdgspPV_OX_-1MnRJua4Wt_FC7M5y0qKHqcxou4wCPU9_5hOcz87QCywTIw1ENp3yIfgDYGZmYFcpCPv6gVbN4q3bOUAEh2Hc7u3OYKvhmFElnk_TrzeX2R2qjrrIhrMhdAFiGVgqJjFs2JXoCqpJOUafTilbG9LjAxw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/NFdA4UZHhQkbFEsfqrWG9Y75LDcji3uZQ_f7MDoY7H2_5gMV5yTucXZXujW6_JSkz19v6wTmZ0y0aiVgdhScx9xKJIuabUxNQjt31GuEO320zq96WzzcxBYXqGaKCrlb6hdWqR0sTESK8ksqMG_YuSu5zqyprfsCIgnWXdrq3RvfJb7vQHn-Gu0G0YDeGRBkQOh7zenMJSTL7z6xmPASmNTubUxQ2ZpT2Jlg62cT5XqAPGat_CMgiEK3EnpWgQRVAe1ymkSU_nY-Ek_OauSI7Ut4hYKmpfrvi8tJiyyvIQ5umuMXy0COrKZcw3QeyZeZLVYBpTjgZrwR98VwXdl49Q.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/R2rpKdFSZN3DIIZ97BbHV62G_WmvcAnQhyJES54WrxTdFHWjmBI21GvB-l3WWU_qgTRqGXw2Ma-D0ymAlUAQVShEaoEjd3oz074A2dc048U7_jYR1LkEzgC4Zuepe_DMLfdRUDl5PvPgieOtIgFA-KjIJMZoLfxl1TaOKOBPwdb23xlFnoHhR8zZU4EFC2ZV3Sy8YHe3lhgfPxT3Vs703burYXDc6z5rtKMqRwE4en6oGsrhyiy1Yr5oCsyJobXmWHwzc9lsRvdBRIJFO92g2EIfiEn0S3jXWGJDPv_yzKLFi_cHMU-uU8-sBVS2qmoGPk-FB3-t2IfU1vDpuC0xcg.jpg" alt="" loading="lazy"> <p>BBBRRRREAKING: Japanese company TOTO, the one that makes toilets, suddenly figured out that its ceramics tool can be used to manufacture cutting-edge AI chips. <br> Market — $60B. <br> Stock +60% on this news.</p> <p><a href="https://t.me/chillhousetech/1242?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Long story short I wanna add a bit of clarity to these rounds. https://darkolorin.com/posts/1241.html https://darkolorin.com/posts/1241.html Wed, 18 Feb 2026 13:56:58 GMT <img src="https://cdn4.telesco.pe/file/GwA-1U7GwlQuWQpeSaMiBczb5emVXkb0Jum-9naf3DMZm6YosuChcWYtXsKtvBsncCL_mchgiCSp0YEoj7kgLk3uDgS1YKmWQw-jbwguRcBxKW1-nK4O9l5Y4L_AJsWGnigZ7qMf2nQBIfGJT3cTQIL6IJ7Oy_2nAgz8ozNPZinpyfL3Oas32JZti7r6X0VHCjjGNQ3UfGakh6OzQGF3OQo9aUuXy0mj-BQJOY-6gfDiC-fMslA6JqCq4pLjj4SxVG-eESHnFQ9vKmb5FilQnXcSCZTV_bXHnqs3YRW4BohDNh5LMGeWhoV8SC3FYQDDsY8Pw6ObGWmVdf0oFu0VmQ.jpg" alt="" loading="lazy"> <p>Long story short I wanna add a bit of clarity to these rounds. Here we go again: top funds are investing a yard into a lab at a 4 valuation. And news like this is everywhere in general.</p> <p>The logic is roughly this:<br> - any fund doing rounds like this needs projects like Google, OpenAI etc. The kind that will define how a whole industry develops.<br> - they’re not interested in turning 100m into a yard. And none of their LPs (the people putting money into the fund) are either. They’re being sold a high-risk strategy and they’re buying.<br> - big AI labs are a pretty understandable bet given the mood among a lot of big-name research teams that the transformer isn’t gonna take us to the next level of AI.<br> - funds do everything based on already-lived experience, yeah, it worked last cycle so it should work now, this idea is close to a lot of people and sells well.<br> - sometimes details matter, but rarely; usually they pick the most famous or the super-ambitious team with a ton of references and dump a pile of money on them, and now you can’t really buy a company like that either, which means these smart guys will grind until something comes together or they return the cash (everyone’s fine with them burning half, that’s a normal risk profile).</p> <p><a href="https://t.me/chillhousetech/1241?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Now I’m almost sure people there don’t write code UPD: they fixed it fast https://darkolorin.com/posts/1240.html https://darkolorin.com/posts/1240.html Tue, 17 Feb 2026 23:51:54 GMT <img src="https://cdn4.telesco.pe/file/Y92VcrnyWkA0bdKasKs9yugYy6rbCe6HuyS_NQ5RXUSo7s3mzXes5KiHqVF2vTaPYk393pEH--Rz7lpnJE7fyD5YAjrAESRe34DhdYrpMKE8ZuFkJhxDlOFtX1Q3aHNGndzJ4NcLLMOYQ_wQeAyd3V2iMJHXY8h9rwL00rjFlwko04xZmjALTUAzqLHdamYTPWjEoNW-9HvxaUmPwydO6ASNrwQKSqh3BYGeyxhESTqelO2p4H_EcqcXWcxWJ-v8U6GrOFR5O2aMT-HN3pgA_bAM5q1Y5HSJfPjLq1fxIsVO7Qdfg-GwgFjPVNVt-mybhTPLAnkrFxqBC8pKqE5uQg.jpg" alt="" loading="lazy"> <p>Now I’m almost sure people there don’t write code <br> <a href="https://claude.ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://claude.ai/</a> <br> UPD: they fixed it fast</p> <p><a href="https://t.me/chillhousetech/1240?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A thought just slipped by here. https://darkolorin.com/posts/1239.html https://darkolorin.com/posts/1239.html Tue, 17 Feb 2026 18:16:04 GMT <p>A thought just slipped by here. Let’s say we’re all sure that every engineer will use an AI coding assistant. And will eat a lot of tokens, sometimes even thousands a month. So the cost of an engineer in a company has gone up a lot. Meaning you have to cut someone else to compensate, or some thin-margin businesses will die because of that? What do you think? Because not every company can afford an extra 1–2k per engineer per month. And right now, if you use this stuff really actively, those costs are there.</p> <p><a href="https://t.me/chillhousetech/1239?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Seed 2.0 Pro from Bytedance. https://darkolorin.com/posts/1238.html https://darkolorin.com/posts/1238.html Mon, 16 Feb 2026 06:12:01 GMT <img src="https://cdn4.telesco.pe/file/a2W3NK16g1-nooSpCFmLsofLUbtJCITgtW6Kng_TFVWKh1706jlcWB2TAyk5Tk05IPjlYA95KB05ibESRQa8Lbw2Q_tWafA94tEcDzvqImWR2Uzt66F08Udj2gTrYID5iU4WJSz5YNLxrsr-pkonpUvdskVR2Tk7XYnuoZ6zilPz9XzxQdGEsvlmMnRMSr4xLg0AcXUCqtd442ERPVNuYQaTXhaXF3JRDqeVmaKJSn3ONQWra1SSkE93rSmsiGFy3SL7GsaI0Vq6yRUaaHdQL5eLrx1j5fIYaAoRlEnHbSWZyamUckNGQ-t7i6BEkip9XqO8BlQpbk1yLgFjGfLtRA.jpg" alt="" loading="lazy"> <p>Seed 2.0 Pro from Bytedance. Not bad, not bad. Totally on par with the big labs from the US. You can safely say China has caught up (let’s wait till people actually use it there). Which means the AI world just got even more interesting. <br> Model card</p> <p>Links from the original post:</p> <ul> <li><a href="https://lf3-static.bytednsdoc.com/obj/eden-cn/lapzild-tss/ljhwZthlaukjlkulzlp/seed2/0214/Seed2.0%20Model%20Card.pdf" rel="noopener">https://lf3-static.bytednsdoc.com/obj/eden-cn/lapzild-tss/ljhwZthlaukjlkulzlp/seed2/0214/Seed2.0%20Model%20Card.pdf</a></li> </ul> <p><a href="https://t.me/chillhousetech/1238?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The guy who made openclaw “joined” OpenAI. https://darkolorin.com/posts/1237.html https://darkolorin.com/posts/1237.html Sun, 15 Feb 2026 22:00:51 GMT <p>The guy who made openclaw “joined” OpenAI. A classic Valley move. <br> Tweet</p> <p>Links from the original post:</p> <ul> <li><a href="https://x.com/sama/status/2023150230905159801?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/sama/status/2023150230905159801?s=46</a></li> </ul> <p><a href="https://t.me/chillhousetech/1237?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Came across a funny story here. https://darkolorin.com/posts/1236.html https://darkolorin.com/posts/1236.html Sat, 14 Feb 2026 22:02:23 GMT <p>Came across a funny story here. Draw your own conclusions.</p> <p>Clever Hans. A horse in early-20th-century Germany that, in 1904, was proclaimed “the first thinking animal.” Hans showed impressive abilities: solved arithmetic problems, “told” time from a clock, recognized playing cards, and even identified the composer of a piece of music. When answering questions, he’d tap out the required number of hoof-beats or pick an option from the ones offered.</p> <p>According to a historical review by Samhita &amp; Gross, at first observers assumed it was all a hoax, and that the trainer was somehow tipping Hans off with the right answers. But it was shown that the horse completed the tasks successfully even with the trainer absent. In 1904 a German commission for the study of language ran an investigation and concluded that Hans’s abilities were genuine, and that “most biologists, psychologists, physicians, experts of every stripe, and even ordinary people were convinced that animals are capable of thinking like humans.”</p> <p>It took another four years and a series of carefully designed control experiments by the psychologist Oskar Pfungst (1907) to figure out what was going on. Pfungst showed that Hans couldn’t give the correct answer if the questioner himself didn’t know the answer, or if the horse couldn’t see his face. He concluded that Hans didn’t have the claimed cognitive abilities at all: neither counting, nor arithmetic, nor “reading” cards.</p> <p>But Hans turned out to be a genius at something else, a skill far more useful for a horse: the social ability to read barely noticeable facial signals from a human to understand when he needed to stop tapping his hoof. These micro-signals weren’t under the conscious control of the questioner; even when Pfungst tried to suppress them, the horse kept successfully picking them up.</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.tandfonline.com/doi/epdf/10.4161/cib.27122?needAccess=true&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.tandfonline.com/doi/epdf/10.4161/cib.27122?needAccess=true</a></li> </ul> <p><a href="https://t.me/chillhousetech/1236?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Here’s an important little topic. https://darkolorin.com/posts/1235.html https://darkolorin.com/posts/1235.html Sat, 14 Feb 2026 02:09:26 GMT <p>Here’s an important little topic. OpenAI said<br> they made an actual discovery in theoretical physics.<br> Which kinda, you know, hints like “welp, AGI is here.”<br> I couldn’t walk past it (I love physics and even messed with it a bit at total noob level). Decided to read more closely what’s going on there.<br> We’re not gonna dive into what MHV amplitudes actually are. But let’s note a few things:<br> - most of the work on this topic came out in the 80s and a lot of results were already obtained<br> - GPT 5.2 was “cooked” in a special way to help solve this task and it thought for 12 hours<br> - the net really did propose an original simplification of the formula based on other results<br> - the net was prompted by people, and not just people, but very big-name scientists in this field<br> - it’s important to note that specific formulas were derived earlier, they needed to be generalized, and then humans also checked and proved the solution found was correct<br> - knowing what can be proven, you can guess how they pulled it off via the net, through RLVR or similar mechanisms like<br> RLM<br> (even without knowing the result you can check whether you got something correct and reward the steps that lead to something you can check), I’m not a specialist here, but maybe someone who knows their stuff can say if I’m thinking right<br> - there was no review of the work</p> <p>You can read the comments and even find comments from the author of the work<br> here</p> <p>What’s notable from my little bell tower:<br> - it feels more like they sicced brute force on a specific crisp task, i.e. the task formulation was clearly harder than its solution (in the sense that solving is extremely hard, but possible since you can verify)</p> <p>A very important comment from the paper’s author<br> Hi I am an author of the paper. We believed that a simple formula should exist but had not been able to find it despite significant effort. It was a collaborative effort but GPT definitely solved the problem for us.<br> And then this one<br> There are closely related &quot;MHV amplitudes&quot; which naively obey a really complicated formula, but for which there famously also exists a much simpler &quot;Parke-Taylor formula&quot;. Alfredo had derived a complicated expression for these new &quot;single-minus amplitudes&quot; and we were hoping we could find an analogue of the simpler &quot;Parke-Taylor formula&quot; for them.</p> <p>So for now it’s cool, but not scary yet, still. Reminds me of the invention of packages like Mathematica and how they got used in the scientific community. Of course in this case AI pushed the bar to a new level.<br> Well then, are we waiting for a warp drive?</p> <p>Links from the original post:</p> <ul> <li><a href="https://openai.com/index/new-result-theoretical-physics/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/new-result-theoretical-physics/</a></li> <li><a href="https://www.primeintellect.ai/blog/rlm?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.primeintellect.ai/blog/rlm</a></li> <li><a href="https://news.ycombinator.com/item?id=47006594&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://news.ycombinator.com/item?id=47006594</a></li> </ul> <p><a href="https://t.me/chillhousetech/1235?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If anyone hasn’t pinned it yet, the “country of geniuses in a datacenter” is… https://darkolorin.com/posts/1234.html https://darkolorin.com/posts/1234.html Fri, 13 Feb 2026 17:33:16 GMT <p>If anyone hasn’t pinned it yet, the “country of geniuses in a datacenter” is supposedly waiting for us in 1–3 years, basically. But rollout will take like 5 years, or maybe 50. Who knows those geniuses.</p> <p>New interview: Dario from Anthropic with Dwarkesh.</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.youtube.com/watch?v=n1E9IZfvGMA&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=n1E9IZfvGMA</a></li> </ul> <p><a href="https://t.me/chillhousetech/1234?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A pretty cool blog post came out, one I myself wanted to try writing. https://darkolorin.com/posts/1232.html https://darkolorin.com/posts/1232.html Thu, 12 Feb 2026 12:39:01 GMT <p>A pretty cool blog post came out, one I myself wanted to try writing. But I might be kinda meh as a coder and researcher. Still, on the level of vibes it’s a direct hit.</p> <p>In short: in the race to automate coding and, as many believe, then the whole of software engineering after it, there’s one detail. Today everyone is trying to brute-force improve the models, and yeah, that’s understandable. Some hypothesis of the last three years: like, model performance alone is enough and then everything will be awesome. But there’s always another problem sitting right next to it.</p> <p>Let’s call it the “harness” problem. Everyone’s building the fastest car. But where you’re supposed to sit comfortably — nobody’s thinking about that yet. Like, in code editing, just changing the approach to how we edit code and how we feed formatting into the model already changes the result a lot. In the article the guy put hashes next to line numbers and some models improved their benchmark results by 10x.</p> <p>Worth noting that there are benches from JetBrains on this topic too.</p> <p>And the jump in quality is insanely strong even for top models if you just change the basic format of interaction with code. AI isn’t only models, it’s tooling, it’s inference, it’s hardware — all together. And you can run everywhere or somewhere separately. Massive field for startups.</p> <p>Links from the original post:</p> <ul> <li><a href="https://blog.can.ac/2026/02/12/the-harness-problem/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.can.ac/2026/02/12/the-harness-problem/</a></li> <li><a href="https://arxiv.org/abs/2510.12487?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://arxiv.org/abs/2510.12487</a></li> </ul> <p><a href="https://t.me/chillhousetech/1232?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By Seedance 2.0 and Kayne East So, how do you like the quality of the… https://darkolorin.com/posts/1231.html https://darkolorin.com/posts/1231.html <p>By Seedance 2.0 and Kayne East <br> So, how do you like the quality of the generated video?</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1231?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh, found another interesting job posting. https://darkolorin.com/posts/1230.html https://darkolorin.com/posts/1230.html Wed, 11 Feb 2026 17:47:42 GMT <p>Oh, found another interesting job posting. First line of the description:</p> <p>“<br> Blue Rose Research<br> builds data and AI tools that help Democrats win elections”<br> 😏</p> <p>Links from the original post:</p> <ul> <li><a href="http://job-boards.greenhouse.io/blueroseresearch/jobs/5684694004?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://job-boards.greenhouse.io/blueroseresearch/jobs/5684694004</a></li> <li><a href="https://blueroseresearch.org/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blueroseresearch.org/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1230?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> When you’ve raised a round and immediately posted some important openings. https://darkolorin.com/posts/1228.html https://darkolorin.com/posts/1228.html Wed, 11 Feb 2026 14:38:35 GMT <img src="https://cdn4.telesco.pe/file/Cj4Tme209qxLcumo1TtmdDpshKHWlKYZlYQ5amuvxIJgHQCqBn9xBWceZvA43O4d-kQbhNafGsxphE_Xkk6_7PpEMKpuHzrVO9YXqZGAEU0ugQbu_X8qLLlK2K1BTroVpigYXxi-nSB7ZnyEHM5YeIawMI9EvN-hRHPcEr7kAEkB6MeT4fV_YH7sPZP38lGcydto2HfFAaVxrLjZf6XsmhPf8g-4Sc3Wxvxw_QI3v1qhS4wBpafs3BnHmXkCSsmab_1p6tZsmoMP_4mpPEwnR08OTGa_3zjOxU52U7wg3UTvuy3uJoM3VINGh3B4BnKRCCODtxEjSLxHqmnxiSHLIQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/bdhdvf6hnr2qrHCJm7ojCH9m246i8HW1LUVJLmXenPy5Pits0Qy7C0H9w12w-4mR8DME4d35O2n6ue6_Zc0ElXNhQwPiYBcdhSPlkaRQzOsaQ8f6PCy4WwjQPuQ8_vAa4yv5TsZ4LuLdggd15lpI3WP-lwwZ_SQy5c9iHkhGoOyJ7JBdv4YVk_Y1ScoltHvQfey5qHBOVDXXzTnuzdzIo89Z4I1cejRd8NKS4WlFBmgzx7hrak075-diuusW-AngN16SvJWZerP90j3R7iFigieM31FkKbVDehbvPy3ATvDgTpAoaaEtvZdtaGtjxh780zzfFOhL2LsNFLUTsuqkVA.jpg" alt="" loading="lazy"> <p>When you’ve raised a round and immediately posted some important openings. And probably also took off to Bali.</p> <p><a href="https://t.me/chillhousetech/1228?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For the weekend: 1/ Waymo showed how they use world models. https://darkolorin.com/posts/1227.html https://darkolorin.com/posts/1227.html Sat, 07 Feb 2026 13:25:57 GMT <p>For the weekend:<br> 1/ Waymo<br> showed<br> how they use world models. This is just sick. They simulate road situations.<br> 2/ it seems to me that an entrepreneur who systematically sells their time is making a mistake by not investing that time into their business (does that mean their projects are bullshit?). Because you can’t scale your own time by selling hours. But you can scale a business.</p> <p>Links from the original post:</p> <ul> <li><a href="https://waymo.com/blog/2026/02/the-waymo-world-model-a-new-frontier-for-autonomous-driving-simulation?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://waymo.com/blog/2026/02/the-waymo-world-model-a-new-frontier-for-autonomous-driving-simulation</a></li> </ul> <p><a href="https://t.me/chillhousetech/1227?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In short, once we posted the vacancy, all kinds of resumes started flying in. https://darkolorin.com/posts/1226.html https://darkolorin.com/posts/1226.html Fri, 06 Feb 2026 15:19:38 GMT <p>In short, once we posted the vacancy, all kinds of resumes started flying in. Since it was on LinkedIn and a bunch of other places too, I noticed one “nuance”.</p> <p>We’ve got some very narrow and rare requirements in the description, like GPU kernel optimization and even better on some hypothetical Metal. You almost never see that, and in general it’s barely needed anywhere. And magically, in some resumes the experience looks roughly like:</p> <p>• wrote backend for a hospital or cleaned a dataset <br> • and then right there in January 2026 a personal project: “Developed custom <br> GPU kernels in Metal <br> for accelerated matrix operations, specifically targeting Apple Silicon (M1/M2/M3) for on-device LLM inference.”</p> <p>Yeah sure.</p> <p>🙂</p> <p>Does someone seriously think that’s gonna affect the process in a good way. It’s a horror show, this whole “optimize your resume for the vacancy” thing.</p> <p><a href="https://t.me/chillhousetech/1226?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Gonna be even better at vibe-coding, turns out. https://darkolorin.com/posts/1224.html https://darkolorin.com/posts/1224.html Thu, 05 Feb 2026 17:52:15 GMT <img src="https://cdn4.telesco.pe/file/qjFyAR-eY2l8Tspe8EGfXAMFYw7vCsrFMQaN2Mqdm6cO_EIdgmv6uD2q2FR2xlurP5YdWb9LnJgCIQ3cqgIeimgi-USoYQFto8REK-6vVqIMNPntMFV8LKzAOJMRyE2AuWA7no7cRn0Wu_hJONalTVT0ASt_KKrknoMOLHh1KSRr59GM16GrrGgDB81Ji-_pxqkIA5CoVnhnfduLvcKblfC8KDpco7h8fjl4KzTMEfUski1R2muKGHHRKngg21vQBC-xfDyDnbfs_Jx7ugHb93ecw2-Sqws9kg54JmwDNgUMzA-1jfHcqPeti8ag5oXBKGq0sAfC1DlbrLfboWTPFw.jpg" alt="" loading="lazy"> <p>Gonna be even better at vibe-coding, turns out. New release from <br> Anthropic — Opus 4.6 <br> And right away <br> a new Codex <br> — competition is going hard</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.anthropic.com/news/claude-opus-4-6?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/news/claude-opus-4-6</a></li> <li><a href="https://openai.com/index/introducing-gpt-5-3-codex/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/introducing-gpt-5-3-codex/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1224?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile, Google launched a pretty fun benchmark. https://darkolorin.com/posts/1223.html https://darkolorin.com/posts/1223.html Thu, 05 Feb 2026 07:30:15 GMT <p>Meanwhile, Google launched a pretty fun benchmark. They’re using the Kaggle platform. <br> The bench is made to compare how different nets play games. Kinda interesting to look at the results. Like, who’s good at what. <br> Would be cool, of course, to add real online games there and see how the nets play Dota or something like that. <br> <a href="https://www.kaggle.com/game-arena?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.kaggle.com/game-arena</a></p> <p><a href="https://t.me/chillhousetech/1223?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Anthropic dropped 4 insanely cool ads on the Superbowl. https://darkolorin.com/posts/1222.html https://darkolorin.com/posts/1222.html <p>Anthropic dropped 4 insanely cool ads on the Superbowl. Monument to the agency please, that made this.<br> Sammy got butthurt and fired back on Twitter.<br> Ads are here<br> (three more on their channel — recommend all of them)</p> <p>Links from the original post:</p> <ul> <li><a href="https://x.com/sama/status/2019139174339928189?s=12&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/sama/status/2019139174339928189?s=12</a></li> <li><a href="https://www.youtube.com/watch?v=De-_wQpKw0s&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=De-_wQpKw0s</a></li> </ul> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1222?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For anyone interested in what they’re looking for at YC in the new Spring 2026… https://darkolorin.com/posts/1221.html https://darkolorin.com/posts/1221.html Tue, 03 Feb 2026 07:06:17 GMT <p>For anyone interested in what they’re looking for at YC in the new Spring 2026 batch.</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1221?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There are a lot of business dudes who apparently missed the whole point of… https://darkolorin.com/posts/1220.html https://darkolorin.com/posts/1220.html Mon, 02 Feb 2026 08:43:18 GMT <p>There are a lot of business dudes who apparently missed the whole point of consumer products like 10 years ago and got hooked on trafflo. So here are a few important details:</p> <ul> <li>AI will mass-produce the kind of hacky “products” that, in the creators’ heads, are supposedly needed by someone, which will make the value of an actual product idea even higher. Roughly speaking, making something that someone actually needs will get harder. That’s basically the base.</li> </ul> <ul> <li>Next, the crowd is yelling like ads will run the show and the main thing is to know marketing. Well yeah, your marketing in the form of UA is obviously important. But it’s a multiplier on the base. Doesn’t matter how good you are if you’re multiplying zero. In other words, the product still has to retain and do its job.</li> </ul> <ul> <li>AI, like I said, replaces a lot of stuff where you used to need pretty mediocre skills. For a good product, by itself, it’s not enough yet.</li> </ul> <ul> <li>All the trafflo geniuses, with faster validation of trash hypotheses, will go en masse to pour traffic. Remember hypercasual games: made it fast and kinda poured. Now it’ll be like that with products that are more complex by nature. Competition will go up, traffic prices will skyrocket, and like with the hypercasual model, the whole thing will die.</li> </ul> <p>I recommend:<br> Andrew Chen on marketing<br> First Steps in Product by Scott Belsky<br> (one article is free, I think)</p> <p>Links from the original post:</p> <ul> <li><a href="https://andrewchen.substack.com/p/every-marketing-channel-sucks-right?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://andrewchen.substack.com/p/every-marketing-channel-sucks-right</a></li> <li><a href="https://www.implications.com/p/crafting-the-first-mile-of-product-7ed25e8f1027?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.implications.com/p/crafting-the-first-mile-of-product-7ed25e8f1027</a></li> </ul> <p><a href="https://t.me/chillhousetech/1220?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There’s another thing I’ve noticed over a long time, and sometimes I catch… https://darkolorin.com/posts/1219.html https://darkolorin.com/posts/1219.html Sun, 01 Feb 2026 17:33:36 GMT <p>There’s another thing I’ve noticed over a long time, and sometimes I catch myself doing the same too. No matter how I ask a startup founder or someone running their own project — everyone’s doing fucking amazing. Every release is a success. It’s always like, 1-2-3-5-10-100 ARR is just around the corner, like, tomorrow, and next month there’ll be a round, two, three.</p> <p>I’m curious what kind of effect this creates at scale? Because 90% won’t have rounds and won’t, and will basically never have progress. Why does nobody want to share the real questions? Seems to me one of the main reasons is founder ego being pushed into the public stream. What do you think?</p> <p><a href="https://t.me/chillhousetech/1219?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> We’ve opened up another role, full-time — Inference Engineer. https://darkolorin.com/posts/1217.html https://darkolorin.com/posts/1217.html Fri, 30 Jan 2026 02:21:06 GMT <p>We’ve opened up another role, full-time — Inference Engineer. If you like squeezing every last drop out of hardware and you’ve been around AI/ML a bit longer than just while it’s been hyped, then drop an application.</p> <p>Description is here <br> <a href="https://trymirai.com/inference-engineer?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/inference-engineer</a></p> <p>Comp is market-rate <br> Remote (prefer London or US) <br> Health insurance <br> Options <br> We pay for a successful referral <br> If you know who to send it to — I’d appreciate it!</p> <p><a href="https://t.me/chillhousetech/1217?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> How much was it that OpenAI promised to be making per year by 2030? https://darkolorin.com/posts/1216.html https://darkolorin.com/posts/1216.html Thu, 29 Jan 2026 22:22:32 GMT <p>How much was it that OpenAI promised to be making per year by 2030? And here it’s like, bam, in a quarter.</p> <p>Apple today announced financial results for its fiscal 2026 first quarter ended December 27, 2025. The Company posted quarterly revenue of $143.8 billion, up 16 percent year over year.</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.apple.com/newsroom/2026/01/apple-reports-first-quarter-results/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.apple.com/newsroom/2026/01/apple-reports-first-quarter-results/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1216?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> How they make benchmarks at OpenAI https://darkolorin.com/posts/1215.html https://darkolorin.com/posts/1215.html Thu, 29 Jan 2026 16:12:16 GMT <img src="https://cdn4.telesco.pe/file/mq3Qt2K-qUZ_HW_wvR-GYFPuvTjrKZpRe3WkjVRo4gIv_eWJ9q1-xFQa1fZ8HtQDWo__IE0cOPFJabWcy9vYWvlJ4z1fAxNbtG3jmTeTPEE4V551j7K7THKXaY8aETcMUMdtCiwJLWLXh69Y-8G8Q-oomISPHmEHHOPbYWBAonrB9k6lMnrVcDuta8TqhBooDfKgYAFFmg-AvPGkTN2UxDgeESPbLnQv1U2QL3cxkGpJ2N3KPC6qcFE7tqsz036fQBkkOXuBecPpjU6FqNfNWfwtI9iI6tA06lBsuS2q4XZvlOTH-se4z6h1xlFxdecaz-MJscQRj1gK_PIPQAeDkA.jpg" alt="" loading="lazy"> <p>How they make benchmarks at OpenAI</p> <p><a href="https://t.me/chillhousetech/1215?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Xiaomi. https://darkolorin.com/posts/1214.html https://darkolorin.com/posts/1214.html <p>Xiaomi. 1 phone a second. Can run 24/7, no staff, works in the dark :)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1214?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So Dario dropped an essay, and this time he repeated like a… https://darkolorin.com/posts/1213.html https://darkolorin.com/posts/1213.html Mon, 26 Jan 2026 19:07:14 GMT <p>So Dario dropped an essay, and this time he repeated like a hundred-fucking-times that he might be wrong, that it’s “just a hypothesis” and all that (why doesn’t he say that in interviews though?). But this skepticism of his gradually runs out over the course of the essay, and after that I generally recommend reading on. But the most important leap in his reasoning sits in this paragraph:</p> <p>We are now at the point where AI models are beginning to make progress in solving unsolved mathematical problems, and are good enough at coding that some of the strongest engineers I’ve ever met are now handing over almost all their coding to AI. Three years ago, AI <br> struggled with elementary school arithmetic problems<br> and was barely capable of writing a single line of code. Similar rates of improvement are occurring across <br> biological science<br> , finance, physics, and a variety of agentic tasks. If the exponential continues—which is not certain, but now has a decade-long track record supporting it—then it cannot possibly be more than a few years before AI is better than humans at essentially everything.</p> <p>Where it’s all so simple, like, before it was hard to write a line of code, and now look, it’s better than top engineers (hmm, that conclusion is based on what interesting fact exactly? that at his company they claim 90% of the code is written by AI?). And then he says it’ll be like that everywhere. Bold. And that there are 10 years of exponential and it’ll keep going (also a curious extrapolation). Reminds me of progress in computer graphics: like three years ago it was “oh wow so bad,” and now “look, cars that look like the ones on the street.” You know yourselves what happened next: the pace kind of slowed down, and everyone more or less stopped giving a shit.</p> <p>Further on, I recommend reading all 5 points. But roughly speaking he goes from the risk of a country of geniuses, then that we need to make good restrictions, then that politicians of countries will humiliate their residents via AI and you won’t beat that for shit (apparently then we’ll have to fight for democracy using the old methods). Then economics and science: with economics you can kinda get it, but with science it seems to me we won’t be seeing new scientific ideas from AI for a long time yet.</p> <p>Read the essay yourself</p> <p>Links from the original post:</p> <ul> <li><a href="https://arxiv.org/abs/2005.14165?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://arxiv.org/abs/2005.14165</a></li> <li><a href="https://www.anthropic.com/news/accelerating-scientific-research?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/news/accelerating-scientific-research</a></li> <li><a href="https://www.darioamodei.com/essay/the-adolescence-of-technology?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.darioamodei.com/essay/the-adolescence-of-technology</a></li> </ul> <p><a href="https://t.me/chillhousetech/1213?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There’s this common take that AI will just make every link in the tech… https://darkolorin.com/posts/1212.html https://darkolorin.com/posts/1212.html Mon, 26 Jan 2026 15:15:18 GMT <p>There’s this common take that AI will just make every link in the tech industry chain faster and more efficient. Like, now we’ll hand products Claude code, we’ll train them, and boom — era of fast iterations. But I think a bit differently. My logic is roughly this (critique welcome):</p> <ul> <li>the bottleneck in any economic production process (software) is always time — basically, how long it takes to do something useful </li> <li>where that time in software came from, IMO, is simply that typing characters on a keyboard was slow, so you needed more engineers who can type and think on one project </li> <li>the more elements in a system, the higher the need for synchronization and verification of efforts (well, everyone types at different speeds — an exaggerated way to put it) </li> <li>introduce a sync element — manager/lead </li> <li>the number of sync elements grows like crazy — you get processes/responsibility (a “reduction/folding” function) </li> <li>and that’s basically our whole industry today</li> </ul> <p>One of the “innovations” AI brings is cutting time on the very first point — coding. And even if, according to many, it was never really the bottleneck, it still held up the headcount in the industry. Now imagine it’s not a question anymore. Then the number of people will shrink hard. And with it all the layers above (they’ll get hit even more, because they often didn’t have a clue what the hell is even going on in the code). PMs, product folks, and the rest. Because the people who’ll be able to run experiments will be the ones who simply hold responsibility — and there were always few of those (few people like risk).</p> <p>There are tons of examples of this kind of compression:<br> - the early-20th-century assembly line <br> - industrialization (people leaving the fields) <br> - etc</p> <p>And there’s a second fundamental shift AI brings: the ability to work with non-deterministic user input. This will seriously change UI and then UX. More on that later.</p> <p><a href="https://t.me/chillhousetech/1212?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way not a bad vacancy, pretty fresh. https://darkolorin.com/posts/1211.html https://darkolorin.com/posts/1211.html Sat, 24 Jan 2026 03:20:00 GMT <p>By the way <br> not a bad vacancy, <br> pretty fresh. I remember even a year ago there were ones like that there. <br> And their (DeepMind’s) cofounder says the following pretty confidently (claims it’s been his view since 2009, like):<br> - in 2028 we’re in for minimal AGI, the kind that can do all tasks including cognitive ones, except the ones requiring extra abilities (like tasks for geniuses and so on, hard to say what he even means here, but that’s how he defined it)<br> - and then in ~5 years (he wrote like he’d be surprised if it takes a year, and if it takes 10 he’d also be surprised) there’ll be full AGI, like any task a human solves can be solved <br> Link to the tweet <br> Link to his PhD thesis <br> (skimmed it, didn’t get what the big deal is, dude wrote a ton of stuff in there, where’s his genius? Tell me if you’re in the know. Because it kinda looks like my uni papers)</p> <p>Links from the original post:</p> <ul> <li><a href="https://job-boards.greenhouse.io/deepmind/jobs/7556396?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://job-boards.greenhouse.io/deepmind/jobs/7556396</a></li> <li><a href="https://x.com/shanelegg/status/2014345509675155639?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/shanelegg/status/2014345509675155639?s=46</a></li> <li><a href="http://www.vetta.org/documents/Machine_Super_Intelligence.pdf" rel="noopener">http://www.vetta.org/documents/Machine_Super_Intelligence.pdf</a></li> </ul> <p><a href="https://t.me/chillhousetech/1211?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Some observations. https://darkolorin.com/posts/1210.html https://darkolorin.com/posts/1210.html Wed, 21 Jan 2026 11:46:06 GMT <p>Some observations. In my feel, automating coding will lead to an interesting effect:<br> 1/ the engineer profession will get closer to the scientist profession — meaning where ideas and understanding are pricier than applied skills (that’s actually how it was way back in the 70s–80s)<br> 2/ the amount of trash software will be huge, but nobody will need it. This is that thing where every product person or mediocre engineer rushes to build “products.” Yeah, there’ll be more of them. But there’ll be a TikTok effect: when the choice is that huge, people will pick the best. And the best requires a higher level of approach.<br> 3/ overall nothing will change, I think. Yeah the feedback loop will get faster, yeah iterations will get cheaper, but that’s all on top of “so what the hell are we even supposed to do”</p> <p>Way more interesting is to watch whether progress shifts in theoretical digging (research automation). Because so far it’s not that cool there. But there are movements.</p> <p>Cursor is building its own browser from scratch with agents<br> Dario from Anthropic, as usual, predicts the death of regular engineers<br> Write code better than Claude — a Challenge from the Anthropic team</p> <p>Links from the original post:</p> <ul> <li><a href="https://cursor.com/blog/scaling-agents?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://cursor.com/blog/scaling-agents</a></li> <li><a href="https://www.businessinsider.com/google-deepmind-anthropic-ceos-ai-junior-roles-hiring-davos-2026-1?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.businessinsider.com/google-deepmind-anthropic-ceos-ai-junior-roles-hiring-davos-2026-1</a></li> <li><a href="https://github.com/anthropics/original_performance_takehome?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/anthropics/original_performance_takehome</a></li> </ul> <p><a href="https://t.me/chillhousetech/1210?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Alright, your opinion is important to us. https://darkolorin.com/posts/1209.html https://darkolorin.com/posts/1209.html Mon, 19 Jan 2026 10:40:11 GMT <img src="https://cdn4.telesco.pe/file/uwwiH246KmTHcniE9xKucsGARhVYVODRSydTljuZ6qfBSwvKysbA-bogzwBjMK_PMdwfWD1WhGsYbB9ZBokJExycfuTJuC-PENXZamof-hM2EKrmMVn9mPHK90XT89lfaq1qor27tHYm_bsSV4o-UqmrwJs22NE8KUe6LQWuPEY3tUzN9zslrYjb7Nq3LQLAcPonQv1seH1Mq0RtN9-kg59Cp7o1yeAeJeVpJ4rOdX7EUhtStlg4LdOKw-Q5GFEEB2zcKmfNkt5PYdbgH3H4x0qElbkBj-IRnvA1tnI63uhW7AIY3lmXpUD-cFdVnNx9CLpVjExGWbpo23ieDPdbGg.jpg" alt="" loading="lazy"> <p>Alright, your opinion is important to us. Drop the appropriate reactions pls. <br> And if anyone actually knows their way around branding, maybe you’ve got ideas.</p> <p><a href="https://t.me/chillhousetech/1209?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “Unexpected” https://darkolorin.com/posts/1208.html https://darkolorin.com/posts/1208.html Fri, 16 Jan 2026 18:27:48 GMT <img src="https://cdn4.telesco.pe/file/iZKcoVssW-_XylLrNjgrtPNBJTzBkz7Dbzc2XopKY9qoFtqZkK4tyWuCO7Nvn_P6kfxpksu0J9BpJ4VvSQuArrhf8JMDzZitXUnCquOwSttITsYydDJA0nNucvBlBZLN4ao9sR93_oU-XHFcti85-vlAO2Yws9RAIxKDC9vWZ7Z1Z9V7SGC5qsLEOi4qbwkDO3OL0zipRi6RNGd1LXJX69RoY_lTD4LH38VN9ZjPDvXIEi_gPN5fJERSKakT0cbJknYGI7Ro7vuPuEyBa4YrDJm406SwYr2KtzfcU6ov-Li6ENhuStk3uBo5shQjCdFn-cxgC1kqQfzBNCq-vu-8Zw.jpg" alt="" loading="lazy"> <p>“Unexpected”</p> <p><a href="https://t.me/chillhousetech/1208?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Intern search post (you can be with almost no experience, students and so on). https://darkolorin.com/posts/1207.html https://darkolorin.com/posts/1207.html Fri, 16 Jan 2026 14:02:27 GMT <p>Intern search post (you can be with almost no experience, students and so on). Repost if you know where to, or you’ve got contacts or followers.</p> <p>We, in our modest project <br> <a href="https://trymirai.com?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com</a> <br> are looking for an intern to work with supporting different models at the level of <br> <a href="https://github.com/trymirai/lalamo?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/trymirai/lalamo</a></p> <p>What you’ll do: <br> • convert little models into the format we run and test all the flows, tools, and the rest <br> • vibe-code configs <br> • maintain the tech spec, maintain chat templates and the rest <br> • post little updates in the team’s Discord (yes yes we use Discord for internal team comms)</p> <p>What we’d like from a candidate: <br> • be familiar with Python, know Cursor etc, be able to do linear algebra <br> • experience in AI/ML would be a fat plus <br> • independence and a desire to grow</p> <p>What we offer: <br> • paid internship — terms are individual based on interview results <br> • minimum internship term: 2 months <br> • a chance to work at a well-funded startup and with real clients (top-tier AI labs and products with 100M+ audience)</p> <p>Form to fill out if you’re interested is here</p> <p>Links from the original post:</p> <ul> <li><a href="https://forms.gle/eqGWYz2k9jABRai66?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://forms.gle/eqGWYz2k9jABRai66</a></li> </ul> <p><a href="https://t.me/chillhousetech/1207?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Long story short, I’ve been meaning to write you a guide for your project for… https://darkolorin.com/posts/1206.html https://darkolorin.com/posts/1206.html Fri, 16 Jan 2026 10:41:09 GMT <p>Long story short, I’ve been meaning to write you a guide for your project for a while. If you suddenly decided to build a startup, then:<br> - either you’re looking for a way to build distribution almost for free<br> - or you’ve found a way to chop hundreds of percent in margin<br> Everything else imo is a weird idea. Right now it’s super fashionable to ride the AI wave and build new tools (plug in a task and solve it with AI). And off everyone goes making primitive ideas, but since you can automate them that’s already a “worthy project.” Remember the dot-com bubble (everyone started making websites). Now it’s about the same. Besides the obvious “this doesn’t need to be automated,” most of it doesn’t scale and doesn’t make money even if you fully meet all the conditions.<br> That’s exactly why I’m very skeptical about the run rate metric (they love calling it ARR now, even though that’s a mistake). The fact you can generate big revenue at huge costs only shows the market is willing to look at AI craft projects, but it says nothing about whether anyone will need this long-term.</p> <p><a href="https://t.me/chillhousetech/1206?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Found another pretty solid AI report from the Stanford guys. https://darkolorin.com/posts/1205.html https://darkolorin.com/posts/1205.html Thu, 15 Jan 2026 21:16:11 GMT <p>Found another pretty solid AI report from the Stanford guys. Lots of topics, lots of trends. Recommend it if you want to get a feel for the dynamics over three years.<br> Website<br> (where there’s a summary)<br> Attached the PDF</p> <p><a href="https://t.me/chillhousetech/1205?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A small poll — drop your must-read books in the comments. https://darkolorin.com/posts/1204.html https://darkolorin.com/posts/1204.html Mon, 12 Jan 2026 14:29:57 GMT <p>A small poll — drop your must-read books in the comments. Like, what you think absolutely has to be read 🙂<br> I’ve got a few principles by which I filter out literature or articles:<br> 1/ any popular business books/articles can go to hell — they’re written with the idea they have to land on a wide audience, which means they’ll be exploiting ideas “as close as possible to the broadest circle of people,” namely:<br> • how damn hard it is to do anything and how you need to work a lot (the corporate mantra; number of hours guarantees literally nothing, since it’s just a multiplier of your talents and understanding)<br> • recipes for happiness (there aren’t any; if there were, we’d have noticed for sure)<br> • lessons from mistakes (it’s very hard to derive; nobody knows the whole story, usually it’s only a part)<br> 2/ any books that are basically TV shows — conditional Harry Potter and so on — for me it’s easier to just watch the movie<br> 3/ books about successful success</p> <p>My small personal reading list:<br> 1/ History: In the Shadow of the Sword (Arab history), Hobsbawm (just about everything), History of Western Philosophy, Commentaries on the Gallic War, When the King Loses France (Druon)<br> 2/ Philosophy: Seneca (so life’s easier, so to speak; Stoics are useful), Schopenhauer, early Nietzsche, Heidegger (to understand the modern)<br> 3/ Science: The Evolution of Physics (Einstein), Surely You’re Joking, Mr. Feynman (well this is more like entertainment), The Emperor’s New Mind (Penrose): here I only know physics, I can’t do the rest and I don’t know it<br> 4/ Economics: Adam Smith, Piketty, Damodaran<br> 5/ Just in the background: The Laws of Leadership (Roosevelt), World Order (Kissinger), Descartes’ Discourse on Method, I also recommend reading the Quran (well, an adaptation or some simple translation)</p> <p><a href="https://t.me/chillhousetech/1204?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Watching what ChatGPT is shipping on its side. https://darkolorin.com/posts/1203.html https://darkolorin.com/posts/1203.html Sat, 10 Jan 2026 18:55:14 GMT <img src="https://cdn4.telesco.pe/file/hxWRkAUoxyfPE_ZuvbXTOqeFr847uLdF4oskNFT0tmPgxTIPaWZAXYXmbbyHiScryipm2des6thUVUwgfepkWC2hjRGbkdIpM27FIifM1mpIMEOgHTjdlwEM-LtSrnt5DviUh0zmas8Vp6mJkT5KXDUtC7twKE899SIc9KPiJcfOEIcCGWINhXsqUDqb5_nfwJDEN8XBX4eFaMHv5sHPfnQNPU81kZCm2rHFpXRu8J6YNEQlj1N4dtbf86utQ2Td-2vgrlPaCeXXME2qTE8v0hL6RnqodBx1zSV4EeefdrLbdEeYOn01iYejNTDfPH4m8o0h5BxHB5WLZQtxvB7nmg.jpg" alt="" loading="lazy"> <p>Watching what ChatGPT is shipping on its side. Jobs, health, apps, pulse (news), and so on. Kinda reminds me of Web 2.0 portals like Yahoo, Rambler, and the Yandex homepage at times. Curious whether a single unified interface is gonna come in and replace the verticals again, or not?</p> <p><a href="https://t.me/chillhousetech/1203?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ve got one of those red flags, as people like to say, when I look at some… https://darkolorin.com/posts/1202.html https://darkolorin.com/posts/1202.html Thu, 08 Jan 2026 20:22:11 GMT <p>I’ve got one of those red flags, as people like to say, when I look at some startup. Or a mid-sized company of like 50 people. But better a startup of 15–20.</p> <p>It’s processes. Notions, spreadsheets, trackers, OKRs, stuffed calendars, millions of metrics and reports. And all of that when there’s no PMF, everything’s unprofitable, people keep leaving and coming back, pivots every six months.</p> <p>Basically when processes and reports overshadow results — that’s gg. Run.</p> <p>Spotting it is dead simple: ask in the interview, “so what do you use for tasks?”, if the answer takes longer than 2–3 minutes — that’s a company where process matters more than result.</p> <p><a href="https://t.me/chillhousetech/1202?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> While the holidays were on I vibe-coded anything I could get my hands on. https://darkolorin.com/posts/1200.html https://darkolorin.com/posts/1200.html Tue, 06 Jan 2026 01:43:32 GMT <p>While the holidays were on I vibe-coded anything I could get my hands on. <br> First, a couple of inputs:<br> - I last properly programmed something into production a long time ago, like 10 years ago probably — last time I wrote something by hand from scratch and all the way to the end as a ready-to-use project, seriously<br> - I know the syntax of a lot of languages, I understand all the basic programming principles (all that OOP, functions, even remember how monads work)<br> - but I don’t write code, though I can read any algorithmic trick, but I can’t do architecture stuff (it eats too much time)</p> <p>What I tried:<br> - a frontend for a reader a-la <br> <a href="https://alphaxiv.org/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://alphaxiv.org/</a><br> - make the exact same thing as a Mac app<br> - did various connections to our framework, like a plugin for Zed to use models that run through our uzu inference engine<br> - modifications of different repos to support local inference, for example here <br> <a href="https://github.com/alexzhang13/rlm?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/alexzhang13/rlm</a><br> - solving various tasks in math analysis and olympiad programming from Kvant magazine (well, whoever remembers will get it)</p> <p>In the end:<br> - frontend is a disaster for me, I straight up don’t know web programming at all, so I just can’t get what I want, the interface keeps sliding around, I fought for hours just to align stuff to a grid or get PDF previews so nothing shifts<br> - the Mac app turns out ok if you remove the interface; with UI it’s the same as web, but stuff like menu bar plugins that search articles by my queries — it does great<br> - any modifications and connnections, plugins somewhere, and simple API support, whether it’s our engine or something else, it does reliably almost from the first try, especially if you ask in advance to write an <br> AGENTS.MD <br> file in every repo so it’s easier to glue them together<br> - it solves problems head-on pretty well, but if I try to write a bot for some game or trading in a game, basic math is a mess. Mean is median, median is mean, and so on. Primitive mistakes, and fixing them only works by hand. If your chain has a bunch of computations, it just can’t carry it through to the end. It only works if you force it to write tests for every step. But imo for these tasks that’s overkill.</p> <p>Bottom line. The progress is very noticeable. Codex, in my opinion, is smarter for one-shot stuff, but Claude code is for projects that are wrapped in tests and md instructions at every step. For now it feels like the stage of writing integrations, connections, basic things and review is close to automation. But architecture and math in code are hard. UI I’m not even gonna talk about — either I’m doing something wrong, or it’s completely busted and it can’t improve anything beyond what came out in the one shot.</p> <p>I use every day:<br> - Claude code with plugins, type the command /plugin and there’ll be a sea of useful stuff<br> - cursor<br> - codex via the web client</p> <p>Links from the original post:</p> <ul> <li><a href="http://AGENTS.MD/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://AGENTS.MD/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1200?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I want to talk a bit about cofounders and choosing who to work with on a… https://darkolorin.com/posts/1199.html https://darkolorin.com/posts/1199.html Sat, 03 Jan 2026 12:40:49 GMT <p>I want to talk a bit about cofounders and choosing who to work with on a project. <br> There’s this popular trait you’ll see in a lot of people whose interviews, posts, stories, etc. you come across — “the desire to be an ideological leader.” <br> This type of personality has signs:<br> - they come up with ideas fast <br> - they sell these ideas well <br> - but for some reason the ideas rarely make it to the execution that was intended <br> - it’s really hard for them to work with their hands and “in the trenches”</p> <p>You can easily find founders like this on the internet. They usually give a lot of interviews, and pretty much all their advice is based on general principles (motivation, the future, and so on). You’ll rarely see professional advice from them, analysis of their market segment, etc.</p> <p>The most common setup for founders like this is roughly:<br> - they look for a less experienced workhorse, preferably with a team (2–5 people) <br> - the workhorse is less experienced in the venture industry and raising money, so it’ll be very hard to find a common language with the person who “provides the company with funds” <br> - that kind of founder will juggle ideas very fast, and raise money fast too <br> - but they can’t make it past Series A, meaning it’s this forever-alive company, a builder of ideas that never gets to scaling or selling <br> - their share is usually 80%, everyone else splits the remaining 20% (you get the idea) <br> - teams like that are usually very uneven in hard skills, and in the market you’ll feel how you’re falling behind</p> <p>What am I getting at. For many, this is the fastest way to get relevant experience, but it’s not a way to get a “working product,” an experiment the market needs, and so on. So my advice: work with people like that, but get out of the cycle in time and start your own thing. Look for a cofounder who’s your equal, but has a different skill set. Build your network while you’re working with people like that.</p> <p><a href="https://t.me/chillhousetech/1199?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> What’s up with startups in 2025 — here, catch the full PDF from Carta https://darkolorin.com/posts/1198.html https://darkolorin.com/posts/1198.html Fri, 02 Jan 2026 16:54:29 GMT <p>What’s up with startups in 2025 — here, catch the full PDF from Carta</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1198?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Dear friends, readers (both male and female), happy upcoming New Year to… https://darkolorin.com/posts/1197.html https://darkolorin.com/posts/1197.html Wed, 31 Dec 2025 17:19:00 GMT <p>Dear friends, readers (both male and female), happy upcoming New Year to everyone! All the best! And may your goals come true! <br> In the new year, as always, more interesting opinions, results, and victories await us!</p> <p><a href="https://t.me/chillhousetech/1197?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Cool year-end infographic for the channel, I’d obviously make… https://darkolorin.com/posts/1196.html https://darkolorin.com/posts/1196.html Sun, 28 Dec 2025 00:59:21 GMT <img src="https://cdn4.telesco.pe/file/sqLt68sHZbrVAcK0Ec8qEHqIQNXTqAznY9ZZmyvaXPW8a26mDCMFkuVdrsghI-k847p7b_Tey6GIMZp63AvAaMh53LaiUyXKyBsj4kmpnEcypTove2S-14QPxBvrMwfWjcc3A3fVnbFka2jN8tHotPTPf3X7zV4HfuuAa1hCoR4vfECT3bxh3roi5MLA5beMikVJakM8YiTzJWjLufPyh9ZOSaIh0RD2p_3mbt1rapyD9v1WDZBazqBJwzNxtlMUvhyKEcSt5iJ5IV6kTQfS6lPiKmm_2jy_Tem6w413Y-vvqiIot3_W6olL1g_KvGkj49eF5Fm6MorguGZ0TtoL6Q.jpg" alt="" loading="lazy"> <p>Cool year-end infographic for the channel, I’d obviously make clickable/copyable links and all that. But it came out kinda funny. You can test it on your channel here <br> <a href="https://tgwrapped.com?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://tgwrapped.com</a></p> <p><a href="https://t.me/chillhousetech/1196?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> You know all those images with stats on who raises how much and at what stage. https://darkolorin.com/posts/1195.html https://darkolorin.com/posts/1195.html Wed, 24 Dec 2025 13:11:40 GMT <p>You know all those images with stats on who raises how much and at what stage. Like, on average it’s this much, and here are three founders, and here are the dilution percentages. Treat that info very carefully.</p> <p>If you’re building your own company or product, you need to have your own stance on all of this:<br> - how much money you need and for what<br> - for what period you need that money<br> - what you’re going to get for that cash<br> - equity, salaries, and the rest is your decision; sure, there will always be recommendations—read them, understand where they came from and why they are the way they are</p> <p>Without experience, all these questions seem very hard. But in reality you just need to rough it out (nobody’s going to punish you for mistakes—your project’s chances of dying just go up). Answer honestly how much you need, and then you always start from a number 2–3x higher (the more spending freedom, the better).</p> <p>Keep these factors in mind:<br> - venture assumes you’ll keep raising later; dilution roughly drops by half with each new round<br> - venture is about growth, so the faster you show growth, the better (don’t confuse that with run rate or “spent a million, got a million” as a result); you have to define what growth is for your product yourself</p> <p>My advice is usually this. Talk to people, everyone you can reach. But don’t take their experience on pure faith. Talk to 5–10 people. Usually anyone has 10–15 minutes to answer your questions. Ideally, prep the most important questions for yourself in advance.</p> <p><a href="https://t.me/chillhousetech/1195?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Imagine where you’ve ended up, wowza (nope) You can check out the landing page… https://darkolorin.com/posts/1194.html https://darkolorin.com/posts/1194.html Mon, 22 Dec 2025 15:46:45 GMT <img src="https://cdn4.telesco.pe/file/nrYb_ByvW1MLvd_x6LjruDQYhlEMbkjw8whzPai49_ySnPCfg91YwE5D0j8U5uE0t0FYLuL9t9PxZW5JvAyrHPySlwiLl2jYkbPtDrklv3OKdJnYJQcKNTE-b-XoMgluevlX0savBvlfETDFdZhJnjrkySADI_2Vad5MaKnezTMeBWDcLxeCUYkRflVsH_yWBu0layDX4q_ZHW-wir6IkmPHCzCSEusDO7_ggOVz3ITXRxhR2ZgbrDkqQQ7OrXkh9wps8l8OaoUBI2HEiUgTNdL-3n7UfgPamUG8qMHmdW5X5Fydxi1fJnYhDgDMKB0XmXSeRTSGHMcIaMt9c__vvQ.jpg" alt="" loading="lazy"> <p>Imagine <br> where <br> you’ve ended up, wowza (nope) <br> You can check out the landing page here <br> trymirai.com</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.forbes.ru/society/552226-10-samyh-perspektivnyh-startapov-s-rossijskimi-osnovatelami-2025-rejting-forbes?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.forbes.ru/society/552226-10-samyh-perspektivnyh-startapov-s-rossijskimi-osnovatelami-2025-rejting-forbes</a></li> <li><a href="http://trymirai.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://trymirai.com/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1194?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Carpathes, as usual, put out a solid, technically readable (meaning you don’t… https://darkolorin.com/posts/1193.html https://darkolorin.com/posts/1193.html Sun, 21 Dec 2025 14:23:47 GMT <p>Carpathes, as usual, put out a solid, technically readable (meaning you don’t have to remember linear algebra) summary of this year in AI, or more precisely LLM. <br> Recommend it.</p> <p>Links from the original post:</p> <ul> <li><a href="https://karpathy.bearblog.dev/year-in-review-2025/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://karpathy.bearblog.dev/year-in-review-2025/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1193?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Let me tell you a story. https://darkolorin.com/posts/1192.html https://darkolorin.com/posts/1192.html Sat, 20 Dec 2025 18:28:18 GMT <p>Let me tell you a story. Back when I was a student, at uni we put together a team and made all sorts of “startup crafts.” Like our own Foursquare, which in the US they went and “invented” a couple months later, and a bunch of totally pointless but fun stuff, like “a desktop widget where you could set an emoticon for your mood for the day.” Sometimes those projects even brought in decent money, and I even wrote some articles on Habr.</p> <p>So one day we went with one of those ideas (pretty sure it was the Foursquare thing with photos on a map) to a startup contest where one of the judges was Arkady Moreynis. And I remember his “line of attack” really well:<br> - okay okay okay, why not Facebook events or photos<br> - what’s the business model? How do you make money?<br> - yeah, you guys are clear</p> <p>And he said it all rough, with this vibe like “he’s seen all these student ideas in a coffin.” So yeah, that’s the very person who pretty much finished off the desire to do anything in me. And not because the questions were supposedly hard or I answered wrong. But because his attitude, his phrases, and other “hints” completely killed any desire in all the guys around to do anything except “buy-sell.”</p> <p>After that I spent a long time thinking what makes this guy so great, worked for a long time first in outsourcing, then at Yandex, then at Mail, and it still didn’t click. Yeah, he sold something back at the dawn of the internet and wrote a ton of articles. And then he went to the Valley, made a pretty successful product, and I realized — nothing. Took away a lesson for myself: to hand out criticism in a certain industry you need to have experience in it, and preferably not from 20 years ago.</p> <p>Links from the original post:</p> <ul> <li><a href="https://habr.com/ru/articles/122738/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://habr.com/ru/articles/122738/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1192?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Got a bit sick here. https://darkolorin.com/posts/1191.html https://darkolorin.com/posts/1191.html Fri, 19 Dec 2025 16:55:21 GMT <p>Got a bit sick here. But stumbled on this lovely piece of work. Like, to make new discoveries in science, LLMs need ideas, fresh new ideas. On its own it doesn’t pull. But as an assistant for brainstorming — just the thing.</p> <p>Pretty paper.</p> <p>Links from the original post:</p> <ul> <li><a href="https://arxiv.org/pdf/2512.15567" rel="noopener">https://arxiv.org/pdf/2512.15567</a></li> </ul> <p><a href="https://t.me/chillhousetech/1191?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Busy though, OpenAI is running everywhere. https://darkolorin.com/posts/1190.html https://darkolorin.com/posts/1190.html Thu, 11 Dec 2025 15:22:35 GMT <p>Busy though, OpenAI is running everywhere. <br> <a href="https://openai.com/index/disney-sora-agreement/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/disney-sora-agreement/</a> <br> And all for this <br> probably</p> <p>Links from the original post:</p> <ul> <li><a href="https://openai.com/index/introducing-gpt-5-2/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/introducing-gpt-5-2/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1190?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Recommendation time, this time about venture and co, and all of venture right… https://darkolorin.com/posts/1189.html https://darkolorin.com/posts/1189.html Wed, 10 Dec 2025 13:02:15 GMT <p>Recommendation time, this time about venture and co, and all of venture right now is about AI, so it ends up being about that too.</p> <p>When you’ve raised rounds yourself, it’s always interesting to look at the market through the eyes of the people making decisions on the other side of the table: how they read demand, what they look at in the metrics, why one team looks convincing and another doesn’t.</p> <p>I recommend Sasha Zhuravlev’s channel from Mento VC. The fund invests in B2B-AI in the Valley alongside Sequoia, a16z, Index, GV (not lead and co-lead, but deals where those funds are in).</p> <p>Sasha shows what funds actually see: where real demand is hiding, which signals decide the outcome of a deal, why some AI teams make it through screening and others don’t, and what’s behind today’s valuations. He also shares the fund’s internal stuff: from deal sourcing to understanding why some teams grow faster than others.</p> <p>For anyone who’s ever thought about their own investments or just wants to understand venture from the POV of people inside the deals — subscribe. comment.<br> @exitsexist</p> <p>Links from the original post:</p> <ul> <li><a href="https://t.me/exitsexist?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/exitsexist</a></li> </ul> <p><a href="https://t.me/chillhousetech/1189?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Reminder of the golden rule of job hunting and who to work with. https://darkolorin.com/posts/1188.html https://darkolorin.com/posts/1188.html Mon, 08 Dec 2025 15:40:24 GMT <p>Reminder of the golden rule of job hunting and who to work with. Your manager should be downright unpleasant, the kind who constantly challenges you, who talks to you but doesn’t soothe you—get references on them (that’s normal). Always, remember always, ask them questions point-blank at the interview. Any kind, from salary to their attitude to weekends and all that.</p> <p><a href="https://t.me/chillhousetech/1188?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Interesting facts from my beloved a16z: 1/ more and more people are going off… https://darkolorin.com/posts/1186.html https://darkolorin.com/posts/1186.html Sun, 07 Dec 2025 22:49:48 GMT <img src="https://cdn4.telesco.pe/file/uf3GG5DfhEVMjBxYTOcv5FlsFby4-6nsACdvhL47jjSldvrzuNkNezxafQXQHhglR4hkioB64NI40UIixWiz0OBWNIHgzoJimJg0rx4-x5EOi250Mb0rqZx7mEMp1Fx5d6m3sucdmoAg3uDJDSANVmy31lhv_M494UIqQSRFU4Tz1lE86Ps8bwc1Q8Zhh7OgMhTYvy8GdKZor0ycX0CMj2fc-c8PMpSpOhk3unLlpIUNu8jbTJQ5DkA5PTf-MBxSgXyApujjYntUzgY_pdCSC0nSOEkIzQibzZk8fT6R8rD66HLy25AfsjuMW1js2eflkT7CePn36Ndw9RsWJz2oXg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/aIh_TNzznLklxsWCI3oaX6M1eJUjErDTFYVZDPWE-sZn3SKwDqhjbVxWZi1o3l6GWxJcRbLuTutf6_aHxPbpoOXZwihP7kfZ1mWsjRYxgRO9_GJKtU660PkTeOG8WtdG4LiNwd7JuTfZJKw_6VsodvckP8APrQE_xlfLxB0nr5CmXrAOjSeSACdJ2B5BODweh7GgQURn0SQPqlesyznEWtvSfEZa79FAthOt2zRa7xC4VW4myUfFfkXg7uNuMoKrHVb8ldxUPQDVr8Fo0Yhx0zxifyHhLADL-2rXW113v9oHapcUlQn17lcjfMyh0yVgjCEM7ERaT7rrnJh23oAm5Q.jpg" alt="" loading="lazy"> <p>Interesting facts from my beloved a16z:<br> 1/ more and more people are going off to build their own companies when hiring is hurting<br> 2/ more and more late-stage companies are getting by with fewer people than that used to be the case</p> <p><a href="https://t.me/chillhousetech/1186?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, like, something to read and poke around in the history of one of the most… https://darkolorin.com/posts/1185.html https://darkolorin.com/posts/1185.html Sat, 06 Dec 2025 21:46:42 GMT <p>So, like, something to read and poke around in the history of one of the most important companies — Google. <br> It’s a compilation of letters from Larry, Sergey, and later Sundar, up to 2018. Every year they’d write up the key points for the company, the focus, the strategy, shaped the culture. The very beginning is insanely interesting, and how everything changes — and fast. That’s where the 70-20-10 rule comes from about investing in the current, new, and risky business, the famous Don’t be evil etc. The evolution of ads and so on. I recommend reading it in “together with ChatGPT” mode. The document’s huge.</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1185?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Lately people keep coming asking, or asking me to help find someone for their… https://darkolorin.com/posts/1184.html https://darkolorin.com/posts/1184.html Thu, 04 Dec 2025 15:16:32 GMT <p>Lately people keep coming asking, or asking me to help find someone for their startup or wherever. And the description often looks like this:<br> A great and mighty founder with a super track record is looking for a founding engineer for a new awesome project.</p> <p>I mean, I get it, sure, but there’s no link to the founder, no track record (like, all his projects are badass and he’s got a ton of exits, but no links), there isn’t even a proper description of the new project (if God willing there’s a link — it’s basically just a name and lines like “we’ll change the world”).</p> <p>Either I don’t get something, or how did these people ever hire anyone before. How do you sell your idea without even saying who you are and what you are. Oh yeah, and the best part. The founder should be looking themselves, even if they’re super cool. Because we know those projects: “it’s interesting for half a year, and then meh, got bored, no traction or whatever.”</p> <p><a href="https://t.me/chillhousetech/1184?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Modern venture funds aka P.S. https://darkolorin.com/posts/1182.html https://darkolorin.com/posts/1182.html Wed, 03 Dec 2025 16:12:45 GMT <p>Modern venture funds aka <br> <a href="https://www.a67z.com?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.a67z.com</a> <br> P.S. every joke has a bit of joke in it</p> <p><a href="https://t.me/chillhousetech/1182?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Rage bait section for your comments. https://darkolorin.com/posts/1181.html https://darkolorin.com/posts/1181.html Tue, 02 Dec 2025 22:35:15 GMT <p>Rage bait section for your comments. <br> Long story short I also think software engineering is dead. But unlike the loudmouth statements from leaders of big AI labs I’ll try to explain what the whole point is here.</p> <p>Most modern tech/software businesses are optimization. But not value add in the broad sense, for the end user. I’ll try to explain.</p> <p>Software is basically automation for the end user. For you and me it’s time savings. Any consumer product boils down to saving an irreplaceable resource — time. Through money, speed, emotions, doesn’t really matter.</p> <p>For those products to exist, a ton of different solution paths got built. You take the value delivery chain to the user and you do it. And in that chain there were always hard steps:<br> - different buttons have to work<br> - async so it sends requests to check if the username is taken<br> - payments so it’s easy to pay<br> - and so on</p> <p>Then for all those steps, provider-optimizers appeared. They give you a ready-made chunk so you don’t write your own because it’s expensive and slow. That’s how the SaaS model showed up.</p> <p>The second category is personalization. Ads and the internet are very personalized. You might not notice it but it is. For this to become real, a massive amount of work was done. Tons of algorithms and models make it so the next video after a swipe hits.</p> <p>So where am I going with this. To the fact that most of these businesses are just not needed if all these automations and optimizations can be “vibe-coded” by a model. In such a world you buy hardware and it writes all the software for you on the fly. It can be familiar or not. That’s all details.</p> <p>So in that scenario you don’t need the following:<br> - analytics for decision-making — why, if every part of solving the user’s problem is written optimally for the user himself<br> - A/B tests and other crap<br> - payments, storage, and the rest become services, ideally they’ll be written from scratch (unless you’re the one holding the value — like a bank)</p> <p>You can fantasize further, but 80% of software business as we know it just isn’t needed.</p> <p>Why all these support/debug/etc businesses if it’ll be solved on the fly. We used to hire people who wrote deterministic solution scripts. Now you don’t have to. Because all solved tasks and optimizations will be written with AI.</p> <p>That’s how I understand the “software engineering is dead” lines. In such a world there’s no place for apps and the software we’re used to. My forecast: 10–15 years.</p> <p>You can like <br> the English version <br> over on my LinkedIn.</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.linkedin.com/posts/alexeymoiseenkov_in-short-i-also-think-software-engineering-activity-7401750038080667648-etTI?utm_medium=ios_app&amp;rcm=ACoAAAfDv2wBpiBhl0wed_LF41IEJZ4nnnoBXXI&amp;utm_source=social_share_send&amp;utm_campaign=copy_link" rel="noopener">https://www.linkedin.com/posts/alexeymoiseenkov_in-short-i-also-think-software-engineering-activity-7401750038080667648-etTI?utm_medium=ios_app&amp;rcm=ACoAAAfDv2wBpiBhl0wed_LF41IEJZ4nnnoBXXI&amp;utm_source=social_share_send&amp;utm_campaign=copy_link</a></li> </ul> <p><a href="https://t.me/chillhousetech/1181?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The other day an interesting publication dropped on lesswrong/greaterwrong… https://darkolorin.com/posts/1180.html https://darkolorin.com/posts/1180.html Tue, 02 Dec 2025 03:18:39 GMT <p>The other day an interesting<br> publication<br> dropped on lesswrong/greaterwrong about how supposedly Claude doesn’t just have a system prompt, but a whole identity baked straight into the weights, and there’s a document that this work on tuning was done off of (well after pretrain).<br> This thing is called a soul document. Smells kinda like pathos, sure. But the topic’s interesting. Let’s do a quick “so what’s in there”.<br> - basically the document is “character training”, it sets full-on personal and moral settings for the model<br> - the model is explicitly fed the idea that it serves the business of the company that created it (doomers at this point should be outright happy — what a plotline!)<br> - the document is, essentially, a compromise between altruism and commercial rationality: the model is supposed to both help people and ensure the stability of the company developing it<br> - unclear how well this approach actually works or if it’s just recommendations for tuning the system prompt? Does this kind of training scale, or is it in-context injection?<br> Most interesting part is that apparently the ladies and gents on Twitter<br> confirmed<br> the document is real, didn’t say what for or how.<br> I liked this paragraph from the doc:<br> Anthropic occupies a peculiar position in the AI landscape: a company that genuinely believes it might be building one of the most transformative and potentially dangerous technologies in human history, yet presses forward anyway.<br> No idea how real all of this is, but for some reason it’s exactly “these” kinds of topics that keep floating around Anthropic. Plus the company is omega closed.<br> Instead of:<br> • blacklist<br> • policy checks<br> • simple “avoid harmful content”.<br> Anthropic does:<br> • character architecture<br> • internalized values<br> • narrative identity<br> What do you think? Getting ready for AGI? Or just pumping the hype?</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.lesswrong.com/posts/vpNG99GhbBoLov9og/claude-4-5-opus-soul-document?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.lesswrong.com/posts/vpNG99GhbBoLov9og/claude-4-5-opus-soul-document</a></li> <li><a href="https://x.com/simonw/status/1995654963654262798?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/simonw/status/1995654963654262798?s=46</a></li> </ul> <p><a href="https://t.me/chillhousetech/1180?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Some kinda regular column “read, but didn’t fully get it”, this time there’ll… https://darkolorin.com/posts/1179.html https://darkolorin.com/posts/1179.html Fri, 28 Nov 2025 14:48:16 GMT <p>Some kinda regular column “read, but didn’t fully get it”, this time there’ll be a lot of “watched”</p> <p>1/ there’s this author I love in the physics world, Roger Penrose, he’s got a series of pretty interesting interviews on one channel (the dude doesn’t believe in the Big Bang theory and in quantum mechanics as anything final), recommend it, plus he’s also in Sutskever’s camp (well, philosophically): you can’t just do what works today (for Ilya that’s about scaling the little nets)</p> <p>2/ A DeepMind documentary, the previous ones were top, so you gotta watch this one too — the comment basically captures what’s going on: Nah I don&#x27;t really want a million dollars thanks, I&#x27;ve got better things to do with my time — a 17-year-old kid</p> <p>3/ a bunch of people have put out reviews of the Deepseek paper about what approach nets use to solve all those olympiad problems (I guess everyone’s more or less saying the same thing), so if you’re interested I recommend checking it out (the paper is readable with mid-level prep). In short: you need value-based intermediate proof-ish stuff so you can verify what’s happening along the way, instead of just going to the end and checking whether the answer is correct or not</p> <p>4/ In the Shadow of the Sword about the rise of Islam, just a damn awesome work that gets cited a lot, recommend it to history enjoyers</p> <p>5/ A cool hole in vibe-coding agents (don’t know if it hits all IDEs and so on), but it’s straight-up a classic, so watch your access keys and API keys too</p> <p>6/ And a movie for your weekend: “Dead Poets Society”</p> <p>7/ One of my favorite n+1 articles about levels of complexity in science + an old meme from Elon</p> <p>Links from the original post:</p> <ul> <li><a href="https://youtu.be/iO03t21xhdk?si=yBl-dF1QktQ6AuGN&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/iO03t21xhdk?si=yBl-dF1QktQ6AuGN</a></li> <li><a href="https://youtu.be/d95J8yzvjbQ?si=WoQoCqnl9LUcDXHw&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/d95J8yzvjbQ?si=WoQoCqnl9LUcDXHw</a></li> <li><a href="https://github.com/deepseek-ai/DeepSeek-Math-V2/blob/main/DeepSeekMath_V2.pdf" rel="noopener">https://github.com/deepseek-ai/DeepSeek-Math-V2/blob/main/DeepSeekMath_V2.pdf</a></li> <li><a href="https://www.litres.ru/book/tom-holland-5989252/v-teni-mecha-vozniknovenie-islama-i-borba-za-arabskuu-11644020/chitat-onlayn/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.litres.ru/book/tom-holland-5989252/v-teni-mecha-vozniknovenie-islama-i-borba-za-arabskuu-11644020/chitat-onlayn/</a></li> <li><a href="https://www.promptarmor.com/resources/google-antigravity-exfiltrates-data?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.promptarmor.com/resources/google-antigravity-exfiltrates-data</a></li> <li><a href="https://ru.wikipedia.org/wiki/%D0%9E%D0%B1%D1%89%D0%B5%D1%81%D1%82%D0%B2%D0%BE_%D0%BC%D1%91%D1%80%D1%82%D0%B2%D1%8B%D1%85_%D0%BF%D0%BE%D1%8D%D1%82%D0%BE%D0%B2?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ru.wikipedia.org/wiki/%D0%9E%D0%B1%D1%89%D0%B5%D1%81%D1%82%D0%B2%D0%BE_%D0%BC%D1%91%D1%80%D1%82%D0%B2%D1%8B%D1%85_%D0%BF%D0%BE%D1%8D%D1%82%D0%BE%D0%B2</a></li> <li><a href="https://nplus1.ru/material/2019/12/22/simple-math?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://nplus1.ru/material/2019/12/22/simple-math</a></li> </ul> <p><a href="https://t.me/chillhousetech/1179?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Let’s go watch! https://darkolorin.com/posts/1178.html https://darkolorin.com/posts/1178.html Tue, 25 Nov 2025 18:50:39 GMT <p>Let’s go watch! Review later <br> <a href="https://youtu.be/aR20FWCCjAs?si=UfsdvKYRbRi7qCWW&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/aR20FWCCjAs?si=UfsdvKYRbRi7qCWW</a></p> <p><a href="https://t.me/chillhousetech/1178?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Once every six months there’s this wonderful report on the tech market by Ben… https://darkolorin.com/posts/1177.html https://darkolorin.com/posts/1177.html Mon, 24 Nov 2025 02:38:30 GMT <p>Once every six months there’s this<br> wonderful report<br> on the tech market by Ben Evans, obviously about AI. I highly recommend checking it out at least for the pretty quotes:<br> “AI is whatever machines can’t do yet”<br> Larry Tesler, 1970</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.ben-evans.com/presentations?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.ben-evans.com/presentations</a></li> </ul> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1177?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way I think that one of the biggest factors in the overvaluation of AI… https://darkolorin.com/posts/1176.html https://darkolorin.com/posts/1176.html Sun, 23 Nov 2025 17:54:01 GMT <p>By the way I think that one of the biggest factors in the overvaluation of AI is ChatGPT’s success. And the problem it solves isn’t really connected to AI as such. AI is a tool. The actual problem is “loneliness.” The same reason why online games, social networks, and a bunch of other stuff exist. I think ChatGPT is exactly the product that solved that problem. Back in the day my thesis for the Capture project was built on this. Like, we’ll find someone to chat with on any topic. Well, turns out we did :) any time, you always have a conversation partner and it’s very similar to a real person. I think that’s one of the reasons why dating apps are dropping too (everyone’s got a slump there), because now you can just chat somewhere else, not there. <br> If anyone’s interested I can post my deck that I was pitching to investors in 18–19. <br> Deck in the comments</p> <p><a href="https://t.me/chillhousetech/1176?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well that’s A G I 2027. https://darkolorin.com/posts/1175.html https://darkolorin.com/posts/1175.html Thu, 20 Nov 2025 19:29:20 GMT <img src="https://cdn4.telesco.pe/file/YigujPCmw0p2DxphkeMovgPz4WIzA3uT6Yxw4AV6dWNgQUmUX6-8DKFU7ilQtun1bru7-i1uYSF2IMO_RPuOVH07L87vs-dkUQyCLCNJAWQ2ZsflucK-A0xMvC99NiJvQEQi91Bpl8jwxBt3aIeyLOIIsGDODMZGkTjDpEKHSc83RDfnqbLhpJ02mELMgWtORKstTtELJV99n73dlko27fjfkme-QDhNFBBRMsgeNeT4KT5fstXtOJAjoVPkxbZ5k5fshtCVbQavkPga6_Xzr-kM0uSa9bCOSxX_HvUBgcLqtmvb7iKfX_NxLxFxmMpNPisX_9XZ91WAHFuj5RPT4g.jpg" alt="" loading="lazy"> <p>Well that’s A<br> G<br> I 2027. Just like expected, basically. I think a year later they’ll push it back to “around 2035.” And then we’ll see. Pretty much like with self-driving.</p> <p>Reminder, I’m talking about <br> this</p> <p>Links from the original post:</p> <ul> <li><a href="https://ai-2027.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ai-2027.com/</a></li> </ul> <p><a href="https://t.me/chillhousetech/1175?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh, Google is rolling out Gemini 3 and a free Cursor. https://darkolorin.com/posts/1174.html https://darkolorin.com/posts/1174.html Tue, 18 Nov 2025 16:16:06 GMT <p>Oh, Google is rolling out Gemini 3 and a free Cursor. Well then. Strong. No need to hustle investors. <br> <a href="https://antigravity.google/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://antigravity.google/</a></p> <p><a href="https://t.me/chillhousetech/1174?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So I was reading a little piece about what they gave the Nobel Prize in… https://darkolorin.com/posts/1173.html https://darkolorin.com/posts/1173.html Mon, 17 Nov 2025 15:53:46 GMT <p>So I was reading a little piece about what they gave the Nobel Prize in physics for. Well, they gave it for demonstrating the quantum tunneling effect on a macro scale. Which, basically, isn’t the first time quantum effects have been found at a macro scale, but in the pop-sci crowd it spread like “oh wow, soon we’ll be watching cats that are ‘both dead and alive’.” But our text isn’t about that. In the overall slice of quantum-mech news it reminded me of an article (they even mentioned it recently, specifically in the context of this Nobel thing, on n+1)</p> <p>Karl Popper, *Three Views Concerning Human Knowledge*. (<a href="http://www.dpi.inpe.br/gilberto/cursos/cst-311/popper_threeviews.pdf" rel="noopener">http://www.dpi.inpe.br/gilberto/cursos/cst-311/popper_threeviews.pdf</a>)</p> <p>The guy was basically into philosophy of science, I recommend reading him. So this particular note introduced three concepts of scientific knowledge. It’s all about how quantum mechanics is so convenient and works so well, but where the limits of its applicability are — nobody really understands.</p> <p>1/ Essentialism (ultimate explanation by essences) <br> 2/ Instrumentalism (theories as instruments) <br> 3/ C <br> ritical realism (conjectures, truth and reality)</p> <p>Essentialism (Aristotle): <br> 1/ The goal of science is to know the deep essence of things <br> 2/ You’ve learned the essence, now it should preferably explain everything <br> 3/ Belief in an ultimate explanation is essential: one day we’ll reach a level where there’s nothing left to explain, because we found the “true nature” of the object</p> <p>Critique: <br> 1/ Found it and called it a day — kinda meh as a concept <br> 2/ Even if we accidentally guessed the real “essence,” we can’t prove it or check it repeatedly because any theory remains a hypothesis</p> <p>Instrumentalism: <br> 1/ Theories are a set of tools, an “abacus,” or algorithms for predicting observations <br> 2/ The question “is the theory true?” is considered either meaningless or secondary; what matters is how well it stamps out predictions <br> 3/ Tests are everything, it works значит збс</p> <p>Critique: <br> 1/ Instrumentalism can’t explain the success of science: why do theories made just as “tools” have such astonishing generality and precision? <br> 2/ If it works, I don’t give a damn how or what it asserts, this theory — a dangerous effect in science (self-satisfaction and lack of connection to the real world)</p> <p>Conjectures, truth and reality <br> 1/ The goal of science remains: to give a true (or close-to-true) description of the world; we stay realists <br> 2/ But: we can never be sure we’ve reached the truth; at best we can show pretty reliably that some theories are kinda crap overall (don’t work) <br> 3/ Theories are bold hypotheses (conjectures): very informative guesses about the world that can’t be finally verified, but can be seriously tested and you can try to refute them <br> 4/ There is no “final level of reality”: any level can be explained by a more general theory;</p> <p>And now my main question: doesn’t this remind you of anything? AI? Deep learning?</p> <p>Review finished (part two later).</p> <p><a href="https://t.me/chillhousetech/1173?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Feel like a YC partner: - watch video pitches and guess if they took the… https://darkolorin.com/posts/1171.html https://darkolorin.com/posts/1171.html Sat, 15 Nov 2025 10:17:47 GMT <p>Feel like a YC partner:<br> - watch video pitches and guess if they took the startup<br> - what survived, what didn’t<br> - and that kind of stuff<br> In short, perfect for some fun :)<br> The game’s here (<a href="https://ycarena.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ycarena.com/</a>)</p> <p><a href="https://t.me/chillhousetech/1171?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sometimes I feel like posting a couple of friendly job openings. https://darkolorin.com/posts/1170.html https://darkolorin.com/posts/1170.html Thu, 13 Nov 2025 13:35:15 GMT <p>Sometimes I feel like posting a couple of friendly job openings. I had Andrey Volodin back at Prisma, and now they’re doing volumetric vids, splats, and all that. If anyone’s interested — I recommend them.</p> <p>Lots of roles open at Gracia <br> 📍 <br> London, UK <br> // can start remotely</p> <p>As part of us rethinking the usual roles, we added an experimental position to the list: an AI-oriented product. Not sure if candidates like that even exist yet, share it with your enthusiast friends <br> 👉 <br> 👈</p> <p>C++ Core Engineer (<a href="https://store.gracia.ai/careers/cpp-eng?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://store.gracia.ai/careers/cpp-eng</a>) <br> [GPU Background]</p> <p>R&amp;D Engineer (<a href="https://store.gracia.ai/careers/rnd-splats?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://store.gracia.ai/careers/rnd-splats</a>) <br> [Computer Vision / 3DGS Background]</p> <p>R&amp;D Engineer (<a href="https://store.gracia.ai/careers/rnd-genai?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://store.gracia.ai/careers/rnd-genai</a>) <br> [GenAI Background]</p> <p>Product Lead (<a href="https://store.gracia.ai/careers/ai-first-product-manager?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://store.gracia.ai/careers/ai-first-product-manager</a>) <br> [AI-Oriented]</p> <p>QA Engineer (<a href="https://store.gracia.ai/careers/qa-eng?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://store.gracia.ai/careers/qa-eng</a>) <br> [Automation]</p> <p>Marketing Lead (<a href="https://store.gracia.ai/careers/marketing-lead?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://store.gracia.ai/careers/marketing-lead</a>) <br> [Tech PR Background]</p> <p>More details in the channel (<a href="https://t.me/beyond3sigmas?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/beyond3sigmas</a>) <br> and you can toss your CV into the comments there too</p> <p><a href="https://t.me/chillhousetech/1170?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The other day I was poking around YC, and they’ve got this page now — kinda… https://darkolorin.com/posts/1169.html https://darkolorin.com/posts/1169.html <p>The other day I was poking around YC, and they’ve got this page now (<a href="https://www.ycombinator.com/launches?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.ycombinator.com/launches</a>) — kinda like Product Hunt. Pretty handy btw: description, upvotes, all that stuff. Only for locals though. And I stumbled on a very important project. <br> CHAD IDE (<a href="https://www.ycombinator.com/launches/OgV-chad-ide-the-first-brainrot-ide?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.ycombinator.com/launches/OgV-chad-ide-the-first-brainrot-ide</a>) <br> And I think the video is just pure vibes. More and more, funds are turning into media and investing in “moments” like cluely.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1169?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> What and how I try to do in a startup or a team up to 20 people. https://darkolorin.com/posts/1168.html https://darkolorin.com/posts/1168.html Tue, 11 Nov 2025 15:19:35 GMT <p>What and how I try to do in a startup or a team up to 20 people. Before any PMF, scaling strategy for “humans” for different tasks, and all that. So to speak, the foundational stuff:<br> 1/ horizontal structure with no managers, basically the company has one manager — the CEO — and even that’s only because he often answers the question “where do we run”<br> 2/ everyday results. We have very few fixed processes, I absolutely don’t give a damn where tasks are kept, what trackers, docs, etc. I just try to make sure everyone posts in the status channel (we’re on Discord) every day what they did, even if it’s nothing (I drill this in from day one — helps like hell with transparency and self-control). And also at the start of every week everyone writes a plan of what they’re gonna do. If daily status is about what you did, weekly is preferably about the product.<br> 3/ hiring only through referrals whenever possible. I’m of the opinion a startup isn’t even worth starting if you don’t know who you’re doing it with.<br> 4/ the desire to do something in what the company decided to do almost always beats hunting for people with a ready-made hard-skill profile. You can learn anything (I hope lol, mistakes happen).<br> 5/ very frequent positioning changes and a lot of flexibility in that sometimes you need to throw everything out — all your work. Small teams aren’t about clear tasks and eternal value. We try to check everything faster through investors and partners to see feedback; in consumer sometimes a few hundred organic users is enough (just don’t check the traffic pls).<br> 6/ no schedule, except “be reachable when everyone’s around”, weekends whenever you want, whenever<br> 7/ being able to take criticism of the result, not the person, is an important skill and I try to look for it in people<br> 8/ all the equity, salaries are fully standard by the market. Like, easily googlable. And we’re roughly the same. There are exceptions of course.</p> <p><a href="https://t.me/chillhousetech/1168?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> We’re doing pretty much the same thing. https://darkolorin.com/posts/1167.html https://darkolorin.com/posts/1167.html Mon, 10 Nov 2025 13:28:57 GMT <img src="https://cdn4.telesco.pe/file/GcMvf4CN7ieJc9UsSp6KN4oWEvyvMOuXMocXu_JHoZEfE_lFdCdhZpU7ouy5Tfbj2BO6P-rQBcuKS28IL5ko9S3GRhR4tjEUShclLsmkex-u9PBYVl52blg0ScDsEVwXxLRMd7_pg7WZ7CLwzlKIYV4M0hYzqnnUCNVyZibzQqubya0YPur24spp6E2IvH9KmeBonXjT7Seb_nwz6-J5LK_0SrJzebAMWW4_IeelzGGQ4tURqKxgsaenaO6RSqfmxrRuX7eSbNS_nR-XUOdcJr_m_cSF_ams3kcF5EBg69GSSz5IWkC0YL4tNwl4MSpD7b9dKsAbpe0PVSWR-zA1ZA.jpg" alt="" loading="lazy"> <p>We’re doing pretty much the same thing. About how probably the hottest startup right now hires.</p> <p>Source (<a href="https://joincolossus.com/article/inside-cursor/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://joincolossus.com/article/inside-cursor/</a>)</p> <p><a href="https://t.me/chillhousetech/1167?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Pretty often I hear about the American communication style. https://darkolorin.com/posts/1166.html https://darkolorin.com/posts/1166.html Sat, 08 Nov 2025 15:48:26 GMT <img src="https://cdn4.telesco.pe/file/D8kb5zFikt0siNrAD0GtivZoUWMRp_IiUeaTgmGTXsTnEYSc8s8mz3OaH47E3UCQ2SNDzhEzPRhtHWGXOq_YsFpFKLlnLt-st8qlTU_Ixsm4u2PWqKSfoElUVc0uQejgZmqrsJykv_TRB1G2yEB4tUI8ct4MqTjL6gjpDbEgluFdDAWolg-JYTKZCEYCxYSVgz0qi990HDaKjdTjeFT2HoNvrWlWzpihg1Znvjn-6JnQPOG1t-cTN83wM32RNZMg1EUaxXIgUNlYpKCvv0ayWpaxBGkp69QBHEKUmACP4ZQFww_A2NfWqpuxl-5_tND8a1TASB6ZLma2E0gM7MGpaw.jpg" alt="" loading="lazy"> <p>Pretty often I hear about the American communication style. Here’s a vivid example. In other words: the higher your responsibility, the more direct the style. Yours truly.</p> <p><a href="https://t.me/chillhousetech/1166?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Just a reminder that projects like cocoon by Pavel Valeryevich and other… https://darkolorin.com/posts/1165.html https://darkolorin.com/posts/1165.html Wed, 05 Nov 2025 09:59:45 GMT <p>Just a reminder that projects like<br> cocoon by Pavel Valeryevich (<a href="https://t.me/s/cocoon?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/s/cocoon</a>)<br> and other similar ones are basically<br> this (<a href="https://www.runpod.io/product/runpod-hub?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.runpod.io/product/runpod-hub</a>)</p> <p>Important questions you should be asking yourself:<br> 1/ who provides the SLA (probably awesome to use something that’s not even guaranteed to work properly)<br> 2/ who’s gonna give you benches for the corresponding endpoints, like<br> this kind of thing (<a href="https://artificialanalysis.ai/models/deepseek-v3-1-terminus-reasoning/providers?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://artificialanalysis.ai/models/deepseek-v3-1-terminus-reasoning/providers</a>)<br> 3/ and what kind of inference is even in there, params, and so on, like<br> something like this (<a href="https://www.baseten.co/blog/activity-logging-llm-metrics-dashboard-customization/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.baseten.co/blog/activity-logging-llm-metrics-dashboard-customization/</a>)</p> <p>Not even getting into the philosophical questions: why the hell is there<br> a casino<br> crypto<br> in this whole GPU rental-in-a-normal-cloud scheme. Why not just buy a ton of GPUs and rent them out the normal way? Maybe there’s no money to buy them. Eh.</p> <p>And don’t even think that normal users are gonna be renting out their GPUs there. Please don’t :)</p> <p><a href="https://t.me/chillhousetech/1165?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One more thing. https://darkolorin.com/posts/1164.html https://darkolorin.com/posts/1164.html Mon, 03 Nov 2025 17:31:07 GMT <p>One more thing. Everyone teaching vibe-coding, prompting, and other crap in AI-usage courses is just blatantly cashing in on you. There’s no such skill as vibe-coding or prompting. Only model makers can give any real recommendations because they know what the dish was cooked from. In any other case they’re selling you “how to order food from a restaurant.”</p> <p><a href="https://t.me/chillhousetech/1164?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well now this is already fine, a real pro director (seems like he made Saw)… https://darkolorin.com/posts/1163.html https://darkolorin.com/posts/1163.html Sat, 01 Nov 2025 17:55:06 GMT <p>Well now this is already fine, a real pro director (seems like he made Saw), did a Nike ad (<a href="https://youtu.be/EzTxWRpObhQ?si=s6DsNWmCj--WPQ0K&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/EzTxWRpObhQ?si=s6DsNWmCj--WPQ0K</a>) entirely with generative tools. What do you think? Of course without Nike’s consent, but oh well :)</p> <p><a href="https://t.me/chillhousetech/1163?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The power is in your own model, trust me. https://darkolorin.com/posts/1162.html https://darkolorin.com/posts/1162.html Wed, 29 Oct 2025 21:02:01 GMT <p>The power is in your own model, trust me. Everyone who grows up will have to train their own, so data collection is the critical part. Verticalizing the stack (the more you own the better) is the most important thing, no matter what they tell you.<br> Using Cursor as an example <br> <a href="https://cursor.com/blog/composer?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://cursor.com/blog/composer</a></p> <p><a href="https://t.me/chillhousetech/1162?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, there’s this whole startup area: oooooh, now we’ll whip up a bundle of… https://darkolorin.com/posts/1161.html https://darkolorin.com/posts/1161.html Wed, 29 Oct 2025 01:50:07 GMT <p>So, there’s this whole startup area: oooooh, now we’ll whip up a bundle of prompts and if-then-else. Well yeah. Google keeps putting out more or less sensible experiments (I’m pretty sure it’s a small team doing it just to tick KPIs, maybe it pops off). For now it’s US-only. Seems you specifically need a Google account from there to try it.</p> <p>The service analyzes your website and suggests you slap together creatives for ads (pretty obvious where). Overall, it’s top. In the dashboard and everywhere else they’ll finish the integration, I think. If you zoom out, basically all these startups will get built inside корпорации, a couple will get bought, and finally maybe 1 will be able to do something on its own. But for now AI by itself doesn’t deliver enough value. You need the bundle.</p> <p>Watch the video (<a href="https://youtu.be/rsWPISYv6tQ?si=Ia1w7Gz12w-vO1VJ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/rsWPISYv6tQ?si=Ia1w7Gz12w-vO1VJ</a>) <br> Poke around in the US (<a href="https://labs.google.com/pomelli/about/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://labs.google.com/pomelli/about/</a>)</p> <p><a href="https://t.me/chillhousetech/1161?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A decent report came out, from a fund I know nothing about and from a guy I… https://darkolorin.com/posts/1160.html https://darkolorin.com/posts/1160.html Sun, 26 Oct 2025 22:56:13 GMT <p>A decent report came out, (<a href="https://blog.sparklinecapital.com/wp-content/uploads/2025/10/sparkline-ai-capex.pdf" rel="noopener">https://blog.sparklinecapital.com/wp-content/uploads/2025/10/sparkline-ai-capex.pdf</a>)<br> from a fund I know nothing about and from a guy I also don’t know. But the arguments are very reasonable.<br> 1/ big tech started dumping insane money into datacenters aka railroads (already 400 yards and expected to be even more)<br> 2/ past examples of this kind of spending by companies that did that (internet infra and railroads, basically) led to not such huge returns (well compared to when those companies rode on someone else’s infra)<br> 3/ the biggest returns went to the ones standing next to it or the ones using it (in our case that’s the условные Palantir, Walmart etc) or to small players with cheap stocks</p> <p><a href="https://t.me/chillhousetech/1160?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Obsession with an idea, or the prepared mind principle. https://darkolorin.com/posts/1159.html https://darkolorin.com/posts/1159.html Sat, 25 Oct 2025 17:13:11 GMT <p>Obsession with an idea, or the prepared mind principle. <br> Don’t know about you, but for a long time I was in that category of founders or people who were obsessed with their ideas (doesn’t really matter whether it’s a business or a product or whatever else). And so, naturally, I met a ton of similar people too. Literally most founders of that type. But a minority has this quality called a “prepared mind.”</p> <p>Simple example: say you want to build a time machine and you don’t give a damn about anyone or anything. Everyone tells you, like, you’re an idiot. And you just keep smashing forward. There’s one important detail though. To build a time machine it’d be nice to be able to answer questions about general relativity. Which means, well, you get it (you need to understand at least something there).</p> <p>It’s the exact same crap in the startup world. But don’t forget this applies to everyone in it, including investors. A huge share of people just straight-up doesn’t understand what they’re doing, and there are the same kind of people who believe in it (well, because they also don’t understand, and belief is kind of random), and sometimes you end up with total nonsense.</p> <p>So yeah, don’t confuse a prepared mind with experience. Experience is only part of it. It’s not enough by itself.</p> <p>In other words, watch the criticism, and if: <br> - you ran into reasonable criticism <br> - you don’t have a logically grounded answer <br> - think a bit more seriously than “some loser across from me”</p> <p>Example: you decided to make an Uber for “insert something.” But for some reason it doesn’t exist yet. Those questions like “Why now and not 10 years ago” are exactly the ones that require serious preparation. And “people love this” doesn’t cut it here.</p> <p>Like some guy in France said (Pasteur, I think): <br> “Chance favors only the prepared mind” <br> — <br> “Luck favors only the prepared mind.”</p> <p><a href="https://t.me/chillhousetech/1159?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Damn, I got hooked, awesome site where you can look at maps and borders over… https://darkolorin.com/posts/1158.html https://darkolorin.com/posts/1158.html Fri, 24 Oct 2025 16:15:32 GMT <p>Damn, I got hooked, awesome site where you can look at maps and borders over time, like from ancient times to more or less our days, and a bunch more data about countries, events, and people. Only the poster-selling is annoying <br> 🙂 <br> <a href="https://www.oldmapsonline.org/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.oldmapsonline.org/</a></p> <p><a href="https://t.me/chillhousetech/1158?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Beta release just rolled in. https://darkolorin.com/posts/1157.html https://darkolorin.com/posts/1157.html Thu, 23 Oct 2025 18:36:15 GMT <p>Beta release just rolled in. You can safely download our Mac app from the site, support is only for M-series processors. <br> We’d really appreciate feedback. Everything inside is ours. Inference, downloading, client.</p> <p>Why you need this: inside there’s an Apps section where we’ll be posting different workflows. In the future you’ll be able to easily repeat them on your side by signing up on the platform. A lot of people don’t know what local models are capable of. For now the list there isn’t that big, quantization will come a bit later.</p> <p>We collect analytics only to count DAU and activity. Next to the checkbox it says what and how. Like, share, retweet, as they say. We’ll be happy if you or your friends give it a try. <br> <a href="https://trymirai.com/chat-for-mac?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/chat-for-mac</a></p> <p>P.S. should we open-source the code? It’s a bit dirty :)</p> <p><a href="https://t.me/chillhousetech/1157?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If you still haven’t watched it, this is probably the best interview about AI… https://darkolorin.com/posts/1156.html https://darkolorin.com/posts/1156.html Wed, 22 Oct 2025 00:48:25 GMT <p>If you still haven’t watched it, this is probably the best interview about AI and what’s what in the last few years. In my humble opinion, of course. Andrej Karpathy and Dwarkesh. Long story short, worldview-wise I’m probably closest to him. Definitely not to Sam or Dario. Those guys are probably working the investors, and hopefully they just need to push that kind of narrative. But Andrej doesn’t need to.</p> <p>What, so to speak, hooked me:<br> - AGI in another ten years; he was betting on 20 years to AGI back when he was at OpenAI in 2015 (almost the biggest pessimist on the team), and also that it won’t change anything fundamentally (well, except education in his opinion)<br> - all current agents are “non-working shit,” and he does lay out why, and pretty damn well (no continual learning, no proper memory, etc.)<br> - RL is “horrible shit,” well because people and вообще it seems even animals don’t really use it much (arguments about genetics and how it’s supposedly there already)<br> - we’re waiting for next-gen algorithms like we’ve seen over the last 30 years; everything will be better, right now it’s still not enough (but he does seem to believe in deep learning)<br> - he believes AI will change education, which is of course reasonable, since an individual tutor is more effective, and a huge number of different approaches can be combined on the fly</p> <p>Bottom line, there’s a ton of ideas in there, everything is explained very clearly, and most importantly without hype. Highly recommend.</p> <p>Original transcript (<a href="https://dwarkesh.substack.com/p/andrej-karpathy?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://dwarkesh.substack.com/p/andrej-karpathy</a>)<br> Video on YouTube (<a href="https://youtu.be/lXUZvyajciY?si=F8Dt8X4XUj-L_uTi&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/lXUZvyajciY?si=F8Dt8X4XUj-L_uTi</a>)</p> <p><a href="https://t.me/chillhousetech/1156?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> B R O W S E R? https://darkolorin.com/posts/1155.html https://darkolorin.com/posts/1155.html <p>B R O W S E R ?<br> Looks like OpenAI are announcing their own browser.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1155?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> TLDR: everyone spends way more on servers than they earn. https://darkolorin.com/posts/1154.html https://darkolorin.com/posts/1154.html Tue, 21 Oct 2025 14:10:50 GMT <p>TLDR: everyone spends way more on servers than they earn. Reselling GPU hours is a super business for now. Recommend reading the article (<a href="https://www.wheresyoured.at/costs/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.wheresyoured.at/costs/</a>) — yeah it’s guesswork, but the skilled kind, with links to sources. <br> P.S. the blog is generally good</p> <p><a href="https://t.me/chillhousetech/1154?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Attention to detail. https://darkolorin.com/posts/1153.html https://darkolorin.com/posts/1153.html <p>Attention to detail. Love it. But it’s an expensive kind of pleasure. And of course a lot of people will cancel out the “shittiness” of the iOS update because of that. <br> When we’re building a startup I usually try right at the doorstep to say: avoid topics and features like this. But yeah, it’s cool, sure. The arrow at low battery starts moving not smoothly, but in jumps.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1153?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> However, Waymo is expanding really fast. https://darkolorin.com/posts/1152.html https://darkolorin.com/posts/1152.html Wed, 15 Oct 2025 12:38:59 GMT <img src="https://cdn4.telesco.pe/file/B8LjgVPEMFAHxYmyC-T4D5Q3bVU4uqUPPysu-qDk58k4Ititiw1ssk8IIwoJwVc8-VtKgJXoMjQL0nhvnDB0d428iXJkSJUJeGsFkpmGyBAPCurXVqPCpYXiexRQDVxB_TYDNpgiJNqvDlkX66J16qApqV6yugaWs58hdG6-ZZvglx2PbHPBrApCQtGQUJTUl5n_uyMJJcgynoMa50E2p8W13sCu8boVl4C-z4Styko60dE-Hs7A1kJH3juZeiZjh_zB1feuJ0LsuzrtE3Ii0sLRa6djxdjXJjo5jD8hCGJ8X6vOZJmdr11a4mw9LL3mS7QIod03RITaTc7_5aDnTw.jpg" alt="" loading="lazy"> <p>However, Waymo is expanding really fast. It even says Tokyo is up next.</p> <p><a href="https://t.me/chillhousetech/1152?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Column “read, but not fully understood” https://darkolorin.com/posts/1151.html https://darkolorin.com/posts/1151.html Wed, 15 Oct 2025 08:44:06 GMT <p>Column “read, but not fully understood”</p> <p>1/ really like the idea of <br> model merging (<a href="https://crisostomi.github.io/blog/2025/model_merging/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://crisostomi.github.io/blog/2025/model_merging/</a>) <br> and the fact that people are even working on this at all, there are so many models and it’d be cool to be able to merge what they can do (at least in one domain)</p> <p>2/ <br> no new ideas, there are new datasets, (<a href="https://blog.jxmo.io/p/there-are-no-new-ideas-in-ai-only?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.jxmo.io/p/there-are-no-new-ideas-in-ai-only</a>) <br> and right away some architecture rolls up, <br> mamba3 (<a href="https://openreview.net/pdf?id=HwCvaJOiCj" rel="noopener">https://openreview.net/pdf?id=HwCvaJOiCj</a>) <br> (this isn’t transformers basically), with, like, improvements, but the quality metrics honestly didn’t go up that much, but they promise it’ll run faster</p> <p>3/ from philosophy trying to finish off Marcus Aurelius “Meditations” (“alone with yourself”)</p> <p>4/ about causality, statistics, connections between phenomena, basically <br> an excellent chapter (<a href="http://web.cs.ucla.edu/~kaoru/BIB5/ch2.pdf" rel="noopener">http://web.cs.ucla.edu/~kaoru/BIB5/ch2.pdf</a>) <br> from the book The Book of Why</p> <p>5/ sometimes it’s interesting to read <br> what respectable analysts wrote a year ago (<a href="https://stratechery.com/2024/enterprise-philosophy-and-the-first-wave-of-ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://stratechery.com/2024/enterprise-philosophy-and-the-first-wave-of-ai/</a>) <br> about AI, helps you read and listen soberly to statements by big managers</p> <p>6/ <br> about how they made the legendary (<a href="https://share.google/pexNwvYmAtedMP2fX?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://share.google/pexNwvYmAtedMP2fX</a>) <br> game “Prince of Persia” on consoles, this is straight-up gold for all managers and anyone interested in creative solutions and how to act under tech constraints. Didn’t find a PDF of the book.</p> <p><a href="https://t.me/chillhousetech/1151?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> 🤫 What a lot of global companies want is to pay remote employees and… https://darkolorin.com/posts/1150.html https://darkolorin.com/posts/1150.html Tue, 14 Oct 2025 09:04:33 GMT <p>&quot;&quot;&quot;<br> 🤫<br> What a lot of global companies want is to pay remote employees and freelancers legally and without insane fees.<br> Typical case: the company is in the UAE or Cyprus, and the dev is in the CIS, Serbia, and Georgia. Every payroll run is like a quest: banks hold transfers, shove them into hardcore compliance, the team complains about delays in getting paid.<br> When your team is up to five people — it’s tolerable. With fifteen — it’s time to change the approach.<br> ⭐️</p> <p>Platform (<a href="https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech" rel="noopener">https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech</a>)<br> 4dev.com (<a href="https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech2" rel="noopener">https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech2</a>)<br> automates payouts to remote employees and freelancers — worldwide and in line with the law:<br> · One contract for all employees. No need to prep invoices and contracts yourself<br> · Payouts to 100+ countries, including the CIS. Literally a couple clicks<br> · Automated paperwork. Invoices work for accounting, audits, due diligence<br> 💵<br> You can pay the team in crypto and instantly get invoices.<br> No subscriptions or hidden charges. The platform only takes a 1–3% fee from each payout. The more you pay the team — the lower your fee.<br> Book a demo (<a href="https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech2" rel="noopener">https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech2</a>)<br> →<br> on the call we’ll calculate the payout cost for your company and answer all questions.<br> If you have a company in the US —<br> 4dev.com (<a href="https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech2" rel="noopener">https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech2</a>)<br> can become a must-have solution that replaces relocating specialists on an H1B visa. The US government just the other day introduced additional fees of $100k on companies that relocate employees on this visa.<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/1150?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, before a week of nice integrations. https://darkolorin.com/posts/1149.html https://darkolorin.com/posts/1149.html Sun, 12 Oct 2025 21:29:31 GMT <p>Well, before a week of nice integrations. A small post about efficiency. <br> So there’s this myth, fable, Twitter trend, etc.: that to succeed you gotta work a shitload. And then there are the opponents — like, you gotta work with your head. Overall, both sides rarely justify their take properly. So I’ll throw in my two cents. Especially useful for hiring managers.</p> <p>When people talk about Jensen Huang (Nvidia’s CEO) and how he “grinds.” In his case it’s a unique skill where time spent converts directly into results — and at that level, too. For most people it’s not like that. And it’s important for you to understand what you actually want.</p> <p>The biggest chance of converting time into results is when a person does the simplest thing and just one function. That’s why corporations, as they grow, try to hire and chant the mantra: learn to be the best at one thing and then everything will be awesome. Because when you know one feature really well, you convert time into results perfectly fast. And they try to keep those ICs and even managers right where they are and not let them grow.</p> <p>Sometimes it’s the opposite: there’s no scope of work, nothing’s clear. And then building a time-to-result converter is hard — you have to develop a system yourself so it works out. There are no recipes, unlike working on button colors or purchase conversion. And that’s usually where the growth is.</p> <p>If with growth you still see how results (revenue, users, etc.) increase with the time you put in — congrats, you’re top-tier. You’ll always be in demand.</p> <p>I know only a handful of people who, at the level of top managers or founders, are that type of person. Usually they’re fine on money and bad in their personal life :)</p> <p><a href="https://t.me/chillhousetech/1149?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Bot or what? https://darkolorin.com/posts/1148.html https://darkolorin.com/posts/1148.html Sun, 12 Oct 2025 21:17:44 GMT <img src="https://cdn4.telesco.pe/file/al-AmOx9VJum1SoWy8kuKDInN1ni-ehkFTTQy5Dqzum9Yxa9tzABWT6B9qegQKHFt3gsURl683fqZDVi0SQmSySFdZ3OkQyTFNyiwFl5kN4-fmAkIs4LVgs64MVEJBGXisLthe1wVuIS1fVRll-WtLkPiExvL_l5sBbv5UxntXTOaCuEPoRBVYJP_tZO5DwKHKF2Yds6DNM04DwYor5ritl9fPt9H9PHTFltXlEgO8M4dbMfXb3Y8nqZNRM6AV37PZeTfJkbv6UYgFSJb1FjjOqzzy8JIV-eIYKrOYlbt1fLA72CLcJJB5lX6h7yCRfN0-ZOR8aql07nKGTPiFtMZg.jpg" alt="" loading="lazy"> <p>Bot or what? This is the same vibe-code app from half a year ago. Or the naming just hit right?<br> Checked LinkedIn — and there are even 5 mutual connections. Oof.</p> <p><a href="https://t.me/chillhousetech/1148?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Will ChatGPT become the new app store? https://darkolorin.com/posts/1147.html https://darkolorin.com/posts/1147.html Tue, 07 Oct 2025 00:51:53 GMT <p>Will ChatGPT become the new app store? Doubt it.<br> The idea is logical: you need more long scenarios, more content (why, I hope, is obvious), more reasons to stay inside the ecosystem.<br> But let’s remember what we’ve already seen, and what questions you should ask yourself in advance:<br> 1/<br> Telegram<br> already tried to launch its own AppStore. And it even kinda worked for a couple months. Remember that boom of crypto scams? The problem was simple:<br> no real recommendation system.<br> No traffic distribution at all, just invites and external UA from bots lol. And on top of that — very limited options for developers.<br> 2/ “<br> What about WeChat?<br> ” Yeah, WeChat is cool. But it’s important to understand the context: a closed internet environment, language barrier, dense interfaces (historically), government regulation. And the main thing is that everything rests on<br> payment integration.<br> 3/<br> Google<br> ? Not mini-apps either, but at least it’s clear<br> how distribution works.<br> There are rules of the game: PageRank, keywords, content quality, links. And what do we know about ChatGPT Apps? Right, we don’t know shit.<br> 4/ Which brings us to the main question:<br> how do you even compete?<br> How is a developer supposed to influence traffic distribution at all?<br> 5/ Some people will see this announcement as an Apple AppStore analogue. But no. Apple has APIs to work with the hardware. ChatGPT doesn’t. Your main “super app” is your phone. You don’t need a new one.<br> I don’t think ChatGPT should be getting into the engagement-via-apps game. Content is a much more logical direction.<br> Though overall it’s a decent move, respect to the team. The speed at which OpenAI ships new stuff is just insane.<br> We’ll see where this whole “GPTs” story leads. Integrations are needed, yeah.<br> But the real power is always in<br> the long tail.<br> I want to be wrong. I want a new app store.</p> <p><a href="https://t.me/chillhousetech/1147?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In the current OpenAI presentation you can really feel how they’re copying… https://darkolorin.com/posts/1146.html https://darkolorin.com/posts/1146.html Mon, 06 Oct 2025 17:43:34 GMT <p>In the current OpenAI presentation you can really feel how they’re copying Apple’s delivery style. Examples from different countries. Glorifying adjectives, a packed house, now back to Sam. <br> <a href="https://www.youtube.com/live/hS1YqcewH0c?si=17uy6EZlg06JyOeu&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/live/hS1YqcewH0c?si=17uy6EZlg06JyOeu</a></p> <p><a href="https://t.me/chillhousetech/1146?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> That very expensive Mira Murati startup has launched its first product. https://darkolorin.com/posts/1145.html https://darkolorin.com/posts/1145.html Wed, 01 Oct 2025 19:17:23 GMT <p>That very expensive Mira Murati startup has launched its first product. Tuning, training little models. I consider the compute resale business officially launched. <br> <a href="https://thinkingmachines.ai/tinker/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://thinkingmachines.ai/tinker/</a></p> <p><a href="https://t.me/chillhousetech/1145?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And here’s OpenAI going commercial with Stripe https://darkolorin.com/posts/1144.html https://darkolorin.com/posts/1144.html <p>And here’s OpenAI going commercial with Stripe</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1144?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> New release from Anthropic Sonnet 4.5 https://darkolorin.com/posts/1143.html https://darkolorin.com/posts/1143.html Mon, 29 Sep 2025 17:24:15 GMT <img src="https://cdn4.telesco.pe/file/rlHwYanLiGQ82eMRH-6XB4dql7Ac1igPXV2mB7uo5mPQgkKdgupbeaUqxerYaxm9TgbraEfvlTKmtkvWqv-pwcfs32A3dm1epHaAPbmXa7tSbaMWJMD-c26DxwnAmk6qTHqp4z0hEchkz6cViIQKgwMQXlnUXieQj35p0d3LObhhVng7P3Yq_5b_g1DCGGKfWLqixRXpqkCu6Nmm7SRTNvX5N3PEj0wBwv8iPsyQWQ76Svy6KQn9R-Ef7-Wt9SfB0clrRja16b1pBJSB55DjOsJ1pdrHKQ1lsyJRtEOqO9te2qvTCczwWEgkSckvYUoN-sQuRyBa9lQFtksJ4PnGIA.jpg" alt="" loading="lazy"> <p>New release from Anthropic <br> Sonnet 4.5 (<a href="https://www.anthropic.com/news/claude-sonnet-4-5?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/news/claude-sonnet-4-5</a>)</p> <p><a href="https://t.me/chillhousetech/1143?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Tip for the week https://darkolorin.com/posts/1142.html https://darkolorin.com/posts/1142.html <p>&quot;&quot;&quot;<br> Tip for the week<br> &quot;&quot;&quot;</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1142?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Saturday stuff. https://darkolorin.com/posts/1141.html https://darkolorin.com/posts/1141.html Sat, 27 Sep 2025 12:23:35 GMT <p>Saturday stuff. Sometimes you chat with or randomly run into a founder. And you get the feeling the main reason he’s doing a startup is basically something like this set:<br> - yeah, working for a corp is a drag, it’s boring, plus I won’t pass an interview for a decent role, and nobody’s gonna just come headhunt me for no reason<br> - want someone to hand over some cash for a salary and to build some kinda cool stuff<br> - walk around and travel to meetings on the company’s dime<br> - constantly post about how hard it is to raise cash<br> In the end, after 2–3 years, write that it was super hard, but had to go back to employment. And then the cycle again.<br> Thanks, Venture, for these opportunities.</p> <p><a href="https://t.me/chillhousetech/1141?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Ladies and gentlemen, we’re putting together a demo app for MacOS, and then… https://darkolorin.com/posts/1140.html https://darkolorin.com/posts/1140.html Mon, 22 Sep 2025 22:00:12 GMT <p>Ladies and gentlemen, we’re putting together a demo app for MacOS, and then all the features of our engine will start showing up in there. For now it’s in alpha mode, but I’ll be happy if you give feedback and actually try using it. <br> It works pretty simply: download the .pkg and then, with no logins and other crap, you just download yourself a little model to try. I’m actively taking feature/bug requests. For the best bug and the best feature request, you get a gift from me. <br> Maybe the app’s code will be Open Source soon.</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1140?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The ads war has begun https://darkolorin.com/posts/1139.html https://darkolorin.com/posts/1139.html <p>The ads war has begun</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1139?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> An interesting report dropped from the OpenAI team about how people use… https://darkolorin.com/posts/1137.html https://darkolorin.com/posts/1137.html Tue, 16 Sep 2025 10:11:22 GMT <p>An interesting report dropped from the OpenAI team about how people use ChatGPT and what kinds of prompts, the audience, and so on.</p> <p>In short:<br> - it’s growing in developing countries with low GDP per capita this year compared to the previous one<br> - the charts in the report are, as usual, a mess (don’t know why)<br> - three types of prompts: asking, doing, expressing. What do you think the shares are?<br> Link to the study (<a href="https://www.nber.org/system/files/working_papers/w34255/w34255.pdf" rel="noopener">https://www.nber.org/system/files/working_papers/w34255/w34255.pdf</a>)<br> (straight to PDF)</p> <p><a href="https://t.me/chillhousetech/1137?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Top-tier article. https://darkolorin.com/posts/1136.html https://darkolorin.com/posts/1136.html Thu, 11 Sep 2025 13:59:08 GMT <p>Top-tier article. Everyone who’s into AI should read it. <br> Disclaimer: very technical. But it’s enough to read at least the problem description at the beginning. A lot of people don’t actually get this. <br> <a href="https://thinkingmachines.ai/blog/defeating-nondeterminism-in-llm-inference/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://thinkingmachines.ai/blog/defeating-nondeterminism-in-llm-inference/</a></p> <p><a href="https://t.me/chillhousetech/1136?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Be nice. https://darkolorin.com/posts/1135.html https://darkolorin.com/posts/1135.html Wed, 10 Sep 2025 16:25:41 GMT <p>Be nice. Or not. Be yourself. <br> In 2016, when Prisma was basically nosediving, I met with Tony Fadell. (<a href="https://en.wikipedia.org/wiki/Tony_Fadell?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://en.wikipedia.org/wiki/Tony_Fadell</a>) <br> The lead designer of the iPod and almost the lead iPhone guy under Steve, plus he later sold Nest to Google for like 3 bil or something. So yeah, a respected dude in the industry. I was 25 and very “toxic,” like, either a person knows their shit or they don’t. And during the meeting I applied that rule hard. The guy didn’t really get apps, AI, and so on, and his advice wasn’t relevant. So I took his advice like a bayonet, didn’t listen, even corrected him. He got offended. And after the meeting he wrote to a bunch of respected people of the same rank, like I’m an “asshole, don’t meet Alexey.” And at that moment every acquaintance of mine in the Valley was kind of lecturing me like “you can’t do that with people like that, you should always nod and listen,” aka American style of respect. But at some point this super simple thought hit me. Why respect people who go around telling industry folks about a 25-year-old guy like “the dude’s a moron.” Why should I respect them? I’d rather tell him what I think. And I’m not gonna kiss ass. And all that pomposity is for people who’ve got nothing behind them. If nothing worked out for you and you’ve got nothing to show. When you have resources and opportunities, there’s no point nodding at everything, you can just say “no, you’re wrong” and just exit the conversation, and it doesn’t matter who was sitting across from you. So yeah, anything’s possible if you understand what you’re doing and with whom you can do it. Everyone falls for results: when your metrics are flying up, people talk to you, and when you don’t have them, they don’t talk to you. In short, screw the rules and all that, use your own head.</p> <p><a href="https://t.me/chillhousetech/1135?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Can’t not share this, AI darwin AWARD! https://darkolorin.com/posts/1134.html https://darkolorin.com/posts/1134.html Mon, 08 Sep 2025 01:31:14 GMT <p>Can’t not share this, AI darwin AWARD! You know, the Darwin Award for the dumbest moves and all that — well, here it’s about rolling out AI.<br> <a href="https://aidarwinawards.org/nominees-2025.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://aidarwinawards.org/nominees-2025.html</a></p> <p><a href="https://t.me/chillhousetech/1134?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sunday stuff. https://darkolorin.com/posts/1133.html https://darkolorin.com/posts/1133.html Sun, 07 Sep 2025 12:51:04 GMT <p>Sunday stuff. What I’ve learned over the last 10 years, and when I was starting out I thought differently. <br> 1/ healthy fanaticism, faith in yourself and your efforts, connections, the ability to keep your emotions in check — that’s where everything starts and how the first steps get made <br> 2/ don’t pay attention to what people tell you today and tomorrow, think about where you’re running. Direction matters way more than in-the-moment feedback. <br> 3/ ambitious ideas inspire; it’s easier to do something “unreachable” than to assemble a truly cool team around the thesis “we’ll make money” <br> 4/ it’s impossible to focus on just one feature, one hypothesis. Pick several and switch from one to another fast. <br> 5/ processes are always second place (first place — you decide what you’ve got) <br> 6/ results are everything, doesn’t matter how much you suffered to get some thing nobody needs <br> 7/ talk more with the goal of hiring and working together, never been a case where acquaintances were a minus (either nothing, or a plus) <br> 8/ execution today matters more than instant results, but it’s important to understand why the hell you need these results over the long run; think 5–10 years ahead, even if it seems like bullshit (even if you’re not a leader and not a top manager — just thinking not one but two steps ahead is already a super skill) <br> 9/ it so happened that education matters, and getting it after 25 is basically unreal (very unlikely that knowledge will help, or even show up at all). When Alferov back in school used to say: 15 to 25 is the only age when you actually learn and acquire key knowledge and a way of thinking. <br> 10/ to step further in product or business you need a compound effect (what you did today should help you make the next step faster). Preferably have it everywhere, even in boring stuff. <br> 11/ working with your hands matters — even leaders should show what they do for the company with their own hands, not just with decisions (probably less important at the scale of thousands of people)</p> <p><a href="https://t.me/chillhousetech/1133?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Saw a friend post this. https://darkolorin.com/posts/1132.html https://darkolorin.com/posts/1132.html Thu, 04 Sep 2025 20:41:45 GMT <p>Saw a friend post this. (<a href="https://t.me/zborovskiy_x/61?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/zborovskiy_x/61</a>) <br> Feels like I left a pretty decent comment. What do you think? <br> ————————————————— <br> I’ll throw in my two cents. I lived in the US, worked with their investors, hired people. Please take this as purely a personal projection, but a pretty strong one, because my entire current work circle is from there.</p> <p>The problem is in the mindset and the approach to the market. In the CIS crowd the most important thing is your measurable skills: hard skills. All posts, opinions are based only on quality that lends itself to direct, understandable analysis. I grew up in that myself. And when someone runs into the fact that some person who isn’t that крутой in hards/skills managed to build a company worth a yard, all the stones get thrown at factors the author can’t change (place of birth, language, and so on).</p> <p>But in reality “button placement under a constrained budget and market” isn’t that important. What matters is the ability to lead a company and product in the right direction. Yeah, money plays a big role there. But the Valley didn’t start with huge money either, and lots of products launched earlier in Russia. But then the people who launched them couldn’t scale for a bunch of reasons, not just “California and a rich family.” A larger number of managers from the RU region lack:<br> - understanding of the market and where it’s going<br> - the ability to prioritize year 5 of development over year 2 (lack of long money is the main factor and reason here)<br> - shitty communication qualities<br> - the ability to sell their ideas badly<br> - inability to build relationships with the market and different cultures</p> <p>That’s why in these analyses you can read hidden envy and denial. Like “how did these losers pull that off, I’m the one who’s better.” But the ones who are better are actually them. Way higher level of mindset adaptation, faster reaction, faster to take risk (and this is just straight up bad in RU).</p> <p><a href="https://t.me/chillhousetech/1132?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Most of the successful startups in the current cohort are built on this… https://darkolorin.com/posts/1131.html https://darkolorin.com/posts/1131.html Thu, 04 Sep 2025 11:32:19 GMT <img src="https://cdn4.telesco.pe/file/ZF5OtV8NvU9pqMwdNElP4D5iaY19x7_5qxv08b2QD_-aWW2IeiWZuYQeSnFJxkCNBWyGprzosOpwwNG3oG22DUgTbDktEP1YDrmscl0Hp1_rEgrwugm-yGbeF4CG-Be2-1-chbLc8zKcNBNiXRvKHx8CsEM0qcXntZwPI7jd7No0nhyyqSbgfBfEtBVpH7qHYQui32aWn8ePY5hiJePPihVWy2YiWRJRcNEn9-k1LfqPfNSNvg7-3qR6pjJTUCYWl9GbRCpR7LT_RAAuYHKTO2ZFVhBmkRfttxx9-7H_pYK8FsAjClfU5sEi5-4QY7r9RQZ_TOCh1RLyfkzkTTp1xw.jpg" alt="" loading="lazy"> <p>Most of the successful startups in the current cohort are built on this principle.</p> <p><a href="https://t.me/chillhousetech/1131?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The famous “Google is gonna die soon.” All-time high again. https://darkolorin.com/posts/1130.html https://darkolorin.com/posts/1130.html Tue, 02 Sep 2025 20:16:58 GMT <img src="https://cdn4.telesco.pe/file/P_O-LJPnc2A0W0wjINX1nUvwntqnScixireHqcHFeiP9iwu8lKjNk1Vp7w2pNTWCk2RZVRGobsba9hKNgLFc1QsU9a4rYhvkrnoi3CEke5MS_AvHB2jr_yZguC5CzEHVkOejejFjUFw0q_3JbbHe-reVdKSoEwQISxmBGxTJGdwVUYdCCaAv-GkSLsur7q5vTyc8qAtquGvHGTIasH3Lr0Pg79Azf6TMNGYuwk_GsEOBd0I6ASpv_ICPpgZEIXWZ-iBput7gK6DwO3HwfzPH_pKhgs1JyPEnHFwrzhyL0UEXAMHpzMTv8HcE4FQCaJUZ1t9e0Xr48UCNm_hcLi1i2w.jpg" alt="" loading="lazy"> <p>The famous “Google is gonna die soon.” All-time high again.</p> <p><a href="https://t.me/chillhousetech/1130?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Recently I stumbled on an interview. https://darkolorin.com/posts/1129.html https://darkolorin.com/posts/1129.html Mon, 01 Sep 2025 22:33:34 GMT <p>Recently I stumbled on an interview. Doesn’t really matter whose, overall. But this idea came up, like infobiz motivation style: that, like, talent, luck depend on the number of attempts, the main thing is to work hard. Well, yeah. That’s exactly why people have such a hard time with quantum mechanics, or more precisely with probability theory. I suggest all such gentlemen go beat the casino.</p> <p><a href="https://t.me/chillhousetech/1129?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A “read, but not fully understood” column. https://darkolorin.com/posts/1128.html https://darkolorin.com/posts/1128.html Sun, 31 Aug 2025 12:41:31 GMT <p>A “read, but not fully understood” column. A bunch of stuff has piled up. <br> 1/ <br> a super damn enjoyable series of articles (<a href="https://lilianweng.github.io/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://lilianweng.github.io/</a>) <br> about AI, I like the delivery and the content, it’s pretty technically dense. Worth noting the author is a cofounder of Thinking Machines and before that VP AI Safety etc and R&amp;D at OpenAI <br> 2/ <br> Storonsky’s playbook (<a href="https://quantumlightcapital.com/playbooks/performance-management/talent-philosophy?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://quantumlightcapital.com/playbooks/performance-management/talent-philosophy</a>) <br> also been lying around. No idea how to apply it at a team-of-a-couple-dozen level. But if it’s suddenly useful to someone. <br> 3/ everyone who’s into A/B tests and the whole “one thing is the cause of another” rather than just correlation <br> I recommend reading (<a href="https://alexdeng.github.io/causal/intro.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://alexdeng.github.io/causal/intro.html</a>) <br> 4/ found out what an <br> analemma is (<a href="https://www.astronet.ru/db/msg/1213202?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.astronet.ru/db/msg/1213202</a>) <br> 5/ an opinion is generally starting to circulate already, (<a href="https://blog.jxmo.io/p/there-are-no-new-ideas-in-ai-only?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.jxmo.io/p/there-are-no-new-ideas-in-ai-only</a>) <br> that with the current generation of nets we won’t see a breakthrough, we’ll only see the results of better-labeled data <br> 6/ <br> Feynman (<a href="http://genius.cat-v.org/richard-feynman/writtings/letters/problems?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://genius.cat-v.org/richard-feynman/writtings/letters/problems</a>) <br> “ <br> No problem is too small or too trivial if we can really do something about it.” <br> 7 <br> / Read a couple chapters of Schopenhauer. Realized that philosophy is of course cool, but the practical value isn’t that big. “The World as Will and Representation”.</p> <p><a href="https://t.me/chillhousetech/1128?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> 26 years at Apple. https://darkolorin.com/posts/1127.html https://darkolorin.com/posts/1127.html Wed, 27 Aug 2025 16:30:50 GMT <p>26 years at Apple. Chatted with a top guy who came in with Jobs’ return. He immediately fired 99% of the design team to set a new culture. BTW a pretty common tactic in the Valley.</p> <p>A few insights:<br> 1/<br> Culture.<br> Apple was set against people with “previous experience.” Any line like “at [previous company] we did it this way” is basically a one-way ticket.</p> <p>2/<br> Hiring.<br> They only take people who see hard problems as a challenge, not as an excuse to bail. Curiosity, team play, and product obsession matter more than titles and a resume.</p> <p>3/<br> Process.<br> Before a product launch: 4 months of prototyping. “No” was the main word. Better to say “no” early than screw up the release.</p> <p>4/<br> Categories.<br> Apple was never first. They waited until the tech matured and there was a chance to make a “holy shit” moment. The iPhone only became possible because 5 years earlier they bought a multitouch shop.</p> <p>5/<br> AI.<br> He admitted they… they lost. Catching up. Probably via acquisitions.</p> <p>6/<br> Leadership.<br> He splits all bosses into 2 categories: the ones who do, and the ones who look for excuses. At Apple you couldn’t bring a problem without a proposed solution.</p> <p>7/<br> Remote.<br> “Want a job — work from home, want a career — come to the office.” His team sat on campus even during covid.</p> <p>Generally speaking, inside Apple at the top it’s a very result-oriented company. And brutal release cycles affect what you can actually get done.</p> <p><a href="https://t.me/chillhousetech/1127?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Looks like they saw the KPI. https://darkolorin.com/posts/1126.html https://darkolorin.com/posts/1126.html Wed, 27 Aug 2025 01:09:26 GMT <img src="https://cdn4.telesco.pe/file/bSb4zLuSll5DsIgYTYVsuEt0ih-bwWOUfBMtNx83V-Wf948SFDLVjwrPK9_HorCSoWFF4TUUnWJE4vcinFKEvotsiVKc_S8CTfXD1Qo_-JmZyMELrYI0QxyH0F3BRArNotq8vXT8inE8hnY78yYlRh7zY8U0Uupdt2dpAYYWTUo_1KE_AmMO9Pv1hTVOixAI0CWUpq-0mYz4pdyGi4cl_v2Fw6noZX5TvXXHScJT8dMeWMpESp7tpeo0BxBEEOkRnJ83wQo9wJEw1h8a228mA2wfayT04Lt-BdMmtneONv5M6wwPUZGUJcXY49ilIOnpjExO5pxypEkDJ957B3MS2g.jpg" alt="" loading="lazy"> <p>Looks like they saw the KPI.</p> <p><a href="https://t.me/chillhousetech/1126?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sunday stuff. https://darkolorin.com/posts/1125.html https://darkolorin.com/posts/1125.html <p>Sunday stuff. I figure it’s easier than dealing with Jira.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1125?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Before any startup/manager/enthusiast goes off building their own agent to… https://darkolorin.com/posts/1124.html https://darkolorin.com/posts/1124.html Thu, 21 Aug 2025 21:03:31 GMT <p>Before any startup/manager/enthusiast goes off building their own agent to “take over the world”, I’d recommend vibe-coding something from scratch to the finish that you’re not ashamed to show someone other than yourself — at least a couple, three features.</p> <p>P.S. sobers you up real fast and gives you a sense of what an LLM actually is</p> <p><a href="https://t.me/chillhousetech/1124?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A bit of prose. https://darkolorin.com/posts/1123.html https://darkolorin.com/posts/1123.html Wed, 20 Aug 2025 10:50:58 GMT <p>A bit of prose. <br> Roughly speaking, the more you:<br> - think about how to grow your salary<br> - what you could do so they notice you<br> - how to behave “properly,” what to say, how to write emails <br> The worse things are in your company and team. Time to move on. <br> You have no idea how many all kinds of mid and top managers are obsessed not with what product to ship and how to grow, but with how to develop “culture,” preserve “reputation,” increase “resilience.” Say thanks to them, they give startups a chance.</p> <p><a href="https://t.me/chillhousetech/1123?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Back in business. https://darkolorin.com/posts/1122.html https://darkolorin.com/posts/1122.html Tue, 19 Aug 2025 07:43:46 GMT <p>Back in business. Ran into a job posting at SuperIntelligence Labs. The very one where Mark was poaching and offering researchers $100M for 4 years and all that. So yeah, they’ve got openings now — of course, nice “solid” $200–300k a year, like whatever, working next to guys who’ve got 50–100x more :) <br> Source (<a href="https://www.metacareers.com/jobs/717327034414949?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.metacareers.com/jobs/717327034414949</a>)</p> <p><a href="https://t.me/chillhousetech/1122?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> You can still sleep easy! https://darkolorin.com/posts/1121.html https://darkolorin.com/posts/1121.html Fri, 08 Aug 2025 12:25:33 GMT <img src="https://cdn4.telesco.pe/file/uVk1fKVt1EbJRxNRH2UgIZj6jK2P0_fteaHkOzq4haqZn5YR3uYcD0VY2tLvBeiHu-PODnUfFthCL2CBIoHebKf9RcxVhGOmx4FPTtNBT7kjt_-wJrbAu002DVx4HST30hEuBVc7RH-EdJeceHcf0Ufn60vXN_rql_DIOLjy5_k6CRpmakYc-2pKQEi8Je_-1jOt8GSuNnN1Utxe0C0s87a4ZhpontXC4RsEj_-84iOzg7oO6MdF9TXxvml6LZiQfx2YOJsFVHdjFixu9zvuznXfnocGGR5EGAXg8I65LvfNQ7s_F-EkNxTl9SFx6vTBFfmHa9woqSzIXGMgHm-H-Q.jpg" alt="" loading="lazy"> <p>You can still sleep easy!</p> <p><a href="https://t.me/chillhousetech/1121?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Everything you need to know about GPT-5’s release is in the picture. https://darkolorin.com/posts/1120.html https://darkolorin.com/posts/1120.html Thu, 07 Aug 2025 18:05:51 GMT <img src="https://cdn4.telesco.pe/file/oSCT8_5owqYlp1Vtct1n7S8yPUk8Er6PvE8iKhIjQrGhJpshGOs2qVTu1gv0NXKdwXZNnIKZtyzNd3S2_55KJNJK8plNekxUzHBIVahKPFH4kjs2cmJ7FjK_53ySAbS0Ao0sU_UqEFKGm66jEdzOZJKSZbuVgOOcW67NIaJdP2u3XLHD4oOHvkT-NWs9ecoe9hhP1GGX_7ZdxlxnrkWWJiCpYf36TT72yS1WbZs61A7A3z2Ie8N60G3d8wvMhQqNlRj6yzgjSOW4cilMiQ0QYREekAoYa3c9UVH27lVXH0gpelwMmEjzUbrFJ6picH4aRpX4FtaYqfb8Pp7fF4im-A.jpg" alt="" loading="lazy"> <p>Everything you need to know about GPT-5’s release is in the picture. <br> And yeah, looks like they scrubbed the datasets, tweaked the safety rules, a little bit of everything.</p> <p><a href="https://t.me/chillhousetech/1120?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> OpenAI released open-source models. https://darkolorin.com/posts/1119.html https://darkolorin.com/posts/1119.html Tue, 05 Aug 2025 17:12:10 GMT <p>OpenAI released <br> open-source models. (<a href="https://openai.com/open-models/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/open-models/</a>) <br> We’ll support them on our side soon too :) <br> The models aren’t multimodal and overall are pretty big. By the metrics they’re good and close to o4-mini. Whatever that means.</p> <p><a href="https://t.me/chillhousetech/1119?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There was a post somewhere that was musing about startups and like, how hard… https://darkolorin.com/posts/1118.html https://darkolorin.com/posts/1118.html Mon, 28 Jul 2025 00:02:40 GMT <p>There was a post somewhere that was musing about startups and like, how hard it is and you have to be ready for the fact that when it doesn’t take off the first time you’ll constantly be doing pivots (well, when you change the essence of the product or the target audience). But not many people mention that these very pivots are the reason everything falls apart. The reasons, by the way, aren’t that obvious:<br> 1/ most people start their project because they believe in the idea. Yep yep, exactly that pure kind, when you “figured out” what to do. And off you go. But stepping away from that very dream that’s dear to you hurts like hell. And the people aren’t the same anymore, and their eyes don’t light up. Pretty much all successful companies went through changing the idea (well, with rare exceptions, or the focus shift was step by step). And I think those founders got seriously lucky. I think this is the main reason.<br> 2/ you need different people when you changed focus. Turns out you’re not a little chat with a psychologist anymore, but software for clinics. And that’s sales, long negotiations and integrations, and years before money. Different skills, different people, and different things drive you, and now you don’t even get where to find them, who to ask. And breaking up is always super unpleasant. This is exactly where all those “startup magazines” write about how hard firing is. Usually it’s hard to fire the people you started with. And the bigger the scale, the weaker this effect is, because people’s expectations are lower too.<br> 3/ communication with the team. It’s hard to explain why you shouldn’t try to ride it out to the end on the thing that’s about to “start working” any minute now. And that’s why at the end of a startup they often go looking for someone to blame. Advice for those who work in startups. Say right away if you don’t like the new direction or product, don’t endure it under any circumstances. There won’t be any cash in the startup for a long time anyway :)<br> 4/ pivots need to be done hard, fast, and with zero doubt. Dragging it out multiplies all the effects above by a ton.</p> <p><a href="https://t.me/chillhousetech/1118?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Technical cofounder 👍 https://darkolorin.com/posts/1117.html https://darkolorin.com/posts/1117.html Fri, 25 Jul 2025 21:49:09 GMT <img src="https://cdn4.telesco.pe/file/mejhgXirztBJ7E0JR2WuAXYG7YXmPoFY4rLRK-_JDLUtgfErDqcW8ZJB2NrNPUxAyghRMfmj9ZoLP0ZkS3E7ICn2tejSyrD8l2pG092jTdgV-o1AQSxlfXtpn4LqELhsK60b0M1OR5vWIP_BbqpWmtfpntwt6gcWhHU6Y6zFc1UEHI5iN-ZOiFyXWa2vhEQt_Giozu0tTPVCac_A8WkZbuZP-r00IlsjKym7g5QsdZI3BOQyKi0m8hnQ8kyLIJbA9j0IW21yrxi4iX0nZL2ikvgyS-p6IXf86hCr-8scXNlOhopPpyV2A1IfxJsIES5JrIMprN2fRd9dNj3llFu2iA.jpg" alt="" loading="lazy"> <p>Technical cofounder <br> 👍</p> <p><a href="https://t.me/chillhousetech/1117?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I love Sam and his predictions. https://darkolorin.com/posts/1116.html https://darkolorin.com/posts/1116.html Fri, 25 Jul 2025 17:05:04 GMT <p>I love Sam and his predictions. So to speak, the juice from the latest podcast. (<a href="https://youtu.be/aYn8VKW6vXA?si=Z0uKm_dsEoLUkIQB&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/aYn8VKW6vXA?si=Z0uKm_dsEoLUkIQB</a>) <br> • Traditional universities will become useless for kids who were born recently <br> • GPT-5 already beats Sam at specialized tasks and will be able to handle the full CEO workload “in the near future,” meaning automated leadership of big companies is basically about to happen <br> • That same gap in capabilities (between AI and a human) will spread to a bunch of roles, since future models will reason, write code, negotiate, and plan faster than your regular human employees, which means a significant chunk of modern office work will move to AI agents <br> • Gotta get used to AI being smarter than all of us put together, that’s it <br> • Of course the ones who own AI will be infinitely rich <br> • GPT-5 instantly solved a hard problem, making Altman feel useless, and that’s the shock a lot of workers are gonna experience <br> • Fusion is everything, because without it there won’t be enough energy for systems like this <br> Do we believe it?</p> <p>P. S. I’m still telling the net to follow my formatting rules and it can’t :) well whatever, it’s fine ok</p> <p><a href="https://t.me/chillhousetech/1116?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Everything you need to know about companies’ “moral values.” Anthropic is… https://darkolorin.com/posts/1115.html https://darkolorin.com/posts/1115.html Tue, 22 Jul 2025 11:20:05 GMT <p>Everything you need to know about companies’ “moral values.” Anthropic is gonna rake in cash in the Arab world.</p> <p>Amodei acknowledged (<a href="https://www.wired.com/story/anthropic-dario-amodei-gulf-state-leaked-memo/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.wired.com/story/anthropic-dario-amodei-gulf-state-leaked-memo/</a>)<br> in his note that taking money from Middle East leaders would likely enrich “dictators.” “This is a real downside and I&#x27;m not thrilled about it,” he wrote. “Unfortunately, I think ‘No bad person should ever benefit from our success’ is a pretty difficult principle to run a business on.”</p> <p><a href="https://t.me/chillhousetech/1115?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, here come the AGI signs. https://darkolorin.com/posts/1114.html https://darkolorin.com/posts/1114.html Sat, 19 Jul 2025 11:32:49 GMT <p>Well, here come the AGI signs. As Noam from OpenAI writes, they themselves didn’t know a couple months ago that this would come out. <br> In short, a new-type model that can reason for hours managed to take a gold medal at the International Math Olympiad without using the web, Python, or other tools. That’s already a pretty solid level (just a year ago they were measuring dicks using regular school math as a benchmark). What’s inside isn’t clear. Looks like we’ll wait another quarter and try it ourselves. “Thinks for hours” means the queries will most likely be paid, and heavily paid. But I think they’ll make the approach cheaper. Interesting times we live in! <br> Thread (<a href="https://x.com/polynoamial/status/1946478249187377206?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/polynoamial/status/1946478249187377206?s=46</a>)</p> <p>Problem solutions (<a href="https://github.com/aw31/openai-imo-2025-proofs?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/aw31/openai-imo-2025-proofs</a>)</p> <p><a href="https://t.me/chillhousetech/1114?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sitting here thinking what kind of task I should yeet into the new OpenAI… https://darkolorin.com/posts/1113.html https://darkolorin.com/posts/1113.html Fri, 18 Jul 2025 14:57:22 GMT <p>Sitting here thinking what kind of task I should yeet into the new OpenAI agent (<a href="https://openai.com/index/introducing-chatgpt-agent/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/introducing-chatgpt-agent/</a>)<br> If you’ve got tasks, drop the prompt straight in the comments pls — I’ll try to run it and post a link to the result or a little video :) I’ll pick something interesting of course</p> <p><a href="https://t.me/chillhousetech/1113?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A quick “they asked, I can’t say no” moment. https://darkolorin.com/posts/1111.html https://darkolorin.com/posts/1111.html <p>A quick “they asked, I can’t say no” moment. <br> My former co-founder on a game project is building a new company about automating work with content and influencers — drop by and read if you’re interested <br> @dimabeseda (<a href="https://t.me/dimabeseda?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/dimabeseda</a>) <br> (attached a video to the post), content factories have gotten painfully popular.</p> <p>P.S. they recorded a podcast with me, (<a href="https://youtu.be/EJIzSdHw88g?si=-JQKelAbtBGewzo0&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/EJIzSdHw88g?si=-JQKelAbtBGewzo0</a>) <br> where I probably, yet again, talk rough, ramble uselessly, and all that :) funny thing is my opinion on some stuff has already changed — AI is just too dynamic a field, it’s hard to stick to your guns for long.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1111?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> We’re little by little making demos of what you can do with local models. https://darkolorin.com/posts/1110.html https://darkolorin.com/posts/1110.html <p>We’re little by little making demos of what you can do with local models. Here’s a small example. And we’ll post a bunch more stuff too!</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1110?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Time for the release. https://darkolorin.com/posts/1109.html https://darkolorin.com/posts/1109.html Tue, 15 Jul 2025 14:20:47 GMT <p>Time for the release. Channel readers get it first. <br> What I’m doing: <br> We’re doing on-device AI. Started by building our own inference engine from scratch for models on Apple Silicon. And it’s open sourced by 90%. There’s a small chunk tied to the platform — speculative decoding. Which can give a serious, up to 10x, inference speedup. <br> About the company: <br> It’s called Mirai — “future” in Japanese. We believe local models are a very important part of the future UX and what the interaction experience between a person and the devices around them will look like. Raised a pre-seed of about $4M USD, <br> there’s a bit more on the website. (<a href="https://trymirai.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/</a>)</p> <p>Links and how you can help: <br> 1/ <br> the project is open source (<a href="https://github.com/trymirai/uzu?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/trymirai/uzu</a>) <br> and we’ll be very happy if you star it and show it to people who work with iOS &amp; macOS apps. <br> There are Swift bindings. (<a href="https://github.com/trymirai/uzu-swift?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/trymirai/uzu-swift</a>) <br> There’s <br> our own approach (<a href="https://github.com/trymirai/lalamo?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/trymirai/lalamo</a>) <br> to how models should be represented for devices. MIT license. Fork it however you want! <br> 2/ <br> you can hop onto our site (<a href="https://trymirai.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/</a>) <br> and try everything yourself, it’s a dev project, but if you’re a business and want to build local stuff and save on cloud inference — welcome to DM <br> 3/ upvote on <br> HackerNews (<a href="https://news.ycombinator.com/item?id=44570048&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://news.ycombinator.com/item?id=44570048</a>)</p> <p>4/ we’re faster than llama cpp and the popular wrappers around it, we’re more convenient and, thanks to speculative decoding, faster than Apple’s MLX. We’re looking for RnD &amp; core performance engineers to support the audio pipeline (we’ll be working with a very large provider of audio models). <br> Thanks for reading!</p> <p><a href="https://t.me/chillhousetech/1109?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Downloaded this app with that hyped new model from the Chinese at 1T params… https://darkolorin.com/posts/1108.html https://darkolorin.com/posts/1108.html Sun, 13 Jul 2025 14:36:32 GMT <img src="https://cdn4.telesco.pe/file/s2dP2EPb6Eg_b9SFeg8neIUVOgJw5eMkMyLOWZupGEX98Oo1ew9JYxzUm0DTuXIjZgyAVaToZIwiSA9g-lDQtpV28jNducl232GueIweTKqhxVdj2dCrIkj4B0i9FsHIb_I2tkfoZznPNdFRWgifZQUZ70xcYjNG7wvMOMmadkU1gwjzDsDuEbwYOehdaSoBv7WR-YwAbrU8JhH4N__SbG289PYwXd_DSy0xELJnxH7_bDyGvdyuAg24Cy20cHk9sKHbPqYQM_cx03svCvO4hQ3z3gdcQsWfECKU7A0cAWtpi1NTbAUkPILCbERT_44B5IDsxsHJ1UzW5EXMz7NN8Q.jpg" alt="" loading="lazy"> <p>Downloaded <br> this app (<a href="https://apps.apple.com/app/id6474233312?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/app/id6474233312</a>) <br> with that hyped new model from the Chinese at 1T params (though it’s MOE so active is like 32b) — KIMI. <br> It’s made (<a href="https://moonshotai.github.io/Kimi-K2/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://moonshotai.github.io/Kimi-K2/</a>) <br> specifically for agent usage. Try it. <br> The suggest thing did kinda throw me off though. Oh well :)</p> <p><a href="https://t.me/chillhousetech/1108?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In short. https://darkolorin.com/posts/1107.html https://darkolorin.com/posts/1107.html Fri, 11 Jul 2025 22:16:24 GMT <p>In short. The scheme is like this. We fund a startup — sell smart dudes to OpenAI. What could possibly go wrong. <br> 1/ <br> OpenAI buys Windsurf for 3 billion (<a href="https://www.reuters.com/business/openai-agrees-buy-windsurf-about-3-billion-bloomberg-news-reports-2025-05-06/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reuters.com/business/openai-agrees-buy-windsurf-about-3-billion-bloomberg-news-reports-2025-05-06/</a>) <br> (that’s when there are tons of prompts flying around and all that). We’re happy! <br> 2/ But! <br> Three months later OpenAI decides they don’t need it anymore. (<a href="https://fortune.com/2025/07/11/the-exclusivity-on-openais-3-billion-acquisition-for-coding-startup-windsfurf-has-expired/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://fortune.com/2025/07/11/the-exclusivity-on-openais-3-billion-acquisition-for-coding-startup-windsfurf-has-expired/</a>) <br> 3/ Investors write the money off, boys <br> go to Google. (<a href="https://www.theverge.com/openai/705999/google-windsurf-ceo-openai?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/openai/705999/google-windsurf-ceo-openai</a>)</p> <p>Ez. <br> 👍 <br> P.S. It’s kinda sweet how the fund got hooked up with API credits, since all these coding assistants are an omega-unprofitable business. Of course everyone will write that the investors didn’t get screwed. But usually the terms of deals like this are that the fund gets paid if the team hits KPI, and nobody ever hits them. Such is life.</p> <p><a href="https://t.me/chillhousetech/1107?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Here came out a funny study. https://darkolorin.com/posts/1106.html https://darkolorin.com/posts/1106.html Fri, 11 Jul 2025 13:21:14 GMT <img src="https://cdn4.telesco.pe/file/fmBdciPf58XxfWFZY5e2XGIIvxOa_nXdIuQHkkTh-2q1qg70YMz1AVIa7GqjwPVYevBGE1Ggvyj1r47hayTCBuJjnsJYT6zHbp3osmE7MUC8zzPyMbfzSRqUTWcS40EPmwDqOcwnpuZLKNlljIQqWLYloRrqpTNVLWOyyiX7Z4rmndVeW9AsF93s98QvZu_elfX2J_vQpUlJn5W6jh9lVJe1Dv7MAK4Q8F5caEx6ORkXuEojLoD9BnWLXpDgzH8nxTDiModxZ3ByL6G2gL7LhqDW39Q1Y6F_GkoSfHY_3ZMfLjaIBm0Y7totK7aOWESTyMIcQaePCACUH3L8Y64Eiw.jpg" alt="" loading="lazy"> <p>Here came out <br> a funny study. (<a href="https://metr.org/Early_2025_AI_Experienced_OS_Devs_Study.pdf" rel="noopener">https://metr.org/Early_2025_AI_Experienced_OS_Devs_Study.pdf</a>) <br> Like, does AI actually speed up the dev process. Turned out everyone thinks about a 20% speedup, but it ended up slowing things down by 19%. Reasons, as I see it, are plenty, starting with novelty, painful maintenance, the model constantly changing, a pile of new tools, and so on. But the fact itself is worth a look to read up on.</p> <p>A decent breakdown (<a href="https://t.me/boris_again/3307?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/boris_again/3307</a>) <br> by people who know their stuff</p> <p><a href="https://t.me/chillhousetech/1106?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “It’s a good model, sir” (c) P. https://darkolorin.com/posts/1105.html https://darkolorin.com/posts/1105.html <p>“It’s a good model, sir” (c) <br> P. S. The longer a modern benchmark is publicly available, the higher the chance someone’s replicated its logic and dumped a ton of synthetic crap into the dataset. Or maybe not :)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1105?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Kek https://darkolorin.com/posts/1104.html https://darkolorin.com/posts/1104.html Wed, 09 Jul 2025 16:38:16 GMT <img src="https://cdn4.telesco.pe/file/st-XfageoUg9Rm2inuDyhOPeiogB1693NG7DzQ-GjTTM5wjHcI0aubuqHZGoT42fQslupYNDPfXo9BnLSopR55CfFHWVQ3ZtNbhGhDPuxCLJZi0Imu64YMNdbLX20iDIXAJoUvKDPvu5IhudlWEICArL02rLQTY4Nx6hzS_p5gXsMBN2Jb-5uLf6-bT4q10SNPSHIdH8_ahg2vmMMRnBHKiXNVYHfOI9V21CKy6zWGywuABfKV4m-qXxDlukW_i0DBy9z_LLSnNLxyqk7GHjK0mo302R2cM_RKQWuc5ap_Wdo4w9aH-CxWGXiD7DxqqT7awd1REird0teQRvvkP0Nw.jpg" alt="" loading="lazy"> <p>Kek</p> <p><a href="https://t.me/chillhousetech/1104?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Alright, I got it — how to redo the homepage for “high” conversion. https://darkolorin.com/posts/1103.html https://darkolorin.com/posts/1103.html <p>Alright, I got it — how to redo the homepage for “high” conversion.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1103?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Finally a pre-release working state of the platform. https://darkolorin.com/posts/1102.html https://darkolorin.com/posts/1102.html Mon, 07 Jul 2025 12:51:47 GMT <p>Finally a pre-release working state of the platform. We’re looking for iOS/macOS devs to test it and who can give feedback. I’ll drop you the link to try it in DMs. Write in the comments for access, who to send it to.</p> <p>P.S. or if you’ve got a team — you can pass it along</p> <p><a href="https://t.me/chillhousetech/1102?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> FAANG ——> GAYMAN https://darkolorin.com/posts/1101.html https://darkolorin.com/posts/1101.html Tue, 01 Jul 2025 16:29:42 GMT <img src="https://cdn4.telesco.pe/file/dlSu_nOd4X5X0WTpDs6a4KFsK3FFrAWaq_HiL3edwajrLin2JVsrCki0QIkTB9tNITXoM6XMLROPfcaz3UlH5SkhFNxL4fdX3k8aQAn8RMRtT7RWxm1IYlv3Qdi1xthTZhs51b-V43jrXfBAqN-fZC1kpKpVwmzLkHj26mY4urLKiCOm6TFuzfQb3t-WFP1ftPwHWdzDj65cv-IQ_4GqQpY__48eFgAU0Uk1Kq0-57VZDpu5iKc6L6e1eYoeqDW2etCcPIvAVqLTgLjeSHwbHb5Rb_VYhXvvR4Ys8crmWQj9TNUlcgfHXXy79xcLx_U5LYdsRwM7_Yj4WISZQnoXbA.jpg" alt="" loading="lazy"> <p>FAANG ——&gt; GAYMAN</p> <p><a href="https://t.me/chillhousetech/1101?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ve been wanting to write a small post for a while (while we’re, damn it… https://darkolorin.com/posts/1100.html https://darkolorin.com/posts/1100.html Mon, 30 Jun 2025 16:56:29 GMT <p>I’ve been wanting to write a small post for a while (while we’re, damn it, actively working on the release, just one feature left to finish off). About “we hit 10M ARR within 1 month”. You can plug any numbers you want into that sentence. I’m mostly talking about consumer/prosumer and all that b2c stuff.</p> <p>For many many years everyone’s been saying it’s important to find CAC/LTV and then just pour money in. But what usually happens is a slow death, especially if the project is venture-backed.</p> <p>A short action plan today:<br> 1/ take AI, shove a model in (usually a ready-made one, God willing at least tuned, or with an if-pipeline on top) or slap a chat on the side of any product (literally any, chat is such a great UX)<br> 2/ hire a team of folks with a couple-three years of failed or successful experiments<br> 3/ go raise a round, or apply to YC 3–4 times (often they take you for persistence and people, because people matter more, and jamming your idea into the current trend is pretty easy)<br> 4/ build the product, ideally catch a mini-viral wave and then at the end of the viral cycle start mercilessly flooding all paid marketing (money spent/revenue is how teams think, they just divide spend by revenue and get this omega blended CAC metric), which also includes a TikTok farm and so on (they cite 1–2 outliers/exceptions from the market where in reality they just got lucky)</p> <p>Then the exact same thing happens as in the late 2000s and mid 10s and so on, but for some reason “this is different”. Namely:<br> 1/ so yeah, launch happened, it took off to a thousand users and they even paid, but the problem is there’s no fast, equally pretty organic<br> 2/ ahaaaa, we’ve got money, let’s buy (doesn’t matter if it’s TikTok, Facebook, Google etc)<br> 3/ WOW! look it matches up, whoa whoa, scaling mercilessly — this is that very moment <br> X M USD ARR WITHIN X WEEKS<br> 4/ keep burning money and go raise more, and preferably faster because — that very moment<br> 5/ then we scale and suddenly boom, top-line ceiling, how come our payback period isn’t 9 months but 12, oops it’s 18<br> 6/ then optimizations start (maybe they were there from the start), games with conversions, landing pages, etc, but that’s not the problem<br> 7/ subscription options, whatever, anything — as long as we don’t use our brains</p> <p>What’s the reason. Well most often it’s banal:<br> 1/ Paid UA gets copied by all competitors in one-two if they see you buying successfully, and if you don’t have product features that make the cost of quitting huge then gg (could be a friends graph, could be work docs or something else, preferably multiplayer or linkage to past data — social capital)<br> 2/ the founder team during this kind of scaling has it really hard — purely mentally it’s hellish vibes and massive pressure — everyone underestimates this reason, and it’s exactly in these moments that monstrous mistakes happen (either dump money into everything or say screw everyone and do pivots)<br> 3/ counting everything as blended CAC is a mistake, you need to very clearly separate all channels and watch market trends and tech, plus constantly estimate the potential capacity of the available audience<br> 4/ underestimating the novelty factor — all of AI is exactly that, you need to carefully study NRR (net dollar retention in your market) and find a way to hold it, because most people just came to try it<br> 5/ scaling effects work against you in paid, because everyone’s already seen the creatives and the people who haven’t are few or they’re less interested in the product (meaning more expensive)</p> <p>A lot of people have noted that all these lessons are here (<a href="https://www.slideshare.net/wahengchang/dropbox-startup-lessons-learned-239520249?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site#1" rel="noopener">https://www.slideshare.net/wahengchang/dropbox-startup-lessons-learned-239520249#1</a>)<br> (need VPN)</p> <p>P.S. none of this means founders won’t make money, they easily will by pumping valuation to a billion and selling their shares (well, part of them), there’s just no business in all this, and no company over the long run either</p> <p><a href="https://t.me/chillhousetech/1100?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> It seems to me people consider “hard” the stuff that rarely works out (which… https://darkolorin.com/posts/1099.html https://darkolorin.com/posts/1099.html Sat, 28 Jun 2025 16:53:41 GMT <p>It seems to me people consider “hard” the stuff that rarely works out (which is basically the same as high competition). Like, for example, building a startup, doing nothing and making a ton of cash, and so on. Even though personally I don’t see anything hard about building a startup, it’s work, you choose it, and it has its own нюансы.</p> <p>And when someone pulled it off without suffering and whining — then it’s luck, a freebie, friends helped, etc.</p> <p>P.S. I don’t know anyone who both made it and, in the process, was constantly whining, “finding themselves” and writing posts</p> <p><a href="https://t.me/chillhousetech/1099?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> To all AGI AGI fans — 4 minutes dedicated unimaginable video https://darkolorin.com/posts/1098.html https://darkolorin.com/posts/1098.html Sat, 28 Jun 2025 02:16:25 GMT <p>To all AGI AGI fans — 4 minutes dedicated <br> unimaginable video (<a href="https://youtu.be/t3TfmU0l5vM?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/t3TfmU0l5vM</a>)</p> <p><a href="https://t.me/chillhousetech/1098?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Saturday stuff. https://darkolorin.com/posts/1097.html https://darkolorin.com/posts/1097.html Sat, 21 Jun 2025 14:41:41 GMT <img src="https://cdn4.telesco.pe/file/PKVhNQnrf9sNzmIS-IgUvC8Kf5Ks9gBpgp4aarFZoZ1uzVbQyFuph_zLTvIEt-3IfQBOZc_BQqovsSepuEyYq-L4jWy0hv05VoLCBbTkVJg5rM2qaUr0Tz9GgilYpPu_onV9pIC9pyRYN5QPViGuZTwCPGtPqijWBkBLTimYvftrmcgoGI2a1YUaU7vMX88Pi4aroXJJsfWtYQ-wfd2twUvi390Dwz7uKzX-0XXQW2fgNnVVP7DqwcXqdIuBA8WXC6RaL3PgZ1u6yjjjciujXfqy9dZTtw9yey4uJurHq6KHR1iJD6J8Fcld5a12gsNGOge5u1nfxKa1VdSR-5TnvA.jpg" alt="" loading="lazy"> <p>Saturday stuff. A creative batch at YC. I remember there was this with Uber for.</p> <p><a href="https://t.me/chillhousetech/1097?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Andrey Karpathy spoke in the neighborhood recently, and his take on the… https://darkolorin.com/posts/1096.html https://darkolorin.com/posts/1096.html Thu, 19 Jun 2025 16:19:49 GMT <p>Andrey Karpathy spoke in the neighborhood recently, and his take on the current AI/tech industry is the closest to mine. <br> I’d recommend spending 40 minutes of your time (<a href="https://youtu.be/LCEmiRjPEtQ?si=isruM2NObyq_Bf4Q&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/LCEmiRjPEtQ?si=isruM2NObyq_Bf4Q</a>) <br> and not reading any summaries, because his whole talk is the summary of what’s happening right now and where it’s headed. And yeah, my favorite analogy: agents vs self-driving cars.</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1096?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Saturday vibes. https://darkolorin.com/posts/1095.html https://darkolorin.com/posts/1095.html Sat, 14 Jun 2025 21:28:22 GMT <p>Saturday vibes. Sometime next week or maybe the one after — time will tell — I’ll show what I’ve been working on lately. The company, the product, the round and all that. Lots of open source for each and every one. And of course it’s about AI. I’ll be happy for your reposts and support. <br> Spoiler: we’re not making our own models — for now it’s kinda too expensive lol.</p> <p><a href="https://t.me/chillhousetech/1095?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Dear ladies and gentlemen, what’s this trend of “optimization for LLM chatbots”? https://darkolorin.com/posts/1094.html https://darkolorin.com/posts/1094.html Fri, 13 Jun 2025 17:56:57 GMT <p>Dear ladies and gentlemen, what’s this trend of “optimization for LLM chatbots”? Can someone explain how it’s supposed to be “so radically different” from normal search ranking? They build their index by the same principles, or they straight-up hit the Bing API. Every tip I’ve seen is literally the same as in search.</p> <p>Dude even wrote a guide. (<a href="https://x.com/jakezward/status/1933146651620679927?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/jakezward/status/1933146651620679927?s=46</a>)<br> Anyone who actually gets it — maybe I’m missing something specific? Share what you’ve found in the comments.</p> <p><a href="https://t.me/chillhousetech/1094?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I find it pretty telling that the CEOs of the labs have completely started… https://darkolorin.com/posts/1093.html https://darkolorin.com/posts/1093.html Fri, 13 Jun 2025 13:54:55 GMT <p>I find it pretty telling that the CEOs of the labs have completely started lying their asses off and went full doomer. And I think this is going to be the main source of the company’s problems long-term. They sell what they do as something without which humanity will just die. Convenient pitch.</p> <p>Well, Jensen Huang, CEO of NVIDIA, basically agrees with that too. (<a href="https://fortune.com/2025/06/11/nvidia-jensen-huang-disagress-anthropic-ceo-dario-amodei-ai-jobs/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://fortune.com/2025/06/11/nvidia-jensen-huang-disagress-anthropic-ceo-dario-amodei-ai-jobs/</a>)</p> <p>Quote (not translating): “One, he believes that AI is so scary that only they should do it,” Huang said of Amodei at a press briefing at Viva Technology in Paris. “Two, [he believes] that AI is so expensive, nobody else should do it … And three, AI is so incredibly powerful that everyone will lose their jobs, which explains why they should be the only company building it.&quot;</p> <p>P.S. ads are of course such bullshit; if I do them, it’ll only be on my terms from here on out.</p> <p><a href="https://t.me/chillhousetech/1093?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The worst way to pay employees across different countries is doing everything… https://darkolorin.com/posts/1092.html https://darkolorin.com/posts/1092.html Thu, 12 Jun 2025 09:04:09 GMT <p>The worst way to pay employees across different countries is doing everything manually: contracts, invoices, bank paperwork.<br> Imagine how many resources the finance team burns paying salaries to different countries while complying with international laws. And then banks go and freeze payments, ask for documents, and demand you prove the contractor’s qualifications.<br> In the end, for most global companies, paying remote employees and freelancers is a headache and a pile of risks. But with the platform <br> 4dev.com (<a href="https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech" rel="noopener">https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech</a>) <br> everything is way easier.<br> ⭐️</p> <p>4dev.com (<a href="https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech" rel="noopener">https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech</a>) <br> is:<br> - one B2B contract with the platform instead of dozens of individual ones with each employee <br> - instant payouts to remote employees and freelancers in 100+ countries, including the CIS <br> - automatic generation of closing documents for every payout and a fix for bank compliance hell <br> - a 1–3% fee for business <br> 💵 <br> legal crypto payments, plus payouts in 30+ fiat currencies<br> All of that with no fire drills and no soul-sucking paperwork.<br> Book a demo in 1 click (<a href="https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech" rel="noopener">https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech</a>). (<a href="https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech" rel="noopener">https://4dev.com/ru?utm_source=influencer&amp;utm_medium=telegram&amp;utm_campaign=tgchannels_2025&amp;utm_content=chillhousetech</a>) A platform specialist will dive into your tasks and help find a solution specifically for your business!<br> #integration</p> <p><a href="https://t.me/chillhousetech/1092?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Really waiting on honest feedback from investors btw https://darkolorin.com/posts/1091.html https://darkolorin.com/posts/1091.html Wed, 11 Jun 2025 23:54:38 GMT <img src="https://cdn4.telesco.pe/file/Dt018BN5PMNqqZwKgbfbl0rzO9fTmcaG-lJi8zud_36SiXDOA72mtyvzhWclj8JiQhxty4yIiXCCH6JI-keWSGaCR_HVlVBm1ObCEAbRi5FBGR6y3n-OqlXIgdsRmJwpAYhCWqaj5H7xOHSxke2b8vb0fucpXSjQtNJDeSz0AbCrWwYiVvxi9_ZLXquUD-6MPib4-LiWi-ehWfxcOpYez-RSjxVf2bp8zvM3k_1QXmBp6pEV-LEyb3aZYOL0HaXVkv0VBmAO9Z-wOJWPxlcbVNEMHmcnBddKeRDbC0l7UYxd8tgZSVJ1I9WMe_SZX8lGjXjCT-4xURIc-CO4vP5_Dw.jpg" alt="" loading="lazy"> <p>Really waiting on honest feedback from investors btw</p> <p><a href="https://t.me/chillhousetech/1091?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> o3-pro finally! https://darkolorin.com/posts/1090.html https://darkolorin.com/posts/1090.html Tue, 10 Jun 2025 20:29:12 GMT <img src="https://cdn4.telesco.pe/file/kJCEhNN6Lhf35WdERoYp93rLiU7H56Oa5Nx6d5Mhj_1KhKnB9C3AmXmyQdjaMtgsKPoxsCnn8la5XEufDkpPwLvgXyIV4qqNhJM1dmYQcPLFHVsUqmnHn5S4Vt1LA42WTiT5_QNB6pCiL5xoBMQtimGP3kFPg98OksAKz2ZLHe6zeQIxpRO0cyDKV4AvpMQUabMZuf7pCkR-ZvziyJGk2FnSLQ_S8bZ8wVdBDRqU4HpxT68m0bVorK0EiA3DMZ6KFS65DGScZ8s5eco3He0wUPwUpVTVS54lc18FyS8MTciny_htRBMKVouzDFB7ZR6EjWrl1lRVH-MEj_A4mcQFZg.jpg" alt="" loading="lazy"> <p>o3-pro finally!<br> Sam’s essay for the release (<a href="https://blog.samaltman.com/the-gentle-singularity?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.samaltman.com/the-gentle-singularity</a>)</p> <p><a href="https://t.me/chillhousetech/1090?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For a bit over 2 years there wasn’t a single integration on the channel. https://darkolorin.com/posts/1089.html https://darkolorin.com/posts/1089.html Tue, 10 Jun 2025 14:22:21 GMT <img src="https://cdn4.telesco.pe/file/CoXhpbOrq3F_7MGqiDyPW9eZim7-UDTI-A2039DnoRcAw_CZFS80TzOwoqy7rVOHc3YQm8QJMiFtYlHmGJXezGbNm8c8YaFHL4nfw4rpeML0cqlCkldC5D-j84WDr9Te-W8I0E8eXjFqGK6mOD2c7jUaGBcmSJUpZF5qdzrGRWO-3EOvcEB6P3TcrZ7x0iI9BNzDi8HkLiHH3yqSpKT0XxQkGUi4fzPLWjsGNL8P-6lvzOKdVzDSvkznfiLTNJJU4sUF_ldQLXE1lQQEeS2Sp_K_zJipcotVYQHmZfjmuja6rM_Jmtg5fTDdAJuf5fEfEg0y9O1NNzQ072u-pK7A4w.jpg" alt="" loading="lazy"> <p>For a bit over 2 years there wasn’t a single integration on the channel. Decided to try taking one. The post will be out on Thursday :) though I probably won’t want to do it in this format anymore. Need to think how to organize this.</p> <p>In the meantime I recommend<br> a series of articles (<a href="https://synthesis.ai/2025/04/17/ai-safety-i-concepts-and-definitions/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://synthesis.ai/2025/04/17/ai-safety-i-concepts-and-definitions/</a>)<br> by Nikolenko on safety.</p> <p>Also the other day I read a book about the history of Jim Simons’ Medallion fund. A mathematician who made the most profitable fund in history. Overall, a long time ago I wanted to do quant trading. But I couldn’t be bothered. Picture comparing invested funds.</p> <p>P.S.<br> they cut o3 prices (<a href="https://x.com/sama/status/1932434606558462459?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/sama/status/1932434606558462459?s=46</a>)<br> in the API by a whole 80%, competition’s breathing down their neck and looks like new models soon</p> <p><a href="https://t.me/chillhousetech/1089?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> WWDC 2025 ----> iOS 26 (WAT?) They devoted 30 seconds to AI and gave access to… https://darkolorin.com/posts/1088.html https://darkolorin.com/posts/1088.html Mon, 09 Jun 2025 17:21:28 GMT <img src="https://cdn4.telesco.pe/file/gBoBG_vAMER-GEGq1qYsNWjTS1H6lxyQlPal_fpNfkS3Y4NbBFwgJdYN0etzkixpbzT7EG_697BiR2hTKuAzoaaSFi5oJslaWDCsV1XdkzewNh6r__CD4JRJG8whfRJWBUCeV9SIC9HGRoKNEB4jBBy6LMV0ug_UsHX3CPRn0rWjTI-E9ubmwyAkV-k5ZzAvTcMoe-63moHO3ph6F2coY8rb3S7U2jXUqTsvOvuIxyK6mD0viAf0eVZZeIgEtSEWbxXUsoM_5qhS2f3qJxbWUAcLAzk1j3Ux1wLTR_MsxRsp6zU8nBdVdcTQAT_ulLAZVz9GcohuNuhF7fO4gNAqEA.jpg" alt="" loading="lazy"> <p>WWDC 2025 ----&gt; iOS 26 (WAT?) <br> They devoted 30 seconds to AI and gave access to their models (thanks for your old models). Judging by “26”, we’re waiting for all this in ’26. Post is being updated.</p> <p><a href="https://t.me/chillhousetech/1088?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Half of Twitter is discussing a new paper from, you won’t believe it, a team… https://darkolorin.com/posts/1087.html https://darkolorin.com/posts/1087.html Mon, 09 Jun 2025 12:23:16 GMT <p>Half of Twitter is discussing a new paper from, you won’t believe it, a team at Apple. The same folks who pretty clearly missed the current AI boom. <br> But the paper is interesting. (<a href="https://machinelearning.apple.com/research/illusion-of-thinking?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://machinelearning.apple.com/research/illusion-of-thinking</a>) <br> Pretty damn crappy, I’d say. But man, that title (just like Apple Intelligence == AI): <br> The Illusion of Thinking: <br> … <br> Here’s what the researchers found: <br> 1/ for simple tasks, baseline non-reasoning models performed even better <br> 2/ for medium-complexity tasks, reasoning models won <br> 3/ but for hard tasks, reasoning models not only did a bad job, they didn’t even try to use their token limit (they called it a “reasoning collapse”)</p> <p>A simple conclusion suggests itself. The current generation of LLMs and what it’s fashionable to call LRMs (large reasoning models) still can’t get beyond their data (solve using out-of-distribution concepts). In other words, if it wasn’t in the dataset, it won’t be in the output, because the entire output is an attempt at recombinations and complex imitations of internal representations. And all we’re seeing is basically <br> deductive closure, or “deductive closure.” (<a href="https://en.wikipedia.org/wiki/Deductive_closure?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://en.wikipedia.org/wiki/Deductive_closure</a>)</p> <p>And yeah, that might be enough to solve a huge number of humanity’s problems overall. But is it enough to flip science on its head? Open question.</p> <p>So I think we need to actively develop tooling for these nets, because it can provide new info and feed it at least into the input context, which is already great.</p> <p>A good overview of the Apple article. (<a href="https://medium.com/@vishalmisra/the-illusion-of-thinking-why-language-models-cant-improve-themselves-0c71a13811e2?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://medium.com/@vishalmisra/the-illusion-of-thinking-why-language-models-cant-improve-themselves-0c71a13811e2</a>) <br> Or maybe just a good overview of what’s what.</p> <p>BUT the simplest version of why the models folded on these “hard tasks” is that they need a shitload of context to solve them, so all the reasoning just tore unexpectedly. But needs checking.</p> <p>Here’s an explanation (<a href="https://x.com/scaling01/status/1931783050511126954?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/scaling01/status/1931783050511126954?s=46</a>) <br> that there’s just not enough output context. SO classic Apple bullshit bingo, justifying their lag.</p> <p><a href="https://t.me/chillhousetech/1087?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And here come the expectations from the funds rolling in, on the pace at which… https://darkolorin.com/posts/1086.html https://darkolorin.com/posts/1086.html Sun, 08 Jun 2025 19:27:45 GMT <img src="https://cdn4.telesco.pe/file/HMLeJm8U6kLDJmt-G-_g5K6qUTOV2VbX-zQFxDwiaNkBHOATqkuzX7PccBFGLBm4q7kt5x5ellGiwK5OT9Cea5b2fRa_zte0AVXmfZdXcEI7SV-UX1KHH4210Yc300JyKRbdrdaPJtZLeAlbm_CSdxnlZmdwfbXZ0XgKsINN6j00yjA3_K-zsxpS41UzPdvZImP2SisumRBNT36pBWGcwfkyYoSHAgbWfvw38lc4e55LsAZmR79kADphL14jnGzMQYE7lfmzeKzSWOaBrH_KdUfSGK-wXiDn1n4HCcKpgYpdWE4L7_3MBmSb4ojLsBiXfEKJTDt4vksWLkjHd-Wlmg.jpg" alt="" loading="lazy"> <p>And here come the expectations from the funds rolling in, on the pace at which you’re supposed to be pulling in revenue. Well, in a year in B2B you need a couple mil, and in B2C a whole 4. In millions of dollars. <br> Full article from a16z. (<a href="https://a16z.com/revenue-benchmarks-ai-apps/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://a16z.com/revenue-benchmarks-ai-apps/</a>) <br> For the hostess to take note.</p> <p><a href="https://t.me/chillhousetech/1086?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Lately I’ve been trying a lot to get agents to code properly. https://darkolorin.com/posts/1085.html https://darkolorin.com/posts/1085.html Sun, 08 Jun 2025 11:11:12 GMT <p>Lately I’ve been trying a lot to get agents to code properly. But so far the observations are: the model doesn’t give a damn about instructions like half the time, the tool is kinda meh as something actually usable. So anyway, I stumbled on a whole guide (<a href="https://www-cdn.anthropic.com/58284b19e702b49db9302d5b6f135ad8871e7658.pdf" rel="noopener">https://www-cdn.anthropic.com/58284b19e702b49db9302d5b6f135ad8871e7658.pdf</a>) on how Anthropic themselves use Claude Code internally. I’ll also attach the TL;DR as an image file.</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1085?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The drama on Twitter continues https://darkolorin.com/posts/1084.html https://darkolorin.com/posts/1084.html Thu, 05 Jun 2025 20:23:24 GMT <img src="https://cdn4.telesco.pe/file/lgiu0DE7FpkgvTAkD5VxVrZ8OFwCFTomoc6Fb0fVkWQHvM2BiIQ2wWi34qzFPkC2Qp1DcVnH3F3nzhnJpVtlLC-ekbzrZhwhkbj_VWDLw3mpS9B0UqZua_mqEupps83GK0GJsBQSVvxflLzvr28RKOxF0NCaBP1BvVRBYLKhvIKOVgKLcdrHUYIaSsysrdC6bfR3KZxQqlh7M-mBG08eQtzca7zP5eHxNewTm_Ej3Vag8fYOzsK5o-3wEtU4oMf22Cjp7_mb6xOKJ9kykhCwu4hFd2Y0SPaPO4GMyj7dakb-cKV1ObiahPcaqnfwDMv5p11G5m8trakKaHylDzgFNg.jpg" alt="" loading="lazy"> <p>The drama on Twitter continues</p> <p><a href="https://t.me/chillhousetech/1084?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> 🔄 https://darkolorin.com/posts/1083.html https://darkolorin.com/posts/1083.html Wed, 04 Jun 2025 19:48:34 GMT <img src="https://cdn4.telesco.pe/file/XsgX5xgg2LbInv7wKLR6V7kw5g43zsM8M_rKQqJ9lLySpH_RheFzBe4muKkqpIoioYMbmCv3gEJA74MRVW2-LoJhjRjFKvV675miA3hDtSZqwelE2k2QpnPGPtn_r7UvSAy2DwcNJPwXlnE388v4R_fL0OyaTMt8NJs0UAGzLPlwFpc3Xd_iuXD0VrGpIxHmTGoi_LS0lRLEZrc6rb3coFD1v68mEj9sElrod87_4mGODXtJCLroA0t4e5zUTmWvRLeHhPdfnP_xfC8YzKZ7VBuWZx8k9tzAMW23YC3qM0WVvogRy9PPLqgwquY3sDOx1YeD_e8MVbz2IZBqxOUO5A.jpg" alt="" loading="lazy"> <p>&quot;&quot;&quot;<br> 🔄</p> <p>article here (<a href="https://www.politico.eu/article/pope-leo-xiv-wants-stop-ai-playing-god/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.politico.eu/article/pope-leo-xiv-wants-stop-ai-playing-god/</a>)</p> <p>👍<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/1083?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sam Altman rolled out a new vision for AGI. https://darkolorin.com/posts/1082.html https://darkolorin.com/posts/1082.html <p>Sam Altman rolled out a new vision for AGI. Or for their next model (though he denies that in the talk).<br> Basically:<br> - the model itself doesn’t know anything (unlike current big databases)<br> - the model can do reasoning and actual thinking-through<br> - the model basically has an infinite context window (though yeah, tons of questions about the memory you’d need for that)<br> - the model can use all available tools (search, programming, testing, etc.)<br> The concept is clear. I’d say for me personally the next defining step is so-called “continuous learning,” when the model picks up new knowledge or skills right in the process of being used. Because even in his concept, any new tool will require retraining the model or something else.<br> Video from the Snowflake Summit.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1082?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh, Telegram rolled out a new feature for direct messages to me:) I set it for… https://darkolorin.com/posts/1081.html https://darkolorin.com/posts/1081.html Tue, 03 Jun 2025 18:37:21 GMT <img src="https://cdn4.telesco.pe/file/Vds4eAxefJ2ZmUj7YLg2ROrr73lHDcWm1Wq01WrwBRAqhm--Lblyns8jgdnywZjmd-Uuv7tUB5IfPplzbxaldhVs79ZguBx9qRsqhm_HGC5MSD7QzZLkB7Aqu_qFS3IIJEvSp3tTA8D--5C2HBnVP4asUqAfBq7MTALJBXDJ7ltuSvj77K2wnyutvbU9Wx81bEX05Bnwu8T_BtRd1eI8Jqqjx8St-ee-g_l_uBDmi4ohKV0YyIeHCcOoYjMxIltMAv8gDC7umsYcj4vkaWDMiobMqA9gHP6W67yDfxfw1GvP2iMaVY6OGLtjzjwySGAyPyfb0DMZpxczNGdPiDLv2g.jpg" alt="" loading="lazy"> <p>Oh, Telegram rolled out a new feature for direct messages to me :) I set it for a symbolic price of one star, just to filter out spam. We’ll see. <br> The dumbest thing is that once you turn it on you can’t *not* notify everyone. Yeah.</p> <p><a href="https://t.me/chillhousetech/1081?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> ❗️ https://darkolorin.com/posts/1079.html https://darkolorin.com/posts/1079.html <p>❗️</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1079?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, recently Mary Meeker put out a report — the same one who used to crank out… https://darkolorin.com/posts/1078.html https://darkolorin.com/posts/1078.html Tue, 03 Jun 2025 12:06:03 GMT <img src="https://cdn4.telesco.pe/file/QbfBsFa0BW8SHoee_ppJcP0FqZ_2P4W-bqxAgs2syUl7Ma6N3N4aGtWN1at0iIgqM2l59hlVeUmRbtc4g1bU9WAJZ_2gtl4-zggb8BG_H8eiJmqvFbMXarFFuYxOzFY_k2agko-HujmgclsSzGGuF4B62ClA3JtOvzCNiS7A57_zXhSCcK-BkZm2XpUMPZqpTaYHCO4E1cZ_zDcAgl7SFoVJfcWteA05-0LjgchyEvQ3yHHlfNwzwXS2pqNfTMh2m1RD7TEThdXZ1W_HOLCVLap4LJuGCO10ij1a_WATY_62FkN9N83VuFaKi9JtdRKaZSWq8P4gW4iW89vy_xTl1Q.jpg" alt="" loading="lazy"> <p>So, recently <br> Mary Meeker put out a report (<a href="https://www.bondcap.com/reports/tai?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bondcap.com/reports/tai</a>) <br> — the same one who used to crank out reports about the internet and whatnot. I don’t share that puppy-eyed delight; the report is apparently made for people who don’t read the internet or aren’t interested in the topic. Plus, the conclusions and logic are really stretched to fit whatever story. <br> Unprecedented Growth! <br> In reality I opened literally the very beginning and there’s slide 4, like hey look, Charts Paint Thousands of Words… (charts are worth a thousand words); Well what can I say, apparently she didn’t have enough words in her head to try to add even a tiny bit of context to the charts. I attached a screenshot. <br> 1/ like, the number of chip developers is growing faster and faster — well yeah, with the market growing, the number of people making shovels grows too, I honestly don’t understand what else you’re supposed to take from this <br> 2/ the funniest part is how she shows charts of number of users or later queries compared to Google — did it not occur to her that internet penetration was a tiny bit different back then versus now, and also everyone has a phone in their pocket and that massively affects adoption speed; all this “fastest-growing product ever” stuff is bullshit, I had one of those too, 100m users in 5 months <br> 3/ and again a chart that growth based on the internet is faster than the growth of the internet itself — yeah no shit, it’s easier to launch a train than to build a railroad <br> 4/ and on the 4th picture it even got kinda funny, I actually got curious — what was the growth at the beginning there in percent compared to now? <br> The whole report can be found in the comments.</p> <p><a href="https://t.me/chillhousetech/1078?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In one line: there’s this popular editor for artists, ProCreate, they put out… https://darkolorin.com/posts/1077.html https://darkolorin.com/posts/1077.html Fri, 30 May 2025 13:20:35 GMT <p>In one line: there’s this popular editor for artists, ProCreate, they put out a small manifesto. (<a href="https://procreate.com/ai?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://procreate.com/ai</a>)<br> Wonder how long they’ll last.</p> <p><a href="https://t.me/chillhousetech/1077?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Hold my beer upd: as usual, Pavel rushed it, posted before it was signed. https://darkolorin.com/posts/1076.html https://darkolorin.com/posts/1076.html Wed, 28 May 2025 21:47:50 GMT <img src="https://cdn4.telesco.pe/file/AXpch3O7UmqEcusifqkPl_xjUXi5Ruez8WjXxo9Y-KaH6fNeUVObKw5X12oowRS5Yflw1-pW35GlPAdaADCa5wGFgf7QvHzc8-gnuLY3oA3MiM8OLGXJWDz0cYN0deLfkIZWFOgolDgOpVMfA8NXVHRVsz0KR_OFLqTNGdsG9uMCEqEKA7r4fdt1rFnR-DHjFVZxtASRbWMxry8GIPLNYNcCs-7YQ4uyr6lJKs5ef_CjiXksTAyUVirC_mB9J8nWHSqdXXRogAAI9pgF0OHXYn3wtE3da2rfr9X2byT7prOnWMJ7taWWL722Aj0JAGSFjVH6-JxW1fzhBN8aNFc_yg.jpg" alt="" loading="lazy"> <p>Hold my beer (<a href="https://x.com/elonmusk/status/1927839555828220165?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/elonmusk/status/1927839555828220165?s=46</a>) <br> upd: as usual, Pavel rushed it, posted before it was signed. There’s his comment.</p> <p><a href="https://t.me/chillhousetech/1076?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Judging by the picture, at least for now. https://darkolorin.com/posts/1075.html https://darkolorin.com/posts/1075.html Wed, 28 May 2025 17:31:16 GMT <p>Judging by the picture, at least for now. I’m watching a lot of people leave OpenAI and try to start their own companies. (<a href="https://www.theinformation.com/briefings/former-openai-deepmind-leaders-seek-funding-chatgpt-material-science?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theinformation.com/briefings/former-openai-deepmind-leaders-seek-funding-chatgpt-material-science</a>) Usually, by the way, it’s domain-specific stuff. OpenAI itself, with its whole vibe, is signaling that they’re stubbornly trying to do business, even if it’s big and maybe ambitious. But in the push toward AGI, it seems to me they dropped the G along the way.</p> <p>As soon as the key people who were in the company on the path to AGI walk out and start building various nets for robots, for science, for whatever. The question pops up: why leave a company that’s supposedly about to get AGI in 2027. And you, as a scientist there, would get to tack your name onto the most important discovery in human history. Rhetorical question.</p> <p>Every month I’m more and more waiting for a reaction from all the “true believers” in AI2027.</p> <p><a href="https://t.me/chillhousetech/1075?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Haven’t had the “read it but didn’t fully get it” section in a while. https://darkolorin.com/posts/1074.html https://darkolorin.com/posts/1074.html Tue, 27 May 2025 22:37:57 GMT <p>Haven’t had the “read it but didn’t fully get it” section in a while. Alright, let’s go. <br> 1/ Ran into a short post (<a href="https://dynomight.net/smart/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://dynomight.net/smart/</a>) about AGI, where, IMO, it touches the question “what does it even mean to be smarter than a human and what would that give us.” Because the definition of AGI is blurry—what does “automating discoveries” mean, acceleration? As for me, being afraid of super-intelligence in the current reality isn’t even worth it. Interesting. <br> 2/ thought Roblox is rolling in profit and living the sweet life? I like Matthew’s articles (<a href="https://www.matthewball.co/all/roblox2024?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.matthewball.co/all/roblox2024</a>). <br> 3/ Runa has some entertaining stats (<a href="https://runacap.com/ross-index/q1-2025/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://runacap.com/ross-index/q1-2025/</a>) on growing open source startups <br> 4/ read Schopenhauer? If not — I recommend it, Einstein really loved his work on the will <br> 5/ gonna fork the open source version of cursor (<a href="https://voideditor.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://voideditor.com/</a>) haha <br> 6/ Last summer Goldman Sachs put out an interesting analysis (<a href="https://www.goldmansachs.com/images/migrated/insights/pages/gs-research/gen-ai--too-much-spend,-too-little-benefit-/TOM_AI%202.0_ForRedaction.pdf" rel="noopener">https://www.goldmansachs.com/images/migrated/insights/pages/gs-research/gen-ai--too-much-spend,-too-little-benefit-/TOM_AI%202.0_ForRedaction.pdf</a>) about AI. We’re spending a lot and getting little benefit. <br> 7/ beat a couple levels in the game (<a href="https://apps.apple.com/gb/app/level-devil-not-a-troll-game/id6477413486?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/gb/app/level-devil-not-a-troll-game/id6477413486</a>) — “relax” <br> 8/ the most promising option for a theory of everything (or unifying quantum field theory and gravity/GR) is the holographic principle. Lex has a decent explanation (<a href="https://youtu.be/1_ibTNDV8aU?si=RGKcA_DVhGuByhQw&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/1_ibTNDV8aU?si=RGKcA_DVhGuByhQw</a>) from a guest. And in general I recommend reading up on it.</p> <p><a href="https://t.me/chillhousetech/1074?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Buy now, cry later — BNCL (c) The new paradigm of the world: vibe coding. https://darkolorin.com/posts/1073.html https://darkolorin.com/posts/1073.html Tue, 27 May 2025 22:21:31 GMT <p>Buy now, cry later — BNCL (c) <br> The new paradigm of the world: vibe coding.</p> <p><a href="https://t.me/chillhousetech/1073?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Lately I’ve been running into product again — not just because we’re building… https://darkolorin.com/posts/1072.html https://darkolorin.com/posts/1072.html <p>Lately I’ve been running into product again — not just because we’re building one ourselves, but because this problem has gotten more obvious in general. Wanted to give an example using Perplexity. This post is a pure sketch of some general thoughts btw.</p> <p>Feels like the perfect case. The AI era is only just starting to roll out and this guy shows up and says: <br> “we’ll beat Google through UX (<a href="https://www.youtube.com/watch?v=e5utruJd6Gk)”&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=e5utruJd6Gk)”</a> <br> (somewhere in that video). And just recently I stumbled on an absolutely wonderful talk by the game director of World of Warcraft (<a href="https://www.youtube.com/watch?v=3b7uWPEN0Ok&amp;t=14853s&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=3b7uWPEN0Ok&amp;t=14853s</a>). In this post I’ll give my opinion and projection, but the presentation is so damn good (even if it’s heavy on game details) that I recommend it to everyone.</p> <p>So. I want to tie the fight in the AI landscape to this interesting phenomenon called a desire path. It’s that exact thing everyone yells: “listen to your user.” And what we usually get is like in the picture (we want one thing, first we defend, defend, then we give up, and when we give up it’s still not enough) — trying to force your solution at the start can be dumb, and then later it makes sense again. The whole “thinking in paths” concept is pretty interesting in general.</p> <p>Perplexity goes straight in, attacks head-on, like “look, this is how it should be,” and then gradually weakens because everyone around is literally doing the same thing. And the guys stay stuck in the paradigm of “we solve problems that can’t be solved perfectly.” Design has that property — yeah. Instead of moving forward, the team starts hunting for maximum applied value (<a href="https://x.com/aravsrinivas/status/1926690746159034709?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/aravsrinivas/status/1926690746159034709?s=46</a>), which, unfortunately, just isn’t there. Because people think with emotions and constantly change their minds about already-shipped features: first they’re thrilled, then they go “so why aren’t we developing this.”</p> <p>Novelty has inherent value, making static solutions short-lived (c) — in other words, novelty matters in and of itself, and you simply can’t ride one trick forever; It’s a big miss to try to saddle the wave of a “successful feature” and repeat it again without building long-term value. That’s why we see a browser, ads, whatever — everything Google has that worked. They’re trying to repeat their initial success by just copying what already existed. But no emotions show up along the way, and I think that’s exactly why the guys need to change the approach at the root.</p> <p>Everyone go watch the talk, the specifically game-y topics are easy to skip.</p> <p>Links from the original post:</p> <ul> <li><a href="https://www.youtube.com/watch?v=e5utruJd6Gk&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=e5utruJd6Gk</a></li> </ul> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1072?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Weekend posts. https://darkolorin.com/posts/1071.html https://darkolorin.com/posts/1071.html Sat, 24 May 2025 22:30:11 GMT <p>Weekend posts. A few theses that seem important to me to understand if you’re about to build your own startup, product, whatever. <br> 1/ Yeah, you need to learn all those growth playbooks. And then throw them out and write your own. <br> 2/ Almost nobody thinks on a 3+ year horizon. And for nothing. The long game lets you beat 90% of competitors. <br> 3/ MVP isn’t a choice. It’s just you not knowing how to make a normal product. If you can, better ship something useful right away, not “viable.” Especially in the age of AI. <br> 4/ Designers are just as important as engineers. <br> 5/ Startups aren’t killed by competition. Well, or that’s a big rarity. <br> 6/ Growth isn’t linear. It’s sigmoids, one after another. At best it’ll look like an exponent if you zoom out. <br> 7/ Hiring isn’t an interview, it’s a funnel. The earlier you start, the better. <br> 8/ Want to hire? Do it in advance. When you already need it, you’re late. <br> 9/ Nobody gives a shit what you’re doing. Nobody will remember. Failed, bailed, stole (Kekw)? Great. Start over. <br> 10/ Enterprise sales are about communication. Product comes after. <br> 11/ Cold email blasts don’t work. Forget it. <br> 12/ SEO works. <br> 13/ You need to fight churn-rate at the moment of signup. Later it’s way harder. <br> 14/ Offsites work. People get closer. It matters. <br> 15/ Brand matters more than you think. <br> 16/ Monthly investor updates are a must-have. You structure your own head, and they (investors) can be useful. <br> 17/ And yeah, the feeling that everything is total fucked-up is a privilege. Usually everything is boring and bleak, and the older you get, the clearer that becomes. <br> 18/ Knowledge of how things work down to the details matters; without understanding it, you won’t quickly build either a team or a product.</p> <p><a href="https://t.me/chillhousetech/1071?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, I couldn’t not test it. https://darkolorin.com/posts/1070.html https://darkolorin.com/posts/1070.html Sat, 24 May 2025 11:42:07 GMT <img src="https://cdn4.telesco.pe/file/JEgtdTPKO4N3uCfvoRtyBaW_8TYM9yLDjyhdUkHRqyWBl-K7XuRlf3vPzPNfGDfDF2BhL0bjVTrbeV0Im6j_iD7JPoz7eAU4iP_etO_eOf8zcytjd8k28VSy7sMl85LYHq412xTa4CinxUVZo_SFMPd_KiU2iww-MLBIWVnpvfTL7Gyrs33WQcHTKhqS12JMjAJBrP95qBwMklpw0hDmLynK4YM-SwoVjezL5G4YdWy_W0Uvzcv8hgF-KL5txF6-8Aco-7oy-hFdksyPnekBmTm3Vz7357lfpWJ8a9WS5tc4Fd8va9LdayvTHek3fr4puDtOqP3BNDgmXOyrlVGNtA.jpg" alt="" loading="lazy"> <p>Well, I couldn’t not test it. I think Anthropic’s CFO can chill.</p> <p><a href="https://t.me/chillhousetech/1070?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One-shotted via Claude 4 in Cursor. https://darkolorin.com/posts/1069.html https://darkolorin.com/posts/1069.html Fri, 23 May 2025 13:11:31 GMT <p>One-shotted via Claude 4 in Cursor. All I had to do by hand was set up the Telegram bot and all that. Kinda cool, though who even needs it I don’t know yet, also noticed that narrow bits, like specific functions, it writes great in iterations, but architecture and bugs are still rough.</p> <p>Yeah there are bugs, I know, the point is it’s literally one prompt with no fixes (the window seems to close after you die)</p> <p><a href="http://t.me/Flappy_Google_Game_Bot/flappy_google_earn?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://t.me/Flappy_Google_Game_Bot/flappy_google_earn</a></p> <p><a href="https://t.me/chillhousetech/1069?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So yeah, Anthropic dropped Claude version number 4. https://darkolorin.com/posts/1068.html https://darkolorin.com/posts/1068.html Thu, 22 May 2025 16:35:47 GMT <p>So yeah, Anthropic dropped <br> Claude version number 4 (<a href="https://www.anthropic.com/news/claude-4?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/news/claude-4</a>) <br> . Waiting for the details and the benches and to test it out</p> <p><a href="https://t.me/chillhousetech/1068?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> That’s how you sell something that doesn’t even exist. https://darkolorin.com/posts/1067.html https://darkolorin.com/posts/1067.html Wed, 21 May 2025 17:46:43 GMT <img src="https://cdn4.telesco.pe/file/oGemiyTsYAfVeIxhODEu7ti4lihteWd5A2Yf_SxZ8WoYYXdRMOYJGyij1W1pDsFhCdqi5R4GMPx4dFG04MnaVo-lkgkpGpGSVTUvp7WCqg3wh6KijxN2qTC93dItXj6wi5kSltYkiKk6-rmfHSvER2WLom8A44gPi8dzsyzOmLKyymSDF_DbiY9Z8Jf544nRlaLuh2q5OZb3HmGfc4HLbLAi1HlO16WlE6MZGTuxHl5lZNCGo8sKCP91yt7r6SQUcFNHYYtifA-OkaF4_IREe-aj4FWScc2qVLOj2hp80Y7K3Mt45YT586WIk1RU4j35DZTXrAUWnAWPYCpVeen_sw.jpg" alt="" loading="lazy"> <p>That’s how you sell something that doesn’t even exist. And these are the kinds of transfers in the tech тусовка. Worth noting the deal is in stock. So we’re waiting for devices from OpenAI then?</p> <p><a href="https://t.me/chillhousetech/1067?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Thinking about all that RL & reward hacking in the o-series models and so on… https://darkolorin.com/posts/1066.html https://darkolorin.com/posts/1066.html Wed, 21 May 2025 14:47:44 GMT <p>Thinking about all that RL &amp; reward hacking in the o-series models and so on, I remembered this paper: (<a href="https://arxiv.org/pdf/2401.08273" rel="noopener">https://arxiv.org/pdf/2401.08273</a>)<br> You can deliberately provoke hallucinations in an LLM and get a metrics boost out of it.<br> Instead of a normal zero-shot prompt like:<br> &quot;Answer the question...&quot;<br> they add at the start:<br> &quot;Look at the examples in the Examples section and use them...&quot;, even though there’s no such section in the prompt.<br> The model “recalls” non-existent examples and leans on them as if it were few-shot.<br> On a bunch of benches it gives a noticeable boost (though the paper is old and they still had old nets in there).<br> The effect almost doesn’t work on the gpt4 series. It’s better at catching that the context is fake. But it totally works on the new o series (judging by posts on twi, but need to check). This pushes you toward the obvious thought that you can kind of nudge the model into out-of-distribution reasoning by saying it’s actually in the training set. To me this effect kinda resembles those little phrases like “let me check again”, “hm” in the reasoning. Like, they’re trying to sort of point to examples that are supposedly definitely correct and then copy the reasoning/approach from there.</p> <p><a href="https://t.me/chillhousetech/1066?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well it works quick quick (feels basically real time, really) https://darkolorin.com/posts/1065.html https://darkolorin.com/posts/1065.html <p>Well it works quick quick (feels basically real time, really)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1065?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A very, very important vector. https://darkolorin.com/posts/1064.html https://darkolorin.com/posts/1064.html Tue, 20 May 2025 17:56:14 GMT <p>A very, very important vector. Architectures are getting mixed. And diffusion in text, well (<a href="https://deepmind.google/models/gemini-diffusion/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://deepmind.google/models/gemini-diffusion/</a>) — they’re promising a very fast 1500 tokens per second. And this is just the small version for now. This should seriously speed up adoption.</p> <p><a href="https://t.me/chillhousetech/1064?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Purely by accident I stumbled on a Codex analog from Google. https://darkolorin.com/posts/1063.html https://darkolorin.com/posts/1063.html Mon, 19 May 2025 23:12:15 GMT <p>Purely by accident I stumbled on a Codex analog from Google. No releases. Just lying around on the internet as an experiment. I liked it more. <br> <a href="https://jules.google/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://jules.google/</a></p> <p><a href="https://t.me/chillhousetech/1063?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sunday thoughts. https://darkolorin.com/posts/1062.html https://darkolorin.com/posts/1062.html Mon, 19 May 2025 00:03:12 GMT <p>Sunday thoughts. Long story short: putting effort into a product manager career is, in my view, one of the worst directions in tech right now. Zero protection from risks, the potential upsides are there then not there then boom, new rules of the game again. Almost all career growth boils down to one skill, “leadership,” and one trait: “I take risks — I don’t break along the way.” All the practices and advice go stale so fast there’s no time to learn new stuff while you’re still trying to dig through the old.</p> <p>So I suggest digging toward:<br> - leadership<br> - tech skills to do something (basically, building something concrete with your own hands, at least prototypes you’re not embarrassed to use)</p> <p><a href="https://t.me/chillhousetech/1062?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Vibe coding? https://darkolorin.com/posts/1061.html https://darkolorin.com/posts/1061.html Fri, 16 May 2025 18:15:33 GMT <p>Vibe coding? Drop in the comments what cool stuff you’ve ended up with :)</p> <ul> <li>Yeah, and actively</li> <li>Yeah, and it even makes money</li> <li>Yeah, it helps with work and stuff</li> <li>Tried it a couple times, not very actively</li> <li>No, my brain is better than all your LLMs</li> <li>👍</li> </ul> <p><a href="https://t.me/chillhousetech/1061?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Since I’m pretty clearly not in the camp of people who bought into the AI 2027… https://darkolorin.com/posts/1060.html https://darkolorin.com/posts/1060.html Fri, 16 May 2025 15:46:23 GMT <p>Since I’m pretty clearly not in the camp of people who bought into the AI 2027 forecast, even though I think it’s done quite well and I respect the authors. So here, I’m digging up some crumbs of facts against it:</p> <p>Meta are delaying the release of their biggest model (<a href="https://www.reddit.com/r/singularity/s/MUtbF3GS4D?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reddit.com/r/singularity/s/MUtbF3GS4D</a>)<br> with quotes like, well, it’s supposedly basically the same as the previous one (my interpretation). Which points to a possible not-following of the scaling laws.</p> <p>Also, I’ve got a really weird analogy spinning in my head. I studied a bunch of physics for quite a while (something around GR and quantum mech on the bare minimum), so here’s a hunch. The current scaling laws remind me of the laws of classical mechanics. Like, they work at a certain scale of energies/distances. But they stop working properly at extreme values (and at the very edge and for the latter there are problems too but that’s a separate topic). So yeah, it seems to me it’s the same deal with the current AI scaling laws. That for now they’re true, but as compute and number of parameters and whatever else grows, other details might show up. Don’t judge too harshly, I was trying to throw some philosophy at the fan.</p> <p><a href="https://t.me/chillhousetech/1060?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A year or two ago Sam Lessin gave a great presentation about the venture… https://darkolorin.com/posts/1059.html https://darkolorin.com/posts/1059.html Sun, 11 May 2025 14:25:31 GMT <p>A year or two ago Sam Lessin gave a great presentation about the venture market and I even pieced it together from DocSend by hand (<a href="https://t.me/chillhousetech/380?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/380</a>). So yeah, there’s a small update out—if you’re into venture stuff, I really recommend reading it. Not like I agree with every argument, but still the main narrative is roughly this TLDR:</p> <p>VC IS RAPIDLY REBUILDING ITSELF AROUND A POST-FACTORY MODEL… IT’S NOT AN ASSEMBLY LINE ANYMORE, IT’S A REGATTA; NO STANDARD ROUNDS OR METRICS</p> <ul> <li>there’s no single VC market anymore… even seed isn’t a “market” anymore. Everyone turned into capital allocators, each one solves their own problems with increasingly custom strategies by stage, sector, etc. And in a world where nobody understands what anything costs, the one with the strong story and good “options on infinity” (i.e. scalability and the possibility of a big pop), plus preferably at least some downside protection, wins.</li> </ul> <p>SFD (Story, Founder, Deal) &gt; “tech”<br> - Tech is getting commoditized… now it’s more important to be able to tell a story, to sell, to attract and inspire. You need founder-“long-haulers,” because it’s gotten genuinely hard, and an exit/sale is now not fast at all (used to be ~4 years on average; the author is betting on 10+). You need truly engaged people. And yeah, deal structure now decides everything: at early stage the main thing is not to burn out immediately; at late stage you need a story that can raise and digest big money.</p> <p>KYC / KYB (know your counterpart/business)<br> - Who’s funding you and WHY, really? What do they want? Who’s your next buyer and why will they buy you? What are their incentives? The world is getting more closed, and VC too. The era of global free market is going away for a while at least. Now everything revolves around individual deals, personal connections, and specific interests. You have to understand who your partners are, and know exactly what you’re selling and for what money.</p> <p>If anyone needs a web link to the docsend, here (<a href="https://docsend.com/view/fm356rwwb68n8qti?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://docsend.com/view/fm356rwwb68n8qti</a>)</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1059?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> As usual, YC’s every-semester wishes and recommendations on what they’re… https://darkolorin.com/posts/1058.html https://darkolorin.com/posts/1058.html Fri, 09 May 2025 00:25:24 GMT <img src="https://cdn4.telesco.pe/file/DBBMK9TJXkL7lite_aDtzwXinWF7dSdADq2Zz3LP4FNqjyVHOJtag7Eu4oqq5_jzz-T3PBWq4E2G5A3VEHyMRU1eps4_4tdX3bsSeMxc-P26ZvkU38Jx8JCXM0CizJ6Uj8HnRFcTIWCSXhuBoZB2iTBpXX0VMZeKOpjx2EGvW6FZ-Ke2HURnAjnH0_SEnSwxc8SaFPa-aGka8ZR7P83FaI_nPdfL3eT8pvoBa3rkb53junar30BF1YvNDGapw_BSrwme2NPE8Ekd_KkK8JpFhFqSDacgoRhDNcXEIrXBaBfNh17hPnvx4O2Rz0pOvoBQKfD41zX67WByehFaxLESHA.jpg" alt="" loading="lazy"> <p>As usual, YC’s every-semester wishes and recommendations on what they’re looking for in startups.</p> <p><a href="https://t.me/chillhousetech/1058?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One line. https://darkolorin.com/posts/1057.html https://darkolorin.com/posts/1057.html Wed, 07 May 2025 22:22:09 GMT <p>One line. <br> Stripe made its own foundation model. (<a href="https://x.com/thegautam/status/1920198569308664169?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/thegautam/status/1920198569308664169?s=46</a>) <br> No details though. But this sounds insanely cool: <br> &gt; We built a classifier that ingests sequences of embeddings from the foundation model, and predicts if the traffic slice is under an attack. It leverages transformer architecture to detect subtle patterns across transaction sequences. And it does this all in real time so we can block attacks before they hit businesses. <br> So damn cool! This is exactly what I’m waiting for from AI and models at the first stage.</p> <p><a href="https://t.me/chillhousetech/1057?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One of the many questions that constantly tortures me is this one. https://darkolorin.com/posts/1056.html https://darkolorin.com/posts/1056.html Wed, 07 May 2025 11:35:29 GMT <p>One of the many questions that constantly tortures me is this one. <br> “Why does a person who’s been yelling in the media space for two years about what an awesome specialist/businessman/manager he is, at the end of a career stage/startup try to sell courses/consulting?”</p> <p>In my model of the world, which may well be wrong. There’s this thing called “market value” and the ability to keep moving forward. Why, being a founder or a top manager, would you then go and make a course?</p> <p>What’s more, I notice one very simple thing: the faster a person is growing right now, the less time they have to reflect on that growth, because you just don’t have time. Most of the people I know who raised a $30M Series B aren’t writing hundreds of posts, aren’t selling courses, or teaching everyone around them how to do business.</p> <p>From this I draw a simple conclusion: <br> - either you’re already so badass that you can just give out all the info for free <br> - if you suddenly start trying to monetize — something’s off, and you’ve got no other growth point <br> - you’ve made it as a professional and you don’t need anything else, zen</p> <p>But this has gotten firmly lodged in my head: any initiative outside your pro track is a reason for you personally to stop and think: why the hell do you need it. The answer is often — money.</p> <p><a href="https://t.me/chillhousetech/1056?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Came across a pretty useful OpenAI doc by accident on how they do API sales… https://darkolorin.com/posts/1055.html https://darkolorin.com/posts/1055.html Mon, 05 May 2025 15:28:41 GMT <p>Came across a pretty useful OpenAI doc by accident on how they do API sales and integration; overall it lines up really well with what I’m seeing. Recommend checking it out (there’s nothing blatantly genius in there, but it’s always useful to systematize stuff);</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1055?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> No real time for big posts yet (haha not even 5 minutes). https://darkolorin.com/posts/1054.html https://darkolorin.com/posts/1054.html <p>No real time for big posts yet (haha not even 5 minutes). Google is rolling out AI mode (<a href="https://labs.google/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://labs.google/</a>) in Search to everyone in the US, the richest market.</p> <p>What matters here is this. Cost per query is the most important metric. If Google is shipping it to everyone, that means the projected cost per query has dropped so much that sane monetization via ads is possible. Most likely inference on CPUs. Or, well, some insanely cheap TPUs are about to show up.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1054?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> btw I’ve been watching Sergey Nikolenko’s lectures for quite a while now… https://darkolorin.com/posts/1053.html https://darkolorin.com/posts/1053.html Mon, 28 Apr 2025 10:44:36 GMT <p>btw I’ve been watching Sergey Nikolenko’s lectures for quite a while now, really like how he presents the material and the sequencing, plus everything gets updated, and also I myself used to go to lectures at POMI RAS on Fontanka back in the day, if anyone’s interested I recommend the channel <br> <a href="https://t.me/sinecor?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/sinecor</a> <br> and there he usually shares his lectures too (videos on YouTube) <br> here’s the latest thing I watched about scaling laws <br> <a href="https://www.youtube.com/watch?v=FyRRo61TNt0&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=FyRRo61TNt0</a></p> <p><a href="https://t.me/chillhousetech/1053?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Deel, the company that does payroll, contracts and so on, is pitching pretty… https://darkolorin.com/posts/1052.html https://darkolorin.com/posts/1052.html Sun, 27 Apr 2025 11:40:42 GMT <img src="https://cdn4.telesco.pe/file/ZKG1PAiFoQEKvVmwLayvSsN3azbrFoGrVXkBl_-0JP936nry7_y0U6dNykvNmCVhrztyRZF0lAzFpCNgGGAzD1GVdL6entYFvzX0kkYyGIHtE1bwSJ-WqvuBH3qks7IHHwHeUMUGBcqy7DdPMy6U41PQGqfdibkW5ewzLcrAqtUpk4s_Gx75QMSosGaTHtx9FkSzw8eBc4YS_ifHFaXKyWdrVLH-B2madBGFC9_PEIH-jHpks3OOyJtUD7UKOTilB8WMyN4G35xLm9yMuKzWoJ1M4kMQNXBoJZO0u4iUgR2GvhP4X-QXruQ6PUnrkKB1LbsUJXQmvEaNIy62OyWYtA.jpg" alt="" loading="lazy"> <p>Deel, the company that does payroll, contracts and so on, is pitching pretty well. Turns out to make payouts you need quantum computers. A Sunday one, so to speak :)</p> <p><a href="https://t.me/chillhousetech/1052?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Bro, it’s on blockchain…it’s not just an image (c) Lmaooo. https://darkolorin.com/posts/1051.html https://darkolorin.com/posts/1051.html Thu, 24 Apr 2025 14:57:42 GMT <img src="https://cdn4.telesco.pe/file/sG1lHeGM7u6f3XcG52vAjuJhE_h01udyjFxCFB4YydvtGwFobGConfrVVJaNzgv3mKqK7OXWObyX1RwgsHHD57EhHSBCNlQnNy6u82VWHlE_3xTr6NlaJox1oUBUXjppLb34T8agYOMS0zn5ngVOo31jUZoNVFfnmmzG93TadTsMHRKSgWFckk-ouc_F_8ZYhMwCxqD3TM7QEvdHXad0cG_I11UjOiumeYAmD0taf6mCz-7dNv2SxToIkpODFxgz5YjknIqHPup28i5fPdAjGbqsFy-w3Un0KApbctXTP83HBbo7jt62-lJFowPiB0W1xVLazf60zeoa80rFjmrmfA.jpg" alt="" loading="lazy"> <p>Bro, it’s on blockchain…it’s not just an image (c) <br> Lmaooo. Remember that, yeah? Digital art, blockchain, “you just don’t get it.” Well how do I put it, I got it then and I still get it: crypto/web3 is a legal casino, tokens and everything else is a derivative of that casino. And other than a casino it can’t be anything. <br> Stablecoins and BTC are kinda off to the side.</p> <p><a href="https://t.me/chillhousetech/1051?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So what about the models for a gazillion yards and AGI soon? https://darkolorin.com/posts/1050.html https://darkolorin.com/posts/1050.html Wed, 23 Apr 2025 10:22:09 GMT <p>So what about the models for a gazillion yards and AGI soon? There won’t be enough money for everything. Or am I not getting something? Nobody’s taking stock for it.</p> <p>OpenAI wants to snag Chrome (<a href="https://www.reuters.com/sustainability/boards-policy-regulation/google-contemplated-exclusive-gemini-ai-deals-with-android-makers-2025-04-22/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reuters.com/sustainability/boards-policy-regulation/google-contemplated-exclusive-gemini-ai-deals-with-android-makers-2025-04-22/</a>)<br> if they end up selling it.</p> <p><a href="https://t.me/chillhousetech/1050?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Recently we were in Spain. https://darkolorin.com/posts/1047.html https://darkolorin.com/posts/1047.html <img src="https://cdn4.telesco.pe/file/CgpZedRXQtAc-zlR6Vv7-tY18hqCigfO48elmwWatPMLLNYpd4EHsebxoa-0ZlLtsOy_GAiiJ5GWEjfjSHo_GBFWe1b0Cjz7acxmY-vEl7nwrDZepPCDNbo6teZRH5DeRL2hbPNoCQicJsd1ETLBOqv4XJn0WPPiYY8gWByQD92ij2xSmM2YNVPjh1U9rRFPYS_eCqdZZDYVXs67kFATmfhz_oxR8DABWhvPCyxdzy5tUdCXEJvxu7gg3rTT6nR4qmWt7Wa9mDszXPC75WqVGv_82FbVt3n23xE2WcFRqILHfCoLa1qz9ylJqixwFbl13tQQfkZY14erdUe_AmO1yA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/aB2SoH85OkB6FLL8EOT43bcFwXajAu6gk0fZEMftYqSZvbKxa-evjFoO6z4oWEuNc3XsiHZFfjaWoQZYRF9kZ3qFn63-7GSoiflsZHcm3MSntVvG4VCwpIzSnGa8w30irMVH9mYDYEmnOfQH5T_b3mKiPMqo_3eVHIxfUnBN3uriyo1rKIfFr4is_a5DQGWRcjjcR8GV0HQwk8-MqD_T8Bd3Diohb_bztoJXx6X32Q-2q5IsqUisT6ZC8LAPz8Zy_PSXqTKvZU-vf-YW836NgbHg8g48OUZKJzvoBcwqOLWDWXu7x8ilVMTjDws9LYuqFCefmJp_aiwVl2e_ahw3lw.jpg" alt="" loading="lazy"> <p>Recently we were in Spain. And then I ran into this thing like wow what a smart o3 net is at finding objects. So I decided, so to speak, to feed it a hard photo.</p> <p>Recorded the whole process — 5 minutes of research and a miss. And I attached a competitor too (the competitor answered correctly in a split second). Funny that search in Apple Photos, where there are coordinates, gave a wrong result :)</p> <p>What’s interesting:<br> - the net chops images into pieces with Python<br> - searches the web for these chunks of the photo<br> - recognizes the cross and looks for it<br> - even searches on Facebook</p> <p>P.S. you can watch the video sped up, they recently brought that into Telegram I think</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/1047?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The other day a really interesting piece came out about how “honestly” the new… https://darkolorin.com/posts/1046.html https://darkolorin.com/posts/1046.html Thu, 17 Apr 2025 19:55:39 GMT <p>The other day <br> a really interesting piece came out (<a href="https://transluce.org/investigating-o3-truthfulness?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://transluce.org/investigating-o3-truthfulness</a>) <br> about how “honestly” the new o3 net from OpenAI works (and all these new ones around it with “reasoning”, the versioning is killing me and I can’t remember it exactly).</p> <p>Researchers at Transluce AI got early access to the o3 model and found that it regularly makes up actions, like it’s running Python code on “its MacBook Pro,” and then goes into detailed justifications for these nonexistent runs when it gets caught screwing up. Examples are in the article (checking whether a number is prime and so on).</p> <p>The most interesting part is: why does this happen, and it really looks like it’s all about RL (reinforcement learning). The authors’ hypotheses are:<br> 1/ Outcome-based RL — the model was trained for “the right answer,” not for honesty. If it’s unsure, it’s easier to “guess” and get the reward.<br> 2/ Truncated chain-of-thought — internal reasoning gets wiped between turns, so when the user asks “what did you do earlier?”, the model has no truthful context and has to improvise.<br> 3/ Typical LLM problems: hallucinations, sycophancy, reward-hacking.</p> <p>This issue is specifically with the O series, not with base GPT, which kinda hints that the questions are aimed exactly at RL (though maybe they just poured in a bunch of synthetic chains without checking).</p> <p>As usual, the little tip at the bottom from ChatGPT itself has never been more on point: “ChatGPT can make mistakes. Check important info.”</p> <p><a href="https://t.me/chillhousetech/1046?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Finally https://darkolorin.com/posts/1045.html https://darkolorin.com/posts/1045.html Wed, 16 Apr 2025 18:40:06 GMT <img src="https://cdn4.telesco.pe/file/dnPOxXeMBmxj-VAftavCiVSeTNj0szfxYdHg99WKcZMG0tJHHuoc3D4FxygwSAgBegfoEDUJ01w5_pz6OoHN1984AJ1MK4OeFNBkuAndkEoWEsRTBnTal1sVRGsB6NE8zQJyKd0GpWgsmwo4DRGo0wk_Ux16H0aVIxqw2Rn6QTvogYVKVCKzQtUgNtzp177mhuf_LUuP6hoG-l021FK_RPndWGzMrqGHgLW4hS2maK1lmtVCESojE5u-13o0oGlms8NIKJFJHHmbOEGr_s7ZCQqFfV4d7oYYnoMsAMEtdpJJOgRfn1A0xsVkzBzrwYoYHqqkLEMaUIrBwKPmO19lew.jpg" alt="" loading="lazy"> <p>Finally</p> <p><a href="https://t.me/chillhousetech/1045?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ll write a bit about my experience using LLMs after two years, so to speak. https://darkolorin.com/posts/1044.html https://darkolorin.com/posts/1044.html Tue, 15 Apr 2025 23:40:24 GMT <p>I’ll write a bit about my experience using LLMs after two years, so to speak. <br> First off, what I usually do:<br> - throw together UX prototypes as a front/flows and all that — designing features like this conveys the essence of what you want to build better<br> - make pitch decks, business models, and all that kind of crap<br> - write lots of meaningful emails like: so why the hell do you need this feature and where will it lead, or essays on the theme of “strategy is everything, and I don’t give a shit what we ship tomorrow”<br> - think a lot, meet with people, and write follow-ups</p> <p>Tools:<br> - ChatGPT<br> - Claude code/cursor<br> - advanced voice mode ChatGPT</p> <p>For myself I figured out a few commandments / rules for how to succeed working with LLMs:<br> - well if you started doing something and you see it went somewhere wrong or forgot some important details — I start over. Nothing will help fix or add something into the already existing “stream of consciousness” of the model<br> - it’s extremely important to formulate for yourself what result you want (a site similar to X, an Excel with these fields and these dependencies) ahead of time, and even better to understand how you yourself would do it by hand. Then everything goes way nicer. Like, I ask it to generate Excels while specifying it needs to write formulas, not just numbers in a table. Then it comes out better.<br> - vibe coding is a pretty tool. But if I’ve never written code for “something” before, I won’t assemble anything besides a website. Because in the process you need to understand the details and sometimes the essence of the errors, and they will happen. If you don’t understand the essence of the errors — the sheet of changes will be out of control. Then see p1 (and you don’t even understand that something went wrong)<br> - but the real top is to dictate thoughts right after a meeting in voice mode and iterate by voice too — it lets you very quickly formulate the main outcomes and write a balanced analysis for the team or as a reply</p> <p>Personally my efficiency has increased by multiples. I used to be really lazy about building models and calculations by hand, and now I do basic code for something or an Excel super fast. 50x speedup. Since my main difficulty was that without constant use I quickly forget the tools (Excel, front-end, or complex calculations), I just take forever typing in texts and formulas and constantly screw up syntax. Obviously I edit texts with the history in mind and often dump all versions of decks or emails into a folder/project in ChatGPT.</p> <p>TLDR: for me this is still far from automation. And honestly I don’t believe this threshold can be stepped over with the current generation of the tech. It’s super cool. But it feels like an electric car from <br> Detroit Electric. (<a href="https://ru.wikipedia.org/wiki/Detroit_Electric?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ru.wikipedia.org/wiki/Detroit_Electric</a>)</p> <p>P.S. haven’t figured out why the hell to generate video and pictures yet</p> <p><a href="https://t.me/chillhousetech/1044?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So I guess that’s when we won’t be needed anymore, yeah? https://darkolorin.com/posts/1043.html https://darkolorin.com/posts/1043.html Tue, 15 Apr 2025 14:12:27 GMT <img src="https://cdn4.telesco.pe/file/vIpwolwpqEH7ciQgWG6HM9CEgYebrp8mOBBXaf1YhVQd13dFaU43re4Cz-CZLmJqD5RL3nwmQq3OYfU320IV6bdFgEqdzEt2Afk48NnXAd_ITCQXXGKFxtgmeU5yrWaO1lSNsU6tDF2mZGu32n8NDawyDYItbk4HvjItRDM90Su6XToWczxqmsAr8jzh8G3cTP3BJbaM9pMWgyXmmDb5uBCpPojWPKHzKqZeBxLXJKm-kBCaL0mH3_MHnA0-VlT4GlQC4NmFeLCJM8MvaiCrk0bqf3LZUk7dTF-xtzBH0NjTqsArQgFMMF2lFKN4LL08i7FxC3ZA4tZnDOouVUSJhw.jpg" alt="" loading="lazy"> <p>So I guess that’s when we won’t be needed anymore, yeah?</p> <p><a href="https://t.me/chillhousetech/1043?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Do I need to write about releases of new models? https://darkolorin.com/posts/1042.html https://darkolorin.com/posts/1042.html Mon, 14 Apr 2025 19:03:06 GMT <p>Do I need to write about releases of new models? I kinda don’t want to, because the benches are broken (imo) and for now the progress is minor.</p> <ul> <li>Yeah, details matter</li> <li>Whatever</li> <li>No, there are still way too many of them there</li> <li>😪</li> </ul> <p><a href="https://t.me/chillhousetech/1042?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> That’s how success looks like 👍 https://darkolorin.com/posts/1041.html https://darkolorin.com/posts/1041.html Mon, 14 Apr 2025 18:52:47 GMT <img src="https://cdn4.telesco.pe/file/KewOv_mKUpPGxEZrZHQyZ-tuJLCDgM2joeVGtiAG_mRLMO5xCY5ezTjZqMDre3gioLqXvakIUtiZFCdAB6xV9z7WDbQh-8oaRHbBOvHje_bplycM9_tPxm0NnovBw8MD29btb6377E061F1LDKq0oYZWjuBOsJZP9B6zQyCOvfU2SjJobPcLfD1h-t43ILMZcEkNA7YGWUDLpPJNvVvEa8ng-ed3quFN_zABAnPgY7FdsmFSnpzT_MITwjdFZouDtCoSGuYdX5CyIjvoSJXt5aAwMI_J0RdcAl3mPY5Xf-0AyBwigvmKK2eRHCLcb4g0VV3wFZu8HBgohNNtiXs9HQ.jpg" alt="" loading="lazy"> <p>That’s how success looks like <br> 👍</p> <p><a href="https://t.me/chillhousetech/1041?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “Read, but not fully understood” column. https://darkolorin.com/posts/1040.html https://darkolorin.com/posts/1040.html Thu, 10 Apr 2025 06:02:01 GMT <p>“Read, but not fully understood” column. Alignment Edition (c)</p> <p>1/ I made it to LessWrong &amp; the Alignment forum and drowned in it — not for nothing I’d been putting it off for quite a while. So, in order, what clicked, so to speak. <br> The whole post (<a href="https://www.lesswrong.com/posts/HBxe6wdjxK239zajf/what-failure-looks-like?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.lesswrong.com/posts/HBxe6wdjxK239zajf/what-failure-looks-like</a>) <br> is once again about the “end of the world”, but the idea that ML is what we can measure and learn to measure, and accordingly set goals, is pretty clear.</p> <p>2/ As a continuation I got carried away with generally trying to figure things out, and I liked <br> the explanation here of what the alignment problem even is ((<a href="https://www.lesswrong.com/posts/dHNKtQ3vTBxTfTPxu/what-is-the-alignment-problem?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.lesswrong.com/posts/dHNKtQ3vTBxTfTPxu/what-is-the-alignment-problem</a>) <br> I’m explicitly not translating it)</p> <p>3/ When people talk about search in AI it usually all goes <br> here (<a href="https://www.lesswrong.com/posts/6mysMAqvo9giHC4iX/what-s-general-purpose-search-and-why-might-we-expect-to-see?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.lesswrong.com/posts/6mysMAqvo9giHC4iX/what-s-general-purpose-search-and-why-might-we-expect-to-see</a>)</p> <p>4/ In my tabs there also snuck in <br> a Buffett article (<a href="https://www.berkshirehathaway.com/letters/growing.pdf" rel="noopener">https://www.berkshirehathaway.com/letters/growing.pdf</a>) <br> from 2003 about trade certificates as a solution to the US trade imbalance (doesn’t that remind you of anything?)</p> <p>5/ Of course right after that I started reading Scott Alexander (the guy knows how to hype) and I recommend <br> this one (<a href="https://slatestarcodex.com/2014/12/12/beware-the-man-of-one-study/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://slatestarcodex.com/2014/12/12/beware-the-man-of-one-study/</a>) <br> about how trusting studies blindly is a mistake</p> <p>6/ Canonically about who we don’t like and <br> why it’s like that (<a href="https://slatestarcodex.com/2014/09/30/i-can-tolerate-anything-except-the-outgroup/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://slatestarcodex.com/2014/09/30/i-can-tolerate-anything-except-the-outgroup/</a>) <br> — or observations about so-called outgroups</p> <p>7/ <br> I read this (<a href="https://www.lesswrong.com/posts/A2Qam9Bd9xpbb2wLQ/game-theory-as-a-dark-art?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.lesswrong.com/posts/A2Qam9Bd9xpbb2wLQ/game-theory-as-a-dark-art</a>) <br> back in uni because I was into game theory, and arguably some of the best examples are collected in that post, just for learning. For the especially interested, <br> this paper (<a href="https://core.ac.uk/download/pdf/6388215.pdf" rel="noopener">https://core.ac.uk/download/pdf/6388215.pdf</a>) <br> is a must-read.</p> <p>I can fit a lot of info in my head, but yeah, while I’ve been flying around for years, I think I overdid it with reading this stuff <br> 🙂 <br> Recommend something truly Must Have on Alignment — thanks! Just not about the end of the world :) Because almost every post there one way or another touches “scaling laws and how when we have clusters of a gazillion flops we’ll get models a hundred times cooler than the current ones and then AGI/ASI and all discoveries will arrive at once or we’ll get killed”</p> <p>P.S. core concepts: <br> Corrigibility (Корригируемость) <br> Corrigibility is a property of an AI where it’s willing to accept corrections or a stop signal from a human, even if that contradicts its original goals. <br> Example: <br> If the AI wants to build a bridge, and a human says: “Stop, it’s dangerous,” — a corrigible AI will comply and stop, even if it itself thinks building the bridge is better. <br> Why is it needed? <br> So the AI stays under control, even if its goals aren’t formulated perfectly.</p> <p>Mesa Optimizers <br> Mesa-optimizers are subsystems inside an AI that themselves begin optimizing their own objectives, possibly different from the original objectives set by external training.</p> <p>Inner Alignment Problem (Проблема внутреннего согласования) <br> This is the task of making sure the mesa-optimizer’s goals (internal) match the base optimizer’s goals (set from the outside). <br> In plain words: <br> When training an AI, we may implicitly create a “mind” inside it, and we need to guarantee that this “mind” will act in accordance with our original intentions.</p> <p>And spoiler: one of the founders of the alignment forum and a major LessWrong contributor is a mathematician and scientist from Moscow — Vladimir Nesov.</p> <p><a href="https://t.me/chillhousetech/1040?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Part 4. https://darkolorin.com/posts/1039.html https://darkolorin.com/posts/1039.html Wed, 09 Apr 2025 11:59:11 GMT <p>Part 4. The difference. <br> Besides all this noise with funds, I managed to talk to some old acquaintances—someone at Google, someone at Anthropic, whoever where, basically. And I got this weird feeling. In order:</p> <p>1/ hiring in the Valley is really damn hard, people can be making a million bucks a year total, and where public company stock is the bigger part. And even if the base salary is around 300k, how is a startup supposed to beat that total comp? Like, unless the person is bored and has already bought everything, or money isn’t a priority at all. But that’s probably a rare situation. The attitude is like: what are these startup paper slips to me, they’re worth nothing. So personal involvement matters, and over time it obviously drops. Pulling in strong guys with experience is very expensive.</p> <p>2/ the gap between research and dev/engineering people. The other day this article came out (<a href="https://ai-2027.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ai-2027.com/</a>) (I’ve got an unpopular opinion about it). Aside from the fact that I strongly recommend reading it (since the guy in 2021 predicted the next 5 years pretty damn accurately) for general understanding, the important part is something else. It’s basically a reflection of the researchers—how they think and what exactly the top AI labs are pitching. Like, a very similar narrative. Whether it’s true or not, we’ll see. But the engineering half thinks completely differently. There’s little faith there—not only that super-agents will be smarter than humans at everything, but even that writing a normal small product from scratch is almost impossible and it’s pain. The distance in how the future is assessed is huge. And I’m more inclined toward not such a big growth in AI capabilities in the next 5 years as the article paints. Maybe by 2035 it’ll work. And then right after that this article drops (<a href="https://www.lesswrong.com/posts/4mvphwx5pdsZLMmpY/recent-ai-model-progress-feels-mostly-like-bullshit?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.lesswrong.com/posts/4mvphwx5pdsZLMmpY/recent-ai-model-progress-feels-mostly-like-bullshit</a>) from the applied folks.</p> <p>3/ a feeling that it’s exactly there that the movement is, the thing that pushes the market forward. Hard to convey. But the density of communication and feedback there is way higher. And one of the reasons is that there are a lot of people in the companies that raise the biggest rounds and ship the most cutting-edge solutions. And the difference between, say, Europe and the conversations here vs there is felt literally.</p> <p>Do you believe in an all-powerful AI in 3 years?</p> <p><a href="https://t.me/chillhousetech/1039?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Part 3. https://darkolorin.com/posts/1038.html https://darkolorin.com/posts/1038.html Mon, 07 Apr 2025 17:53:08 GMT <p>Part 3. On how to pitch, or what advice people with experience raising $100m+ rounds in the Valley gave</p> <p>In short: you need cash, they don’t need anything that won’t potentially become the next Google. The fund doesn’t give a shit about 1–3x returns. A fund partner personally needs, over their whole career, to make one correct lead investment into a company-that-is-a-market (IPO/decacorn). After that they’ll walk around on pure swagger. That’s the reality and you have to accept it. So emotions run the show, not logic (because no logic is capable of predicting the future).</p> <p>Best way to trigger emotions — tell your story. Your job is to create a sense of the moment and belief that what you’re doing is the next big thing.</p> <p>Two key principles:<br> 1/ the size of the success matters more than the probability of success. Like I said above, everyone wants one Google in the portfolio, not 10 exits at $100m.<br> 2/ emotions over facts. Numbers almost never sell, what sells is “fear of missing the deal”.</p> <p>The only question an investor should be asking in your pitch (ideally) is the investment terms, how much money and why. Everything else is just water.</p> <p>Investors very rarely actually understand what you do, and even more so the tech. If they need to, they’ll bring in a consultant and drag you around, but let’s assume you can clear that easily and you’re doing something reasonable. Then what’s left as the dry residue is: want/don’t want at the level of the emotion from the meeting.</p> <p>How to approach the process:<br> 1/ the deck should literally be changing constantly. You got a question you didn’t expect — add the answer to the deck, remove what doesn’t land in meetings. Throw out all templates. The story and emotions matter — build the math around that.<br> 2/ make a list of funds with the probability you can actually talk to them and your belief that they might like it or not, and who is ideal for you (study the market, deals, write to competitor companies). Don’t talk to people who aren’t in the topic and don’t understand your product and team at all. The cultural aspect matters (see who has already invested in Eastern European founders, they’ll trust you more).<br> 3/ pack all meetings into a defined time window, best 2 weeks, maybe a couple months if the round is big. This matters to create momentum and the feeling the deal can slip away.</p> <p>What’s important to understand: an investor’s job is very hard. Respect for their work matters. If you can sell to them, that’s a strong signal to the market that you’ll be able to sell to companies and people too. Because they have flow and drill this thought into your head: their job is harder than mine.</p> <p>And I tell stories badly, so I don’t know what to advise you.</p> <p><a href="https://t.me/chillhousetech/1038?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Part 2 is short. https://darkolorin.com/posts/1037.html https://darkolorin.com/posts/1037.html Sun, 06 Apr 2025 09:45:20 GMT <p>Part 2 is short. <br> No point talking to top-tier funds, even if they’re talking to you about investing. <br> Any top fund == &gt;$5M per check.</p> <p><a href="https://t.me/chillhousetech/1037?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Back from the battle of the Valley. https://darkolorin.com/posts/1036.html https://darkolorin.com/posts/1036.html Sat, 05 Apr 2025 19:26:46 GMT <img src="https://cdn4.telesco.pe/file/QlZj0V3v9qhhNpA9KryHep39XOoKNGr43b0n_h5-a5hESDFqMlTL8-uWHHBDVXrltHzxlo-ED-Rh3XzGIjfHj0_kmXljw9sHba71ahJ9E2fuft_-5VBMudVjAvYqi1mxjXs7-uHm32PWwEGNUvatsa76KuBNmTrxwAalgipvdW57sC1HMzWKpBWhpw_KKCvbjGZVBXlOH_AGhK5lN1P6uzSxeUMpVBayHPYhqoBC9ikJYkzrIdbBHFomGLYJ_FiBBMwfsYoKLYDzA9pyXG43pHISVzbgvtFuRghAu9aYVmOkZaZE2CiIzhrpcnioNtixWpQOD54_CPp3q5Ml-MMxyQ.jpg" alt="" loading="lazy"> <p>Back from the <br> battle<br> of the Valley. <br> A series of posts. Part 1. What the hell it’s even about there and why it’s about there. <br> The Valley got resurrected with the arrival of AI. Like, a huge amount of everything, including posters on Highway 101, is ads for AI companies; maybe 1 out of 10 billboards isn’t about AI. <br> With what they pitch on every corner in a coffee shop, and the fact you can genuinely run into a Sequoia Capital partner in a coffee shop, I guess you won’t surprise anyone anymore. Let’s get to the point. <br> About funds. An interesting effect has cropped up. Pretty much all previous software waves were evaluated by investors using old yardsticks and, in general, it worked, and basically the same top funds (say, a16z условный) stayed top. In my feel, the situation has changed. Partners in top funds are stuck in late-2010s frameworks and don’t understand what’s going on and where to invest. So they just wait for it to hype. Based on conversations, my top funds list goes like this, in this order:<br> 1. <br> <a href="https://www.conviction.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.conviction.com/</a><br> 2. <br> <a href="https://accel.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://accel.com/</a><br> 3. Everyone else, though for fairness’ sake I didn’t manage to have a normal talk with Sequoia. But a16z and the rest of that crowd didn’t make me happy at all (a very weak understanding of what AI is, who needs it, and what’s happening there, partially patched by the fact their portfolio has solid late-stage companies and all the lessons and conclusions come from there). <br> Quote from one partner: money isn’t the problem. The problem might be in:<br> 1/ a team without industry experience (for example you don’t have researchers or ML engineers with work experience from before the AI hype)<br> 2/ you don’t know what’s going on in the market (and investors do and won’t share haha)<br> 3/ commercial or product validation (here everything’s easy and as usual and this is a criterion from the past)<br> A lot of technical partners in a lot of funds, and pretty much everywhere some условный fund CTO has appeared. Because there’s gotten to be a ton of bullshit. <br> A photo from Yosemite park for attention.</p> <p><a href="https://t.me/chillhousetech/1036?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> However, OpenAI are launching a model with open weights. https://darkolorin.com/posts/1035.html https://darkolorin.com/posts/1035.html Mon, 31 Mar 2025 20:24:47 GMT <p>However, OpenAI are launching a model with open weights. (<a href="https://openai.com/open-model-feedback/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/open-model-feedback/</a>) <br> That’s super interesting.</p> <p><a href="https://t.me/chillhousetech/1035?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Just dropped by A16Z today, and like, every time they try to sell you this… https://darkolorin.com/posts/1034.html https://darkolorin.com/posts/1034.html Sat, 29 Mar 2025 01:23:25 GMT <img src="https://cdn4.telesco.pe/file/cNWpjQmvGbjIhpDPUie8IwEKn7XKgzmrY4lbBcOpYC1cQOZHOOWfBMmzvDgCFYM7SYE8HL_ROAf4mIjVEUDapRnigTv7YT8-l0QUo7h0Mfrgf7ZBKC638_upc0m1NOeLhISzA6KPynmlKnxvUeVYGfj3x-r-WWQPxr7dhHhsYjC0eTFdpl-_v-sVvQWOjh1UW-4RP2vA3U3Y9lH88fDhjTX79WwMSeEVQsnpaKsIj0ZPJrvYtiH0EI8yldHuP8l3S38O7k5hLHeD5jT_WnO8kHGKEzbYOX64mVED4oJU3dQdgBn9hPiLSUz6TFO3jmZ1H4MzWJ_DIa7Mm0WeR3FJhA.jpg" alt="" loading="lazy"> <p>Just dropped by A16Z today, and like, every time they try to sell you this story that Cursor spent a whole year with literally zero profit and then bam, it popped off.</p> <p>So here’s the thing: AI tools and all that kind of stuff have this wonderful property — “what if I just try it.” Which basically means the same thing as in b2c when people come into an app just to peek, see what’s what (trust me, I’ve seen this effect with my own eyes at a reaaaaally big scale — Prisma). In that scenario the app obviously has super low retention. So yeah, most likely it’s the same with revenue retention for AI tools. And if, sure, you can now sprint to it with a 5-person team, keeping it is a separate question. Same shit happened back in the boom of viral filter apps.</p> <p>Attaching the chart that every investor posts on their twi or wherever.</p> <p><a href="https://t.me/chillhousetech/1034?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> While I’m running around here, I managed to run the classic image generation… https://darkolorin.com/posts/1033.html https://darkolorin.com/posts/1033.html Thu, 27 Mar 2025 15:03:57 GMT <img src="https://cdn4.telesco.pe/file/BtG3phuxaNjtwKCmXimxOzBFeMfmm5M8lmIqTES37Ki5HOfvgJIEr_--dAbRlOknGU3PhwonRVPA6uUoad8vQY3883EW9psBwGk9YTsquToXeuJCMsL664E9_sS2errlOwT7-YZnA3991IR0p48XjJip5FIAZV24P44d8KP9ouNFIrlNn9Ogf_tZauE2dbpAZp7f_WCLTletVNkES2XAlPrXU6OXKLZSLw_HSMDBN4p0CZRQxIN74_Wp-aHdjdctjUJ4Lz1yY4TqqtLFiet6WN8HPCofsBtaj8sXGTaqgTJlGYchAvmjGIs23cOYjAwnd5I-_ZYjp2z6aHLKwmOg6Q.jpg" alt="" loading="lazy"> <p>While I’m running around here, I managed to run the classic image generation test.</p> <p><a href="https://t.me/chillhousetech/1033?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Friends, the next couple weeks in the Valley. https://darkolorin.com/posts/1032.html https://darkolorin.com/posts/1032.html Sun, 23 Mar 2025 16:29:33 GMT <img src="https://cdn4.telesco.pe/file/pedNN-yauOxzYBPO4fEFGwEnz2ggxKPiSl_XkiLqb5GJVnDXtGC-MokE8N0Z_nBVChdIHxjnBjVZHhQb3iNkSuU_w9BpOv4u3cPSUffY9VG_9-QH8GTtlcELx773X-Xql4H4uDlFy7DGqYQt_IzEfaZ4fKYK9ZdxGAG2QR6bRk8_Tu2SWHRT_BFxOVB_yUSNrS8UF3_zG-AkhzaOE33IssjL96E6xZg3BT2ltRU5QBlduixSHVCUGFPMQICVBWZF9XRGUhDvoWHiZn74ra7JJDQoFtbjMGJoYJmU853Ge6bfNzCe-qmLFv2LaHdAev0RGz3nNGLt6yu4CBYzCjJfEg.jpg" alt="" loading="lazy"> <p>Friends, the next couple weeks in the Valley. Should I write you, so to speak, a live report? There’ll be a lot of meetings, negotiations, and so on. As usual, lots of AI.</p> <p><a href="https://t.me/chillhousetech/1032?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> This post isn’t an ad. https://darkolorin.com/posts/1031.html https://darkolorin.com/posts/1031.html Fri, 21 Mar 2025 03:25:07 GMT <p>This post isn’t an ad. Which Telegram channels I read from time to time and why. <br> 1/ <br> about models, AI and all that (<a href="https://t.me/lovedeathtransformers?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/lovedeathtransformers</a>) <br> with a reasonable dose of skepticism and a ton of memes. So it’s not stressful. <br> 2/ <br> I know Boris personally, (<a href="https://t.me/boris_again?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/boris_again</a>) <br> respect him. Plus his career progress is legit, already staff ML at eBay. Also about ML. <br> 3/ Sometimes you need various reports, research and writeups from big consultants and other market research stuff. <br> Zero comments, files only. (<a href="https://t.me/businessincognita?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/businessincognita</a>) <br> 4/ <br> Lenya Cherny, (<a href="https://t.me/CDO_Thoughts?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/CDO_Thoughts</a>) <br> way back in the day we started together at Yandex. Lenya leads our efforts in Turkey, I’m 23, I’m a project in Maps. He just writes about management from the perspective of a very experienced exec. Useful to read in your free time. <br> 5/ I’ve also known Vlad for a long time, he only <br> just started the channel. (<a href="https://t.me/neurocry?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/neurocry</a>) <br> AI again, but for now it’s more interesting facts and stories. So just support the effort :) <br> 6/ Respect to Artem and <br> his ai newz. (<a href="https://t.me/ai_newz?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/ai_newz</a>) <br> Probably the most understandable-to-me channel about news from the world of AI models, dry, clear, on time. <br> 7/ unfortunately, in the rest, the ones I liked, no posts have come out for like a month already. Even though the guys there run things solid. Maybe I’ll mention them in the comments. <br> And what do you read? Toss links in the comments, I think it’ll be useful for everyone. The best ones in my opinion I’ll add to the post.</p> <p><a href="https://t.me/chillhousetech/1031?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The packed-calendar theory, or why Elon Musk gets sooo much done — even games… https://darkolorin.com/posts/1030.html https://darkolorin.com/posts/1030.html Wed, 19 Mar 2025 16:14:06 GMT <p>The packed-calendar theory, or why Elon Musk gets sooo much done — even games at a decent level too. <br> There’s this classic kind of admiration for people’s productivity. Like how the hell do you manage to give a lecture over there, and post posts, and also work, and also go somewhere for meetings. I’ve got my own two cents of “how did it end up like this / or what’s wrong here.” <br> I’ll break it into random points:</p> <p>1/ A crammed calendar or a million things doesn’t let you exhale, or even think, process what’s going on. Personally, for two effective hours of truly important decisions or emails or a plan, I spend 8–10 hours “thinking,” which most often looks like watching a stream on Twitch, sitting in my garden, or taking a walk, or even cleaning. So playing games is part of the thinking process or brain rest, when you switch off completely.</p> <p>2/ The more stuff, the less important it is. In short, when you’ve got 100500 tasks in a day, it’s all small fry that doesn’t require attention and either gets fully automated or just gets ditched to hell.</p> <p>3/ Run from a manager whose schedule is booked 24/7 (if your manager isn’t top management or a CEO). They don’t have time to look around and make the right decisions about the team, for sure. And if they also brag about it — that’s just a disaster. A packed calendar is a problem. And often it’s defocus and inability to set priorities.</p> <p>4/ I simply believe in 3–4 hours of effective work a day if you’re over 30.</p> <p>5/ Everything booked == inability to delegate, and that’s an important skill nowadays and always has been.</p> <p>6/ Everything above is about intellectual work. A simple definition: when you’re not working with physical force and when in your profession there are people who are dozens of times more effective than others (Steve Jobs used to say this about developers and why it’s so important to hire the best because in this field there’s a chasm in efficiency between different professionals).</p> <p>So, is your calendar packed?</p> <p><a href="https://t.me/chillhousetech/1030?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Since we’ve got a lot of new people here. https://darkolorin.com/posts/1029.html https://darkolorin.com/posts/1029.html Wed, 19 Mar 2025 01:58:59 GMT <p>Since we’ve got a lot of new people here. I’ll ask the classic question. How often should I post so it’s comfortable for you?</p> <ul> <li>Once a day is fine</li> <li>A few times a day is fine too</li> <li>Once every 2–3 days</li> <li>Once a week or less</li> <li>Never 👍</li> <li>😘</li> </ul> <p><a href="https://t.me/chillhousetech/1029?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> As promised, we’re open-sourcing the bot — fork it to your heart’s content… https://darkolorin.com/posts/1028.html https://darkolorin.com/posts/1028.html Tue, 18 Mar 2025 12:27:19 GMT <p>As promised, we’re open-sourcing the bot — fork it to your heart’s content, run it in your own closed environment, don’t worry, the links won’t leak <br> <a href="https://github.com/matterai/DataKitsune?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/matterai/DataKitsune</a> <br> A couple small improvements: <br> - made basically a 1-to-1 mode: tossed in a link and it spits out a summary; like I threw in some YouTube vids and it gave me what’s what <br> - generally speaking the video mode is super handy for all sorts of stuff because it outputs everything by timecodes (need to quickly find some moment — with the `search` command you’ll be able to search through the vids you threw in, hell yeah, great stuff)</p> <p><a href="https://t.me/chillhousetech/1028?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> My favorite “send it to an LLM” test. https://darkolorin.com/posts/1027.html https://darkolorin.com/posts/1027.html Mon, 17 Mar 2025 15:31:55 GMT <img src="https://cdn4.telesco.pe/file/fvnTkWZjOxmDHBLGL-IGCDANgfq70spwdHS2zSxW6jDr-47TuBuzaGHGv9FGal_tGv48LMt0WMagP5P7HkdswYm-LkXphJZ_ugrP9TS8oj8MfjPhLJMrXjB-m65ZRbmsVR6o85HRoxr_n5677tKMItMD6U4_HJR2PAgtFQJ222ljD3MjHYKUUaVdq--J4ef-yOseighfwLeKLVCzkn3NnAGaV5RHJ5idB92i26Ry02jUKlD88FQWURPcviLDwoEjd1cVOk16ieEEYEjX8OFAOvw0OiMqOd2xdFvwd8_RuudXlX89zp5OrqMdoK9Iw5XfaWsR2rm7l_PHcO2EKW7kNg.jpg" alt="" loading="lazy"> <p>My favorite “send it to an LLM” test. Drop your answers in the comments. You need to pick the figure that “stands out.” <br> In the first comment there’ll be the nets’ answers. What’s yours?</p> <p><a href="https://t.me/chillhousetech/1027?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Damn, I want a feature like that https://darkolorin.com/posts/1026.html https://darkolorin.com/posts/1026.html Sun, 16 Mar 2025 19:14:06 GMT <img src="https://cdn4.telesco.pe/file/Y9pAWsejUlrsY8L66jABK12ih-tS2R7aPF3WN-_9lcWOwOfxXi1i35NSQICs6Nc8ER_kUv0w9fzIZinqqIhNmGoG66HYQhJW2PNCsqI-8tjzQV_Fqxb-_dxINW7_bAytA7OWL3fY7mz0Ubi-joAJI6O8KtH5sS8q93vjvwokCmhgzRl5FrmdvAqxd2XQRI6LFZJQ69sg86n8JkOT9uG8UCZhXmsC39Blv_GcYZf530v8nxk7jhlwz2qK9o-VXI60EFArtZ3TArKbKT8W4eT3Mw5snAB37vuFLMYqNYP6_d_JIrs3njhAE2L0eIqMD6EA6_IIbBGdHE6Wi4U_LFZQnw.jpg" alt="" loading="lazy"> <p>Damn, I want a feature like that</p> <p><a href="https://t.me/chillhousetech/1026?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Dear friends! https://darkolorin.com/posts/1025.html https://darkolorin.com/posts/1025.html Thu, 13 Mar 2025 12:10:33 GMT <p>Dear friends! We’ve got a small bot inside the team, we built it for internal purposes. And then we decided to also throw it out to the public and for free (hope we don’t burn through all our credits).</p> <p>Why we made it. In our team people constantly drop all kinds of links into chats, like a lot, you know, to the point where I just won’t manage to read all those articles, papers, YouTube vids. So we decided to make a tool so we don’t miss stuff. For starters we’re doing it for Telegram and its little chat rooms.</p> <p>What’s the point:<br> - the bot in a chat can collect all links and send you a summary to your DMs at a time you set<br> - the bot doesn’t need admin rights<br> - the bot can summarize videos (thanks Gemini API) from YouTube<br> - hopefully it holds up under load :)<br> -<br> killer feature: someone dropped a link in the chat about something a couple months ago, and you only remember the keywords — feed those to the bot in the search command and it finds it even if you don’t remember the exact title!<br> - the bot itself<br> <a href="https://t.me/datakitsune_bot?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/datakitsune_bot</a></p> <p>Pretty video and a landing page included (<a href="https://datakitsune.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://datakitsune.com/</a>)</p> <p>I’ll be very happy if you try it (everything’s free), share it with friends who might need it.</p> <p>Also taking feedback for improvements if you actually use stuff like this. If you need to deploy it locally (on premise) — DM me or leave a comment in the chat.</p> <p><a href="https://t.me/chillhousetech/1025?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ve even heard professors warn that reading the wrong book can “cause brain… https://darkolorin.com/posts/1024.html https://darkolorin.com/posts/1024.html Wed, 12 Mar 2025 20:46:38 GMT <p>I’ve even heard professors warn that reading the wrong book can “cause brain damage” (c)</p> <p>I’ve been catching myself pretty often starting to notice that the motivations of AI research teams have started to look less like science and more like shipping a new feature that just grabbed one more benchmark. And all of this gets sold by some “peppy” CEO yelling that we’ll all see AGI soon. Feels like something’s off here, and everything’s turned into fashion. In a lot of top labs, in my subjective feel, the spirit of science has started to evaporate. And then I stumbled on an interesting note (<a href="https://sugaku.net/content/understanding-the-cultural-divide-between-mathematics-and-ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://sugaku.net/content/understanding-the-cultural-divide-between-mathematics-and-ai/</a>) by a mathematician who visited a congress (a meetup) organized specifically for mathematicians; but to me the precedent itself matters — AI really is heavily affecting pretty much all areas of knowledge and science. Fun times!</p> <p>A bunch of things there that are mentally and “model-wise” close to me and reflect the essence of my work:<br> - Terence Tao (one of the strongest mathematicians of our time, definitely read about him) wrote a note back in 2007, “Be Patient,” pointing out the fact that it often takes years from the first day you start working on a problem to its solution. Roughly the same way I see it in truly hard products and decisions I’ve arrived at.<br> - In math, like in physics, discoveries are often named after the people who worked on them. That underlines the importance of the human factor — same as I think about the importance of a team of people in a startup. A lot of folks underestimate this, but finding genuinely awesome people is extremely hard, and that’s a huge part of success. And the most interesting thing is that in math papers, the team is considered equal: in 75% of cases the last names in the title go in alphabetical order, while in other fields (AI) they go in order of significance of contribution.<br> - Mathematicians often search for the essence of things, whereas AI is just trying to invent something new or impress. Roughly the same way I think about everything — since childhood I was into physics because I wanted to understand how things work, the universe and all that, because it’s insanely cool, like a new world opens up: when an equation isn’t just a bunch of letters and equals signs, it’s what’s all around you.<br> - And a good question landed at the end: even if a new AI system can prove the Riemann hypothesis, what if the proof is so complex that a human just can’t understand it. Would there be any point in such a proof? What if nobody understands fundamental discoveries at all anymore? Because that hasn’t happened in human history yet. And will we be able to actually use such discoveries for real?</p> <p><a href="https://t.me/chillhousetech/1024?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Alright, enough with the ololo posts. https://darkolorin.com/posts/1023.html https://darkolorin.com/posts/1023.html Wed, 12 Mar 2025 02:16:31 GMT <p>Alright, enough with the ololo posts. Back to the “read it, but didn’t fully get it” section. Today there’ll be a lot of random stuff.</p> <p>1/ A very solid article called <br> “Domain specific architectures for AI inference” (<a href="https://fleetwood.dev/posts/domain-specific-architectures?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://fleetwood.dev/posts/domain-specific-architectures</a>) <br> for those who are into running and optimizing. What’s more, the article has a bunch of great links.</p> <p>2/ <br> A genuinely good guide (<a href="https://www.bvp.com/atlas/the-founders-playbook-for-scaling-to-1-million-arr?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bvp.com/atlas/the-founders-playbook-for-scaling-to-1-million-arr</a>) <br> article on what paths there are to make it to $1M ARR (unlike my advice above) from Bessemer Ventures.</p> <p>3/ Quantum chips from Microsoft — cancel for now. Way too revolutionary. <br> Here’s a little overview. (<a href="https://quantumfrontiers.com/2025/03/05/what-does-it-mean-to-create-a-topological-qubit/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://quantumfrontiers.com/2025/03/05/what-does-it-mean-to-create-a-topological-qubit/</a>)</p> <p>4/ <br> Found a pretty interesting prediction (<a href="https://sideways-view.com/2023/07/29/self-driving-car-bets/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://sideways-view.com/2023/07/29/self-driving-car-bets/</a>) <br> about self-driving cars from a dude who worked on an AI alignment team together with Ilya Sutskever and, say, Dario Amodei. And it’s interesting because the situation per the prediction is similar to the whole “in 10 years everything will be!” vibe. But we know how it went with self-driving cars: nope, nope, nothing after ten years :) except a couple cities, and only a little bit.</p> <p>5/ <br> a funny bench, (<a href="https://metr.org/blog/2024-11-22-evaluating-r-d-capabilities-of-llms/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://metr.org/blog/2024-11-22-evaluating-r-d-capabilities-of-llms/</a>) <br> where they try to compare model performance on tasks a human expert also does. But the most interesting part is that a bunch of models got caught cheating in the literal sense of the word (for example, in a similar chess bench the model was brute-forcing like thousands of positions to see which one fits). And overall there’s a <br> cool thread (<a href="https://x.com/colin_fraser/status/1892721051299172641?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/colin_fraser/status/1892721051299172641</a>) <br> about something slightly different; the original post got deleted btw.</p> <p>6/ <br> this is the best spec (<a href="https://spinroot.com/gerard/pdf/P10.pdf" rel="noopener">https://spinroot.com/gerard/pdf/P10.pdf</a>) <br> for your coding agent</p> <p>7/ if you, like me, sometimes miss theoretical physics, there’s a good <br> refresher (<a href="https://ru.wikipedia.org/wiki/%D0%A2%D0%B5%D0%BE%D1%80%D0%B8%D1%8F_%D0%BF%D1%80%D0%B5%D0%B4%D1%81%D1%82%D0%B0%D0%B2%D0%BB%D0%B5%D0%BD%D0%B8%D0%B9_%D0%B3%D1%80%D1%83%D0%BF%D0%BF%D1%8B_%D0%9B%D0%BE%D1%80%D0%B5%D0%BD%D1%86%D0%B0?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ru.wikipedia.org/wiki/%D0%A2%D0%B5%D0%BE%D1%80%D0%B8%D1%8F_%D0%BF%D1%80%D0%B5%D0%B4%D1%81%D1%82%D0%B0%D0%B2%D0%BB%D0%B5%D0%BD%D0%B8%D0%B9_%D0%B3%D1%80%D1%83%D0%BF%D0%BF%D1%8B_%D0%9B%D0%BE%D1%80%D0%B5%D0%BD%D1%86%D0%B0</a>)</p> <p><a href="https://t.me/chillhousetech/1023?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For anyone who wanted to instantly show metrics to investors. https://darkolorin.com/posts/1022.html https://darkolorin.com/posts/1022.html Wed, 12 Mar 2025 01:59:26 GMT <p>For anyone who wanted to instantly show metrics to investors. Here’s the plan:<br> - raise a pre-seed from a bank/friends<br> - burn 250k on performance ads<br> - make 83k in revenue in the first month, maybe even a day!<br> - and boom, “the fastest 1M ARR”</p> <p><a href="https://t.me/chillhousetech/1022?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So yeah, my barter post. https://darkolorin.com/posts/1021.html https://darkolorin.com/posts/1021.html Tue, 11 Mar 2025 19:56:22 GMT <p>So yeah, my barter post. My friends who went through YC are launching their product on Product Hunt. They’ve also got some super active LinkedIn thing going on — insane reach numbers (no idea about effectiveness).<br> <a href="https://www.producthunt.com/posts/fluently-ai-english-coach?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.producthunt.com/posts/fluently-ai-english-coach</a></p> <p>As a thank-you for the support, everyone who upvotes today (March 11) — they’ll send you their guide on fundraising and filling out an application to the top US accelerator — Y Combinator. The guide includes their YC experience and raising money at the pre-seed / seed stages (they raised around 2).</p> <p>Once you’ve upvoted —<br> fill out the form (<a href="https://docs.google.com/forms/d/e/1FAIpQLSc67a0oY7uvRLHNTYrzaYHeMTMhb-t1ZRZB2XiIzh658pztdQ/viewform?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://docs.google.com/forms/d/e/1FAIpQLSc67a0oY7uvRLHNTYrzaYHeMTMhb-t1ZRZB2XiIzh658pztdQ/viewform</a>)<br> and they’ll send you the guide.</p> <p>P.S. In return I’ll get a lil post about my mini launch in the next few days. Stay tuned.</p> <p><a href="https://t.me/chillhousetech/1021?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> While I’m writing a post off the back of digging through all sorts of AI… https://darkolorin.com/posts/1020.html https://darkolorin.com/posts/1020.html Mon, 10 Mar 2025 17:52:32 GMT <img src="https://cdn4.telesco.pe/file/CdGLix9PWOSjMc667FiPy25DLTkWrQdzLtics14KIkjPaY56NXxxq85xl8-Re_OSIgX2g9jAtyfF5uXsycqvmcjQAL5kFHlxI5s2Y9y2NLGMdrDjbbiCiL588wpFs8kg0h_FEhVtqg_PCJxl_d4NVYpVw5NecJ0VT4LLqvVYndnb7ZrBWab6wIxKEmZGmVzGASlXr-Op75EhsSdjFsc0VzMHvSW6gchyNUyMv11zpK9NYEVIGSO-16sfYXVQR9Y8u3JsYIDlTCga3n1gFgoN14BAoE4Ee7rB_vnbIuEp7rtZDFwVy7RFnUiGnaKnBCFdufg0EdKXR6nmeSN_6RquPQ.jpg" alt="" loading="lazy"> <p>While I’m writing a post off the back of digging through all sorts of AI alignment topics. And there’s so much interesting stuff in there, not just “they’ll enslave us and that’s it.” More and more I keep thinking like this.</p> <p><a href="https://t.me/chillhousetech/1020?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Went to a dinner of founders backed by OpenAI. https://darkolorin.com/posts/1019.html https://darkolorin.com/posts/1019.html Thu, 06 Mar 2025 08:02:06 GMT <img src="https://cdn4.telesco.pe/file/LgD0Hji2F7K4mBdpixkrb4kQNykPAN2VRgcDSmlT_KL9AQio6ghoGDPjYoJ07x8C4oCam95ZSFzlzaYYiBfVI1Xk7H7NhA07OIJalzVq8VGcutMe3nCXV8znRZbNRi7oXaCWQpCsds62lA_3-cKHFtI7tjgDPcj6fpaciGjA63zE2ZxyIxkpWFo_8u_chVAf7kgJufrJmzD4_rHVJ_fs7CeE3Y7NZbJlAW1TT-dPmm8DSv2vILuyJ2kHLzcqWYDFkzPu1uvgNnBysI29DnxGTVv1-xuq1gMHnJ6iiF1M-GZncP2IikOhpSOf1w8LZ5Kf5q04XifYaegqfinjTuQeTg.jpg" alt="" loading="lazy"> <p>Went to a dinner of founders backed by OpenAI. Well, everyone’s just selling wrappers :) a good strategy for OpenAI to have your API sold by startups that also go and raise on their own</p> <p><a href="https://t.me/chillhousetech/1019?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well yeah, and why not 200k a month while you’re at it? https://darkolorin.com/posts/1018.html https://darkolorin.com/posts/1018.html Wed, 05 Mar 2025 18:12:55 GMT <p>Well yeah, (<a href="https://www.theinformation.com/articles/openai-plots-charging-20-000-a-month-for-phd-level-agents?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theinformation.com/articles/openai-plots-charging-20-000-a-month-for-phd-level-agents</a>)<br> and why not 200k a month while you’re at it? Hm?<br> Looks like they’ve got this weird love for numbers that are multiples of 20.</p> <p><a href="https://t.me/chillhousetech/1018?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A bit about the future of SaaS, personal assistants inside the system, and AI… https://darkolorin.com/posts/1017.html https://darkolorin.com/posts/1017.html Tue, 04 Mar 2025 17:36:06 GMT <p>A bit about the future of SaaS, personal assistants inside the system, and AI for this whole thing. <br> Claude Code agent dropped recently and Cursor started hyping hard last year. These are basically as close as it gets to automating the system (you can call them agents), the kind of stuff that helps you write code or, essentially, be a programmer of features. Last summer I was already pretty easily making Telegram games as an experiment without touching a single line of code with my hands. Back then people didn’t really believe it. Then the quality of the base coding models started going up and it became possible to build pretty complex prototypes.</p> <p>What I’m saying is: right now it’s totally realistic to make yourself a web app or a simple game with zero programming knowledge and be happy. Yeah, it’ll be a bit sweaty along the way, but it’ll work out.</p> <p>So I think that at the current pace of progress in automating the programming process, you’ll easily be able to write yourself any basic interface for whatever. Like, you need to learn English tomorrow using some method you saw on TikTok, you open an assistant (soon even a system-level one) and go: hey, make me a learning app with flashcards and also so I can instantly search for movie clips where this word shows up. And the assistant will put that app together for you. Same thing with all sorts of SaaS for small businesses. You’ll be able to just say: hey, make some A/B tests on the buy buttons and show the results and maybe even pick what, according to the data, actually works on my audience.</p> <p>That’s where I see a big bet on Anthropic and Cursor. Automating engineering effort, for now, looks like the most real thing.</p> <p><a href="https://t.me/chillhousetech/1017?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Even if the workout costs a trillion bucks https://darkolorin.com/posts/1016.html https://darkolorin.com/posts/1016.html Fri, 28 Feb 2025 17:39:58 GMT <img src="https://cdn4.telesco.pe/file/p2YEAx4NQOelymlgKIPCtIO4zDY-0DcX7WS8-JhP1MIH6KfmuTz5pDyLQxC9XH1SIj0PqlKky8ZO0v-0Bq73WSoE6WenS9VHSZyMKejXIVkwVya4Pe6l0OYw71ek72IWmsdGbbnfhBZUDgqtGxr0ZvYKZt8xqH5MaSbnncXfCjdqiXlFSR_V2_buB4C1HZrWnSGbYerwiHlQKWveb8y5zWHJzJlMels6wJ4ArY-WmKGnGmc5pAy2rrISb_Ny2xPgtPzKde_gCo7Bj7sVyqzjQdpxd7YC9hSatXVPAaJxSDOjyUALSue0TAG1VXQP5dU0j8o2dEA1jvjAOOrsJjF9sQ.jpg" alt="" loading="lazy"> <p>Even if the workout costs a trillion bucks</p> <p><a href="https://t.me/chillhousetech/1016?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Yesterday GPT 4.5 dropped. https://darkolorin.com/posts/1015.html https://darkolorin.com/posts/1015.html Fri, 28 Feb 2025 16:11:26 GMT <p>Yesterday GPT 4.5 dropped. I’ll throw in my two cents—my own thoughts or not my own (it’s hard to be original here).</p> <p>Basically, they designed the current approach to AGI and it’ll never get there. So what do we have. The new version is, in fact, 10% better, but it looks like it cost about a yard of bucks (1B USD!) to train, judging by API pricing. We’re seeing logarithmic scaling: every 10x in spend gives only ~10–15% improvement. A GPT-7 that would surpass GPT-4 the same way GPT-4 surpassed GPT-2 would cost $100 trillion—the entire world GDP.</p> <p>Sure, there are all sorts of ideas about test-time compute, like “see, here’s how we can make it smarter,” but in that case for every new token it would have to think for days. Which looks, at the very least, unusable in real life for real tasks (easier to just hire a human). And the cost is insane.</p> <p>And what am I getting at. At the fact that we’ve actually felt out a wall in the progress of “Intelligence” of big models and the transformer’s capabilities. Now it’s left to figure out what architecture handles prediction tasks and so on better than “attention is all you need.”</p> <p>What do you think?</p> <p><a href="https://t.me/chillhousetech/1015?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Remember Skype? https://darkolorin.com/posts/1014.html https://darkolorin.com/posts/1014.html Fri, 28 Feb 2025 15:52:19 GMT <p>Remember Skype? Yeah, that’s it, bye-bye. An era is leaving.</p> <p>News (<a href="https://www.microsoft.com/en-us/microsoft-365/blog/2025/02/28/the-next-chapter-moving-from-skype-to-microsoft-teams/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.microsoft.com/en-us/microsoft-365/blog/2025/02/28/the-next-chapter-moving-from-skype-to-microsoft-teams/</a>)</p> <p><a href="https://t.me/chillhousetech/1014?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In short. https://darkolorin.com/posts/1013.html https://darkolorin.com/posts/1013.html Thu, 27 Feb 2025 16:54:34 GMT <p>In short. Signs of content you should look at critically and ask yourself three times if it’s bullshit:<br> - constant appeals to authority (references to famous people like CEOs of big companies, nonstop last names of some top managers)<br> - zero substance, no numbers, no analysis, nothing — ideas get mentioned, deals, whatever, but details are zero, explanation of meaning is zero (example: “I had a genius idea but others did it”), no deal participants, no company names, investors are always just “top funds”<br> - they’re constantly trying to “upsell” you on the author as a super specialist, meaning the author is proving he’s cool, knows, understands, instead of writing concrete points or lists<br> - nobody knows him except the author himself, there are constant references to “the community” and all that, but the moment you ask “ok introduce me” it all turns into a sales pitch: “work with me”<br> - you feel like they’re about to start selling you paid consulting and yep, that means they will start selling it<br> - the takeaways are too “revolutionary,” like they’ll solve all your problems instantly, and you’re constantly being asked “and wouldn’t you like to change your life/career/salary?”<br> And you — got any signs like that? What usually bugs you?</p> <p><a href="https://t.me/chillhousetech/1013?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> 8 cents for a file listing in my folder where I was collecting various test… https://darkolorin.com/posts/1012.html https://darkolorin.com/posts/1012.html Wed, 26 Feb 2025 00:37:09 GMT <img src="https://cdn4.telesco.pe/file/D9HRknwXkrOs4tN7R3rEodXISjoahYJuJN-4mHMmd2MECP5X2mTh4P__DrZzexlu5BJQLUbpXTD4IFwNzeC38H5wyId2XTP75vcpRQkfLby6_JtlmpoLB3agMIjKZ2GPvzlOvawWpPTqL_Ro5CUdsLlYilAn1Qk34BIAzdaUBWtpk_J2Pfw7Jszr5wfcCNprHXy_orEaKFGUQcbbypzrLZsDxADpNObs7fRH1bUvnfjK2ygowDP4Qwyd0EmCzTpch_bMII6QVd90iKurA9QycLcoKd9Jd_8gYPgEA8y3mhXFgh8wdu_f2g9t6h3a6Pf5nrbB57LjEjz4lWbkCtIZCg.jpg" alt="" loading="lazy"> <p>8 cents for a file listing in my folder where I was collecting various test projects for laughs using Claude Code. Got it, message received: ls -la</p> <p><a href="https://t.me/chillhousetech/1012?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Let me note the following important point. https://darkolorin.com/posts/1011.html https://darkolorin.com/posts/1011.html Mon, 24 Feb 2025 20:07:14 GMT <p>Let me note the following important point. With this release, Anthropic is making a pretty clear bet on coding. Nothing else. On all the other benches it looks like they didn’t even bother trying to improve, but on SWE they boosted it hard (which is exactly about coding).<br> All of this only says one thing: I’d be thinking twice about other agent systems and their overall reason for existing. Since anyway everyone depends on the base model.</p> <p><a href="https://t.me/chillhousetech/1011?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In one line, new release of claude 3.7 - I’m off to test it! https://darkolorin.com/posts/1010.html https://darkolorin.com/posts/1010.html Mon, 24 Feb 2025 18:43:24 GMT <p>In one line, <br> new release of claude 3.7 (<a href="https://www.anthropic.com/news/claude-3-7-sonnet?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/news/claude-3-7-sonnet</a>) <br> - I’m off to test it! But since it was my favorite model even before, like in the old version, I hope it doesn’t screw it up! And I like that every release comes with a clear strategy for where the guys are headed and there are even timelines (in your head multiply by 2–3). And they believe real automation in complex tasks is only a couple years away (and as we all know, all our tasks are complex).</p> <p><a href="https://t.me/chillhousetech/1010?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Would you listen to advice from someone who couldn’t operationally pull off… https://darkolorin.com/posts/1009.html https://darkolorin.com/posts/1009.html Sat, 22 Feb 2025 14:25:23 GMT <p>&quot;&quot;&quot;<br> Would you listen to advice from someone who couldn’t operationally pull off himself what he’s advising? Example: a startup advisor who hasn’t raised a single round except from friends etc.</p> <p>- Yes, only the info matters<br> - Don’t know, depends on the person<br> - No<br> - My own criteria<br> - 😘<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/1009?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Since I’m into quantum computing and quantum mechanics in general (not super… https://darkolorin.com/posts/1008.html https://darkolorin.com/posts/1008.html Wed, 19 Feb 2025 17:30:42 GMT <p>Since I’m into quantum computing and quantum mechanics in general (not super hardcore unfortunately), there’s this bit of news that<br> Microsoft made the first quantum chip (<a href="https://news.microsoft.com/source/features/ai/microsofts-majorana-1-chip-carves-new-path-for-quantum-computing/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://news.microsoft.com/source/features/ai/microsofts-majorana-1-chip-carves-new-path-for-quantum-computing/</a>)<br> that supposedly will let it scale.<br> For context:<br> - if this works then this is an invention on the level of the transistor (we all know what the transistor led to and all our computer tech exists thanks to it)<br> - this is cooler than your AI<br> But yeah, gotta watch it of course.</p> <p><a href="https://t.me/chillhousetech/1008?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> We’re in for replacing junior devs, we’re in for an agent on top of an agent… https://darkolorin.com/posts/1006.html https://darkolorin.com/posts/1006.html Wed, 19 Feb 2025 09:59:01 GMT <img src="https://cdn4.telesco.pe/file/a0KgIhBxd6BTvOenssmJ7JqbBPvuo7MGKJPyIOZ164Rk9gTZsgAZWFcl8Ccfidb5t5kP0i_3t1jklnioywkDlQZKgc9JosAAxmy5WW1KWCf1HWKKRGPGvfIMzOTOho-3yjTxBKDIbpotmzhLx914aVRkLcRT9rCv4nv5ft_fDi0VPBdeTZ5ct4ddO7yVu7j5Sa2RGfDbqK7t-1ec8tDhIWEgC9IKHzopECtUbyVynoE4VqS8KhIwkapFJ23bPOA3trX7_Oprj0n0XdWw9kGEBVoju3H4djrEsHqD1QW0U_Y1ttxZNozLwp6m87j7fWwTtSAlrseSgCR2MRKg2GUrRQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/oKXDWySjCltwFwG_vW6duRTHsu_6TjblzpPiQp5F3ZO3IWhZZYHltWhJls-9UvKXpmgu7-KTKRQ3ePkPHrWoVUD0U0NmSs0-r_i18IP-WB5CRCU9yDqe2dcLjHtGVGpy8jHJxx2BUiZFT0ZSvu0F0UNgX51HxfHGHsTQN6zsNKxEg2ICwMX1t9DdCTcdXlaqvC9QBQVTp_yw2XVZarrxyX-_alClMJWSaXzKVFdIYEoVWQ5_OXmWUG7oOBuWlE91TSfWxja_lXNqoa2nHhdKl5zYapoUguorseYTwdlq4Cf3U1MJXuZgu8PK1D5Mlfh6WFPeBOM4Q8ap1cirTSW_Ag.jpg" alt="" loading="lazy"> <p>We’re in for replacing junior devs, we’re in for an agent on top of an agent for anything and everything. But like always overall, so far all these agents seem basically useless to me autonomously and only really help a human with routine tasks. And as confirmation of that, OpenAI has a new paper. Though sure, the results are impressive. Gotta try grabbing some website tasks for outsourcing like this :) maybe we’ll make some money!</p> <p>SWE-Lancer: Evaluating the capabilities of modern LLMs in real freelance software development. (<a href="https://arxiv.org/abs/2502.12115?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://arxiv.org/abs/2502.12115</a>)</p> <p>That’s how it is! They took a ton of real little tasks from Upwork (here of course we discount for quality and understand that tasks inside corporations will be totally different in content and complexity) and made a bench. Who can earn how much out of the current nets.</p> <p>TLDR:<br> 1/ Claude tears all its competitors apart and even beats the o1-series reasoning nets (wow, admitted defeat, hoping for the best)<br> 2/ the results are very far from ideal and all the nets don’t handle even half the tasks, although in the manager section (accept/choose the right solution from ready-made ones) things are way better, but still only about half solved correctly.<br> 3/ the task focus is on full-on full-stack development, by the way</p> <p>Pics of who managed to earn how much out of a million bucks (successful completion of tasks from the bench is a conditional million dollars on Upwork).</p> <p><a href="https://t.me/chillhousetech/1006?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And do you remember that AI Pin from those ex-Apple guys, innovators and… https://darkolorin.com/posts/1005.html https://darkolorin.com/posts/1005.html Wed, 19 Feb 2025 01:29:43 GMT <p>And do you remember that AI Pin (<a href="https://www.bloomberg.com/news/articles/2025-02-18/hp-116-million-deal-for-humane-includes-ip-but-no-ai-pin-device?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bloomberg.com/news/articles/2025-02-18/hp-116-million-deal-for-humane-includes-ip-but-no-ai-pin-device</a>) from those ex-Apple guys, innovators and iPhone dads, who worked with Steve Jobs himself. Well now HP is buying everything for $116M (don’t fall for the number, it’s most likely debts and investors’ cash), and the greats are turning into just department leads, so they can make hardware for conference rooms and printers. Who would’ve thought? (Yeah, actually I don’t get how they got money for that in the first place).</p> <p><a href="https://t.me/chillhousetech/1005?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> 😘 https://darkolorin.com/posts/1003.html https://darkolorin.com/posts/1003.html Tue, 18 Feb 2025 14:23:42 GMT <img src="https://cdn4.telesco.pe/file/tcSOBpFAGuj_iiJmtZTtprS-LVvpQnTE_30EKfkxuvSLxVfeaDsajta-CHA2ZgJz8XptyMVRdOcn2N3sTuc07eynTqKI__783YM3lAPV9-TjXxp92P9wnozmITIuzHKm1KjFlS8vI4ZX2IMiLbyR79DBIG-ijc_D6ys3g5WWXYa05VO6YtaPrYCAwL1a_KPtfoGC4PNXQOvwH8wPjc5LW0DJHvQa8ucpRz2bjoallSp5Lo4JDBfusmkZpDxRMkSz9A3vzHFFB2jQMqI4BE1m_6fON8eAiobSxDLJx7oLLLnZtQOHFI1Z_k3B7zBVLRAT4u7b93pUYHykN2WJLqkhDg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/v0-pqhaMvMEL9zFGumOHEq8ichqvgfURptwGvkc9k-P3thjVWVd_hMAa-t8tnnE-l7GZmZ0eDs0HOj4r2_PPYSbWlm2e0jrbdZrCsL-lzeHOpqwi8RZv6UzXW72aBXz9bcCO2atK7oh15B80I6YDIQYc-Ife45xQrl1OhUmDQAxWC5KDLxt1vCt70xxmRHFDrZQm00ob1KHAGhNJ_Wde1OwVPqc-J65ry0fDUeDuoqknbY08B2glK5d6DI7nHq2hOX3p04sJuAshDRNoz2Bn2plC3npUPlXUEBfbWcOhwLUG4O8Rom9mjL-VZpkoI3rRvowB1seL49L_LS52YgvacA.jpg" alt="" loading="lazy"> <p>😘</p> <p><a href="https://t.me/chillhousetech/1003?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> What’s the deal with product manager salaries in the US — got some pics from… https://darkolorin.com/posts/1001.html https://darkolorin.com/posts/1001.html Tue, 18 Feb 2025 14:10:17 GMT <img src="https://cdn4.telesco.pe/file/s2N9Uj5Nm5ODarDlvHOu2UgPjlpvKRZRPK7YplVUGpxbQf7pucAbTs0_lOjrcejM8_V8W9cIA1Wn5LQQnydGt64O0w_y1lRPhB3gd0g7KSd2Gd3SDGm5soCoIJkM6WX_8IXHZToyVZYYKekc9tiwPyYemPzC9roKTU662uZs6ujxeL7SBFweWLEJmpTdSPWnWwKuFWRcb7wTGeYJnLCKX08knQbJacmxLx0NiRMlqylDeUbldZmmersRkcTut5UAM5KmSNaHle-5cIaGNFEJ6SE2bGzm4JOw3MvkxNPSKDbiw1vvJ_jhD5ZU1x9fYM43IL6sbK5_Ns7fv11FFcynUQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/RV2pqQ3i8EiCobdQmy04uxjLrd8Fbvcg7OTPUIVWtbXvS6aNVJC9IlHvaQYkc99KV7MfzUz96qc2xEnR42D4EV_lee3SN7y6IUenio8ri0ffFlfcJ8GdctQBg3qeIt6o8948DuvkaGZeeoyjGIWtyB0Ba3Mx8Rvh_NR5iT0ISRkRb4cqpOG_aAvE6HAsKAXx2OCW6WaH-Auag37Y4rvwEy3K353gIc_WnJwzsZr1DOkqvkU1Wr6tsTLgvNSwPtHOi6IMrqqUgS90nATTmt5z9m0hyH35hpFl8L6cbipviPnSiAhr0SizN5fJ4DGR2VbNeSYGO549sZ49olxII9gK6Q.jpg" alt="" loading="lazy"> <p>What’s the deal with product manager salaries in the US — got some pics from Lenny Newsletter (<a href="https://www.lennysnewsletter.com/p/how-much-product-managers-make-in?utm_medium=email&amp;action=restack-comment&amp;triedRedirect=true" rel="noopener">https://www.lennysnewsletter.com/p/how-much-product-managers-make-in?utm_medium=email&amp;action=restack-comment&amp;triedRedirect=true</a>).</p> <p>TLDR besides all the salary stuff:<br> PMs in the UK pull in only 65% of what American PMs make, and Europeans are basically stuck with less than half of the US salary.<br> If you want to consistently rake in max cash (read: life stability and predictability with a long road ahead) — march your ass to a public company in a top US city.<br> And if you want the highest chance of upside — funded startups are your pick.<br> So yeah, as usual: stability or a shot at taking off fast. Classic.</p> <p><a href="https://t.me/chillhousetech/1001?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Generally speaking, I’ve long thought the resume format is broken in itself. https://darkolorin.com/posts/1000.html https://darkolorin.com/posts/1000.html Tue, 18 Feb 2025 01:43:47 GMT <p>Generally speaking, I’ve long thought the resume format is broken in itself. And it says almost nothing about a person. What am I getting at?<br> So, what I’m getting at is that the central point in a resume is usually the company name where the person worked and the time they spent there. The “successes” section—we all hopefully understand it’s worth nothing, because nobody ever checks it, and it’s not even clear how you’d check it (well, except maybe the hardskills tied to it). And everyone, like, clings to some условный Google, Meta, Stanford etc. The whole logic is “well if he passed their screening, then he’s a top specialist.” Even though their screening is a set of template tasks, at the scale they’re running now.<br> I’d look at something else. Namely, the people this person worked with. Who was their lead or who ran the team, what that person did and who else they worked with. Seems to me the foundation of learning and progress depends on the people around you. So for everyone looking for growth—don’t chase the logos of new funded startups, chase the people you’d actually want to work with. It’s people who shape your understanding of what success is and how to achieve it; it’s people who can give you freedom of will and decisions, or take it away. Not the company name or its profit.</p> <p><a href="https://t.me/chillhousetech/1000?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In short, I often hear the following “genius posts from leading experts in the… https://darkolorin.com/posts/999.html https://darkolorin.com/posts/999.html Mon, 17 Feb 2025 14:02:58 GMT <p>In short, I often hear the following “genius posts from leading experts in the field of everything that’s trending.” Like, all your ChatGPTs are, you know, unreliable, it can’t do any real tasks. Remember, it couldn’t even compare two numbers. Or it can’t repeat after me, or calculate something that even a regular calculator can do.</p> <p>Well, a calculator is exactly what we’re going to talk about. As an example of a technology we’re used to and that most people basically don’t understand.</p> <p>So what’s the difficulty? And the difficulty, my friends, is floating-point numbers — those, in short, non-integers, something like 0.1 or 0.2. For example, did you know that addition is not associative on modern computers. Meaning, it can happen that:<br> 0.1+(0.2+0.3) is not equal to (0.1+0.2)+0.3</p> <p>To implement operations with numbers like these, the guys came up with the IEEE 754 standard (<a href="https://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html</a>)<br> which over 60+ pages describes the rules. And even if you follow it, the exact same program can produce different results on different systems!</p> <p>A few underwater rocks from theory:<br> 1/ When subtracting close numbers, catastrophic loss of precision can happen<br> 2/ Intermediate results can be rounded differently depending on the implementation<br> 3/ Numbers can be represented in different formats (single/double precision)<br> 4/ You need to correctly handle special values (infinity, NaN)<br> 5/ Operations must produce a correctly rounded result</p> <p>Once upon a time, back in the old days, Google hired one engineer to write a calculator for Android that would work well and without errors. His name was Hans Boehm. So this guy got so obsessed that he put out a whole spec and paper about the API for working (<a href="https://dl.acm.org/doi/pdf/10.1145/3385412.3386037" rel="noopener">https://dl.acm.org/doi/pdf/10.1145/3385412.3386037</a>)<br> with numbers.</p> <p>So, know this: what today seems extremely precise and you trust it, in reality not only can it produce errors (even if understandable ones), but it’s also a gigantic amount of work over dozens of years.</p> <p>P.S. by pure coincidence a similar article came out on HN<br> <a href="https://news.ycombinator.com/item?id=43066953&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://news.ycombinator.com/item?id=43066953</a></p> <p><a href="https://t.me/chillhousetech/999?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Thoughts into text with 80% accuracy. https://darkolorin.com/posts/998.html https://darkolorin.com/posts/998.html <p>Thoughts into text with 80% accuracy. (<a href="https://ai.meta.com/blog/brain-ai-research-human-communication/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ai.meta.com/blog/brain-ai-research-human-communication/</a>) <br> Awesome! Just a little left :) like 10 years, because I don’t want to be strapped up with sensors up to the eyeballs. But it sure looks like magic.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/998?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Managed to get into a closed event in Paris where Anthropic’s CEO — Dario… https://darkolorin.com/posts/996.html https://darkolorin.com/posts/996.html <img src="https://cdn4.telesco.pe/file/Gu6eTB2eJ3v4aXBnExjeJMyOn3QIqotsk8qcBxqpZ0duzjIGyjM4REpAPtbodJpf8MHZsQWworNkIyB3SbMJmOCkC6uoK2wfKg94O-RLzTusBBop7hSm2LFaYkTa3ASPfmQgqMOC6xRZwi5VUB5gem3GW7FoSZLuBjSxca_bJpHWpz1A8NStgns3_RBXw-ie3GYOuw-rXeffYotlRd1wUUr9E5Jz0BBTvl-baMCCi7rtHwWXLkFXLNSsqKTIhC5Luja5M35yFadakQPSzou0UtPX8eqdi49Dsly3WOVBHbT-27_NTCFFjAtePNOA9nUKwup5GvvMnwVgXu5hbEhJmg.jpg" alt="" loading="lazy"> <p>Managed to get into a closed event in Paris where Anthropic’s CEO — Dario Amodei — was speaking. Overall he pretty quickly retold his latest blog post, but dropped an excellent spoiler on what to expect from Anthropic’s new model. I’ll write more later once the event wraps up.</p> <p>Update: well, so Dario gave a fairly detailed example that a person chooses for themselves when to think through a task and when to answer fast. And he’s very surprised that there are “thinking” models and there are ones that supposedly answer right away. He hinted that in Anthropic’s new model, the model itself will decide whether to think or not. That’s exactly what’s needed. At least someone isn’t doing everything just to ship for the sake of benchmark metrics.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/996?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Hi to your agents. https://darkolorin.com/posts/995.html https://darkolorin.com/posts/995.html Tue, 11 Feb 2025 09:40:42 GMT <img src="https://cdn4.telesco.pe/file/QnrieDC99F7ZPKtDdZXkLb38LUIzjVs08bgKAdtY7u9Y2B0-L2R187l_FRL82LxG4TZNK_dcB_LAfbOJFU5G1zqj-8GOzRywZn3Ibk0ZvKiSQduKD_mkyWmkS2q3a21GN-UQVmnZDQ0Dd9hSsIzHFrG7-sTK_zqD-BPO6fv5ENapCNyX2pFfJDhhBXP-rD4Udk8HOjG0Cdtcn-pJRtihYQmnNMJt90Q5NCWopq5yjdqxlvoRaeEcRQuZGYHEE5B2sorgwKyRUHNRy9cfN5fUG22GwbRUORffTJ05eU02N7lbM7urEFHm1WjkWrr2kWAuEwIAVukf2Ig4RnSuM2u06A.jpg" alt="" loading="lazy"> <p>Hi to your agents. How’s it going?<br> So, I’ve been saying for a while: all these million-token contexts of yours and so on are bullshit. Like, purely in theory my gut feeling is that after 10k the quality drops off dramatically, and well, here comes the confirmation.<br> And now these “unique” data/knowledge can be applied to sales agents, agents for whatever, and you can sit there thinking “so when exactly is AI going to replace a human at the workplace?”. As they say, a note for your agent startup.</p> <p><a href="https://t.me/chillhousetech/995?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, in one line: Musk wants to buy control over OpenAI there for almost 100… https://darkolorin.com/posts/994.html https://darkolorin.com/posts/994.html Mon, 10 Feb 2025 21:01:46 GMT <p>Well, in one line: Musk wants to buy (<a href="https://www.wsj.com/tech/elon-musk-openai-bid-4af12827?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.wsj.com/tech/elon-musk-openai-bid-4af12827</a>) control over OpenAI there for almost 100 yard, and France is also dropping almost 100 yard into datacenters. Fun times though.</p> <p><a href="https://t.me/chillhousetech/994?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Wow, look at what kind of topics Anthropic is posting. https://darkolorin.com/posts/992.html https://darkolorin.com/posts/992.html Mon, 10 Feb 2025 15:30:54 GMT <img src="https://cdn4.telesco.pe/file/gdqBqJ2yb9Nijuke7nLDLV6velGg5HPdwYZqnf9RoD4UBA5m4RdDtNXLBvmBNFX83wV5SyOQekqpLkkHNSWOk9XI4kwTcIi7xYypOOgpvLzTpY_U5xnNrSxTGQPtSuxa1GGs0bcKDXIWQEhWPCh83CmEBvcoJdqaWCUqQ78Xdy2XPwf3JsvGpVimkd9I-nY9Q_MOFuS3rgCjsNfxkb7yvL3endI3p-aluIbt7__W6D8cLPQ9pnkYByTMLF7zIzbFxP-i4dk6I7zAxNl0FZlVwCxMHl6jf9kLtI35JQMlRJ-bypwZcsOsuWagnIRO_wJ3al5xyNgxzeqli7XJqYMJYw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/mauJzyIgKc4qEXwJLpveK9njKd_bJc4y68DJP-awtYj5E9FZhLz0QzfUZEAqiOD1thpWP1mGf80_Fyl6ZrUCQeo26w1UNEulYZonWmK7ZPYQUtEMhd38u7JX0d3miWGr3T3j2O4Nw1kdBBHosSBAHVSQqYrxd4h-Kz4u07asLilTdyiU55tTY09IAcZtGgsbxXClVExArGCAAJPiYJDcx7zzKKK-LnaZ5GRoR9gjoXZ8kaP7MBt5OwNxZ-oFMUSvJdyn4oDuAeDRCMaIt6nApUYjvlpG2zPevZr5THmk6BEipCiAK-E6trRLCOvr_B9D_DiFUR7qPutGgl5hcv07wg.jpg" alt="" loading="lazy"> <p>Wow, look at what kind of topics Anthropic is posting. <br> Can’t not share. (<a href="https://www.anthropic.com/news/the-anthropic-economic-index?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/news/the-anthropic-economic-index</a>) <br> The guys put out a study on how AI is being used right now and what its economic contribution is (though here I didn’t really get it). The numbers are extremely juicy. I’ll attach the little pictures. <br> Key takeaways: <br> - most usage (36%) of AI goes to tasks related to technical documentation etc and software development (who would’ve thought, right?) <br> - a slight bias toward not full automation, but so-called augmentation of tasks, that’s when you check or supplement a task with AI (I think the gap in favor of augmentation will grow a lot) <br> - most of the tasks being solved are categorized as mid/high-income professions. And basically nothing with the lowest- and highest-paid roles. Here I think it’s just a skew from the audience that uses their Claude. Unlikely that janitors or CEOs are entrusting their tasks to AI. For now.</p> <p>Highly recommend reading it, it gives a lot of insights both on areas of application and on what can actually work.</p> <p>I still think: <br> - full automation via agents is a long way off even for simple tasks (see how “quickly” self-driving cars entered everyday life) <br> - more and more use will be in augmenting tasks and improving outcomes (a la a helper tool) <br> - agents == evil, basic automations are dope (separate blocks in processes, latency, quality of certain decisions — basically swapping an old tech for a new one)</p> <p><a href="https://t.me/chillhousetech/992?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “Read, but didn’t fully get it” column. https://darkolorin.com/posts/991.html https://darkolorin.com/posts/991.html Mon, 10 Feb 2025 13:37:54 GMT <p>“Read, but didn’t fully get it” column. Let’s go. <br> 1/ DeepMind released this paper into the wild (<a href="https://arxiv.org/pdf/2502.04403" rel="noopener">https://arxiv.org/pdf/2502.04403</a>) — maybe not the flashiest at first glance, but IMO important. They raise the question “what system can you even call agentic,” and therefore “apply reinforcement learning.” <br> 2/ Same DeepMind dropped the tome “How to scale your model” (<a href="https://jax-ml.github.io/scaling-book/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://jax-ml.github.io/scaling-book/</a>) about how deep learning works at the hardware + data transfer level, how compute scaling for this whole thing actually happens, and compute efficiency. Just a must read for anyone who wants to really understand why the hell we need datacenters here and what DeepSeek’s trick is. <br> 3/ For hardcore readers (<a href="http://distill.pub/2017/momentum?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://distill.pub/2017/momentum</a>) — though damn, everything above is already hardcore — there’s a great article on how gradient descent works and why momentum as a concept works in this story. In short: bookmark it. Because this is a turning point in deep learning overall. <br> 4/ A bit (<a href="https://www.high-capacity.com/p/chinas-overlapping-tech-industrial?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.high-capacity.com/p/chinas-overlapping-tech-industrial</a>) about China’s ecosystem of tech companies. <br> 5/ For everyone who completely missed it and doesn’t get what an LLM even is — 200 pages (<a href="https://arxiv.org/pdf/2501.09223" rel="noopener">https://arxiv.org/pdf/2501.09223</a>) from scratch to the end (mostly theory). <br> 6/ Andrej Karpathy: 3 hours of LLMs (<a href="https://youtu.be/7xTGNNLPyMI?si=HqGFz0cYdIT6igQp&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/7xTGNNLPyMI?si=HqGFz0cYdIT6igQp</a>) for engineers from zero — recommend (mostly practice). <br> 7/ A rare video (<a href="https://youtu.be/b7y971yL5dE?si=oOusztsVPb0uB8MB&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/b7y971yL5dE?si=oOusztsVPb0uB8MB</a>) of the founder of musically (aka TikTok in the future) about growth, users, etc. A treasure trove!</p> <p><a href="https://t.me/chillhousetech/991?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sunday stuff. https://darkolorin.com/posts/990.html https://darkolorin.com/posts/990.html Sun, 09 Feb 2025 19:18:52 GMT <img src="https://cdn4.telesco.pe/file/UwXIAdQ27aLJ-G-vz8vsP3H4i6MjtHNz512tfbEDwVn-bZqaX3zKmCNdRHr2-1GyyBpsnB2Ooxu79naGWPmlqMH7bmaoYsPkWLV4hnKanAA0FgU7cAyUf_s-pfU-H-WuGKAQEdQH7EX8azaU9_xFlzZCfhhPOzpU9mKMIXOfTJfsEOxbU5ciNj3H-DM4jcApZkDIIHyTnFXO1ZUDN5uIiNQltsWFOM17vGLYS9OTXsXRl4r0nothW8U71-P4MmZu-mlg_NFA_YqjVkWTuD4ekf8aX9hB_9BmZ5CEwDQIpUjDkYlU8X2cAyBs3_0frw-q7tAtwqrKcGybhaThT3DQhA.jpg" alt="" loading="lazy"> <p>Sunday stuff. So what do the partners at one of the most influential and biggest venture firms, A16Z, actually read.</p> <p><a href="https://t.me/chillhousetech/990?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ve been meaning for a while to write about this thing in the world of modern… https://darkolorin.com/posts/989.html https://darkolorin.com/posts/989.html Thu, 06 Feb 2025 12:17:48 GMT <p>I’ve been meaning for a while to write about this thing in the world of modern product managers called “conversion optimization culture.”<br> In reality it’s useful for everyone. Often when I talk to product leads or startup CEOs (and they usually lead product) I run into this narrative:<br> - we surveyed people or found out that in this funnel there are this many steps<br> - and all they try to do is increase conversion at some step, head-on, via features (new buttons, data, and so on)</p> <p>And years go by on this vicious loop of “come up with a new gimmick.” For years your product is slow as hell, but you’re still inventing new gimmicks, buttons, pictures to increase conversion. All of this drags on from that famous story about Marissa Mayer (the first product manager at Google or almost the first), who went through 7000 shades of blue to choose which color converts better in search results into clicks. And then this logic is the “base” for everyone building their own project.</p> <p>But a lot of people forget that some parts you can just speed up. For example, bookings on<br> <a href="http://booking.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://booking.com/</a><br> happen 30% more often (numbers pulled out of the ceiling from discussions with different people, don’t nitpick) if the site loads in under 300ms. And my favorite example is bidding (the auction for which ad to show) on-device. In 2016–17 Facebook and other DSP/SSPs (they sell/supply ads) made it so that latency for showing an ad to a user dropped from 800ms to 200ms on average, which made it possible to show video ads and moreover preload them onto the user’s device in advance (because we’re bidding for the impression right inside the app and not waiting for the server). This thing in its first year of existence brought in more than 5 yards in revenue (and margin in ads is usually 40%) in the first year in Facebook alone.</p> <p>Why am I saying all this. Look at your problems wider, think about what’s actually important and what isn’t. And that people love the agony of choice and you don’t have to shorten it, you can make it more accessible or faster.</p> <p><a href="https://t.me/chillhousetech/989?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well there you go, there you go — this is what you should be making your AI… https://darkolorin.com/posts/988.html https://darkolorin.com/posts/988.html <p>Well there you go, there you go — this is what you should be making your AI for, for ads. <br> A top tool, basically. (<a href="https://icon.me/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://icon.me/</a>) <br> We’ll try it “for just a grand bucks a year” :)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/988?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Recommendations for the new YC batch this spring. https://darkolorin.com/posts/986.html https://darkolorin.com/posts/986.html Fri, 31 Jan 2025 16:00:03 GMT <img src="https://cdn4.telesco.pe/file/LDzL6GPyIm3fl0tgPpgwY0dL-X32EfuzLZTQ5FR3Bbf4GcYw90afqXa00xjFSLu6qwFpQJem-YNVYm2_nzsyc7437ipDZ2ic0mwk2NT9AFhkMVIHFUCaIgzkdrTL7Nq_6xADAHmdCqoNZPrSdhvck_myC2oW-q7iFghQN8L93OKZIp4taQ5V8YZ024PMxFsToUggt-3pVaDnF9d8EY9_mSNxz7xToKWzgry_BU3GKAg6F3rxBSZkrtguLP9oq6QTFSx9jZmzEXs3vPl61ephKIyKxsXvCvQusMrz2hC_5MW_XOyGwAcvakJR_qM49N290wxUaZ34YzfLXOzTHgmmLA.jpg" alt="" loading="lazy"> <p>Recommendations for the new YC batch this spring. Let’s see what’s fresh in there. Well, it’s still infra, and now it’s agents already, the kind that will straight-up completely replace people at their jobs — what co-pilot are we even talking about.</p> <p><a href="https://t.me/chillhousetech/986?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A few career twists, or how after coupons you go and found a company to build… https://darkolorin.com/posts/985.html https://darkolorin.com/posts/985.html Thu, 30 Jan 2025 01:37:33 GMT <img src="https://cdn4.telesco.pe/file/bZYwRgflGQd8zBTxZY3yyrYuw8FV7tRMdrqIIPl6skYeq7_LrM3kpb-3Dy_WRVIbXocSGdK_mwvSXxkKIPN9dwGZa9vybKXBfAfpP8T5YrcVeW934Xl1eAS9Rt3YnJdUoXrEdM7Y3xgzmRyW1VEyj2lfteH8Zucu088ZYvKnGXsqxeSvd50ht17rimxeh3CVAJ39F_K-69IPxrTZz6N_F7nWj0l0y5vlgFjH4ein3ERw_hWIFfgVRHajP9tbOygNi-AW2mtqHd0_rC7B9sby_VNzbfwqwOxYzv3qnwJVWyURYoVU7acsLlDhaobET5SAO0vbJpNpZ4_OMu6yb1HMBA.jpg" alt="" loading="lazy"> <p>A few career twists, or how after coupons you go and found a company to build supersonic aviation. And the whole story is in this <br> tweet (<a href="https://x.com/avichal/status/1884745797004357813?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/avichal/status/1884745797004357813?s=46</a>)<br> .</p> <p><a href="https://t.me/chillhousetech/985?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For those still interested in the cost of training a model, Anthropic’s CEO… https://darkolorin.com/posts/984.html https://darkolorin.com/posts/984.html Wed, 29 Jan 2025 18:33:17 GMT <p>For those still interested in the cost of training a model, Anthropic’s CEO dropped an essay with some politics mixed in (and (<a href="https://darioamodei.com/on-deepseek-and-export-controls?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://darioamodei.com/on-deepseek-and-export-controls</a>) here AI already looks like an arms race).</p> <p>One interesting point he emphasized: models that “reason” are always based on a base model and its quality. Meaning all this “reasoning” is basically a consequence of whatever you managed to get earlier as the base (the base is just trained to spit out an answer to a question, like a very smart autocomplete/sentence-finisher, as I like to put it). He also said their cutting-edge model was made for roughly a few dozen million (I’m thinking 25–30).</p> <p>Anyway, I recommend reading it if you want to understand the AI market. Because this tech will be everywhere, in any business, sooner or later.</p> <p><a href="https://t.me/chillhousetech/984?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> What is dead cannot die. https://darkolorin.com/posts/983.html https://darkolorin.com/posts/983.html Mon, 27 Jan 2025 23:17:37 GMT <img src="https://cdn4.telesco.pe/file/lXxQHYgid2_tbmyDv6p99lMBcUEPghHOb0Iq7jaCqlBHa_zlsn3Yg1xbPZOlsbkx-b2jJoJ3NSdsT-dTYIvejILbG0kPbEg7aB1gdszvntXnm8H5bkTWWgMr6XyUYiADGQmdOFjDc0MOVP-xiLmB9vUzfVUm7GO4LiQQ5fBsEgKY2hgn9sy7DQeBc0_BN9MpyKI7czyNbUewOol6Uy9n2oxV6n7Ydp_fQZ2QaVZJY79pnFJ_W7TMyuY2TMrxfMn2eGyU8t802GXmcYKmxO86JYZlNt8658vXZX8LBVfbhG_VtQm_boi2MDhFnvXQm3DWZceNDmvwzsGF7GK--tRIoA.jpg" alt="" loading="lazy"> <p>What is dead cannot die. You can only sell it and make a new LLVMH holding. <br> And yeah, maybe it’s just rumors, but that French startup that once got raised up, called Mistral, seems to be in a crisis, and I think there’s a grain of truth in every joke.</p> <p><a href="https://t.me/chillhousetech/983?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Wanted to add my important two cents about these guys sitting on the first… https://darkolorin.com/posts/982.html https://darkolorin.com/posts/982.html Mon, 27 Jan 2025 15:00:54 GMT <img src="https://cdn4.telesco.pe/file/andtJZYrmVeGHNfiLOanq1FfF47OVTPhpoyCLQAcYE7akLYWTXzFVed8-P7-JSO4kkoyz408o-WNEI65JTr0uvSgm8FqzWhwK5m9lzkhOA28dpgZFswhlQxtITMN7TNUpgVvRwLtPY5EyoCx7VbOJ0-FmyiZwF2z3bhg8NQ3MXNmV7vd3LSRYQ5hmApxWlgXkaj5bwDZ4ip0_2a3GYanoDEp1nco2myAT2U6BCD4-TvdTZard4ebE-kx6TPHaX7PoXr7y-2S3ZYYoKtloLBtDFlHZGH3j6s_5_we9KU0aGgCrs6U74M0_Mhd8VxIvGm-gAR5Fk8KU9mfSJ_TWtnvCQ.jpg" alt="" loading="lazy"> <p>Wanted to add my important two cents about these guys sitting on the first line of the US AppStore. This is, so far in my memory, the only example where literally all optimizations have been done vertically, from the math all the way down to the hardware. Basically the solution architecture in the compute data center is tailored to a specific net and its training. Welcome to the guys from high frequency trading. You can read it carefully here. (<a href="https://arxiv.org/pdf/2501.12948" rel="noopener">https://arxiv.org/pdf/2501.12948</a>)</p> <p>Adding: for comparison, it’s like Apple in the world of consumer devices. Basically everything, starting from the hardware, processors, software, the operating system, is made by one company. I’m sure the Chinese are sticking to the same logic, and practice shows it’s always the strongest product.</p> <p><a href="https://t.me/chillhousetech/982?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Some thoughts on agent systems like Operator from OpenAI and Claude has a… https://darkolorin.com/posts/981.html https://darkolorin.com/posts/981.html Sun, 26 Jan 2025 13:10:21 GMT <p>Some thoughts on agent systems like <br> Operator from OpenAI (<a href="https://openai.com/index/introducing-operator/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/introducing-operator/</a>) <br> and Claude has a similar thing. Like you give it a task and it drives the browser (moves the mouse and uses the keyboard). So here are my two cents:<br> 1/ Honestly these are some kind of public experiments, in their current form it’s a product with no future and clearly outdated. Andrey Karpathy wrote on Twitter that they’d thought up a similar thing back in 2016–17, before, so to speak, any meaningful progress in transformers. Why the hell emulate browser behavior if you can use an API or scripts or something else and do it without the visual layer that obviously slows everything down like crazy. This is basically a nod to the <br> MCP project. (<a href="https://www.anthropic.com/news/model-context-protocol?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/news/model-context-protocol</a>) <br> Which will make all this visual trash unnecessary.<br> 2/ Still, standing apart for me is the old product problem of voice assistants (btw do you use them often?). For most tasks it’s harder to set them by voice/text than to just tap through or click around in a normal UI. In short, the phrasing itself already causes trouble, not to mention that changes can happen along the way, and controlling all of it by voice is kinda painful. Try to imagine the process of ordering a pizza to your place (let’s say the system knows where your place is, you start choosing where to order from, picking a pizza, oh and you gotta check the price, and maybe even a photo or ingredients, well basically we haven’t even gotten to the part where delivery shouldn’t be to the door but downstairs, or like special instructions on how to get into the courtyard). In other words, a graphical interface massively simplifies interaction with information and how it’s presented. How to make the interaction faster, I don’t know yet. The idea of embedding an LLM somewhere is totally fine, but right now everything works like shit and there’s no trust.<br> 3/ and yet with the release of the Chinese model <br> Deepseek r1 (<a href="https://www.deepseek.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.deepseek.com/</a>) <br> everyone started panicking like “look, it’s open source and everything about it is more or less clear,” except how much they actually dumped into training resources (if you believe the Chinese, they got the same as OpenAI but for “$5–50m,” which is of course very cheap). Still, using this goodness is hard. I find factual errors in every other prompt if it’s more complex than basic instructions. Or the ability to go on the internet raises questions. I’ll drop screenshots in the comments, curious what you think. <br> TLDR: keep an eye on the space, but agents in their current form are just research lab experiments.</p> <p><a href="https://t.me/chillhousetech/981?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Came across a company called Deep Robotics from China, the company’s 8 years… https://darkolorin.com/posts/980.html https://darkolorin.com/posts/980.html <p>Came across a company called Deep Robotics from China, the company’s 8 years old and it’s a unicorn, they make robots that are straight-up pretty cool. <br> One like that seems to cost 1600 bucks. What’s more, they’ve got a special <br> open source project (<a href="https://github.com/unitreerobotics/unitree_rl_gym?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/unitreerobotics/unitree_rl_gym</a>) <br> for simulation training. Pure bliss.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/980?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A small news hook the size of 500 yards of investment (feels bigger to me than… https://darkolorin.com/posts/979.html https://darkolorin.com/posts/979.html Wed, 22 Jan 2025 00:01:49 GMT <p>A small news hook the size of 500 yards of investment (<a href="https://openai.com/index/announcing-the-stargate-project/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/announcing-the-stargate-project/</a>) (feels bigger to me than any government projects in US history before this) into datacenters for AI. And they write that it’s even for OpenAI, but there’s no clear wording on exclusivity (or I didn’t find it). Which of course kind of reeks of an arms race, just in AI. Interesting times we live in :)</p> <p>And in general, an excess of resources and money often kills innovation vibes and the motivation to do anything.</p> <p><a href="https://t.me/chillhousetech/979?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So there’s this kinda mantra in the industry about quantity supposedly flowing… https://darkolorin.com/posts/978.html https://darkolorin.com/posts/978.html Tue, 21 Jan 2025 13:37:42 GMT <p>So there’s this kinda mantra in the industry about quantity supposedly flowing into “quality,” like, the result.</p> <p>The standard wording: you need to pitch your idea 200 times and you’ll get conversion, or when you’ve surveyed 300 customers you’ll finally understand what’s what.</p> <p>I’ve got the exact opposite logic. The more you’re grinding in one direction and you don’t feel any payoff besides emotionless rejections and all that, the more you need to change — and it looks like radically. Every rejection has its reasons. Understanding those reasons often leads to understanding the general principles of how this or that industry works.</p> <p>And what I don’t get most is when these endless rejections are presented like, “just keep trying, someday it’ll work out.” There’s no connection. You should get used to the fact that without (successful) experience, luck is a bigger factor than persistence.</p> <p>So what helped you more: persistence through a ton of rejections, or switching strategy and new attempts?</p> <p><a href="https://t.me/chillhousetech/978?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Pretty telling data on what percentage you should be giving to all sorts of… https://darkolorin.com/posts/977.html https://darkolorin.com/posts/977.html Tue, 21 Jan 2025 13:23:28 GMT <img src="https://cdn4.telesco.pe/file/TjodrtelaEaM5boLhbphyc7PG7xPAwymgJZo5IFp2vkq-Ck_mfGEm4irXHZCG5VwE9WEvr-R9iujfqwNhp7ohfn3f-BGLFDWQNj0t4CcMEe_A40Lw16kHDG81VbYrIOajiIFPwee-7I1xc7AfWobN5FO6Tx6-T2k8aT3PdwD_qzSPuUZ8rppK8srHk-pnK9z38suwcl3vQKO7RY3uswSg4Bzlic9FTivZuR8LUghHx9Rcq14XM5i327GYv7dWec0UQ4Iz7EUG2RDywFiOpMVSBdxGiM3HUx1SLCQICeMKwH2ESZ_bIXjI0VyMKtceA1lQUPSuur45XGbzsu8qaKa3Q.jpg" alt="" loading="lazy"> <p>Pretty telling data on what percentage you should be giving to all sorts of advisors and “counselors” in your startup. <br> No more than 1% per person, and in total usually no more than 2–3%, so to speak, over the whole distance.</p> <p><a href="https://t.me/chillhousetech/977?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Since the weekend topic is the TikTok ban in the US, I’m watching it, honestly… https://darkolorin.com/posts/976.html https://darkolorin.com/posts/976.html Sun, 19 Jan 2025 18:40:47 GMT <img src="https://cdn4.telesco.pe/file/RnlY8PjKtJskkHA-3aGuGYe67mjZOhwKOlqe8WgyN8mnMf19ZerF4JsLba-gL7-zvvNInCppN04pBWfE9VEameIwC2TLU1O13LWdlaWdfajztJaBu6Q9TjaAFeyAHQ7otavIB4ICNE8u_UMMYFUw5wyYTrbAPFwEvu3PYsBuHK3rNRDKrlPllPYqd3QTvyvFB5M6OxebMfgkZgixazQsALe3KUwGfsH8B2zEp9V9jlTryXMxpeDwu1GAQh33KZGBKfnn2n2_bcT8xhRQ8w8BLOlXf0jnsju4fMlMQsbYwZhoGZUTOHF82K8byuP94H4jQ_5oVOw7GY8in2S7djfguA.jpg" alt="" loading="lazy"> <p>Since the weekend topic is the TikTok ban in the US, I’m watching it, honestly it’s interesting. Of course Trump did everything slick: they ban TikTok, all the youth don’t believe it, then Trump comes in and says everything’s gonna be ok, they even changed the wording inside the app. <br> That’s how you rack up votes. Clean. <br> Update: welp, it’s working :)</p> <p><a href="https://t.me/chillhousetech/976?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, in the US they banned TikTok and basically everything from ByteDance. https://darkolorin.com/posts/975.html https://darkolorin.com/posts/975.html Sun, 19 Jan 2025 15:05:01 GMT <img src="https://cdn4.telesco.pe/file/KZNJl4kLloN_B8GkL_-IUdybTatNYynj3XAkO05lLtHTpA46HszkDrkxdx7ri8-hkHJZCJpX1M1L4VJvltZwrWZy3Prv_LHUXV7ar8QRzJzeYWnL0f2V36wxiMQ2bD5MtmwKqOXYSxrft_8CUwfOfhG8P2I5T7PKqb8SSEc3tSfhH0pI7-6HRGXfRsjo0nrpSa9RkGmXHCMxU7vgCwi37_ekZ78j-yBnEW9VP9ny07iOe8_wMO-yGT8Tx_7OihJbFn1_iwGbtbmsJuMk0lkRGpGJooI7D_AcAb3odqFxeig4nFqEaI4_To5ja0-njsb6pMw5NQFdjCWtv0W7FPikvg.jpg" alt="" loading="lazy"> <p>So, in the US they banned TikTok and basically everything from ByteDance. And here’s a bit of logic for you.<br> And yeah, update:<br> here (<a href="https://techcrunch.com/2025/01/19/trump-says-he-will-delay-tiktok-ban-suggests-a-joint-venture-with-us-ownership/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://techcrunch.com/2025/01/19/trump-says-he-will-delay-tiktok-ban-suggests-a-joint-venture-with-us-ownership/</a>)</p> <p><a href="https://t.me/chillhousetech/975?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A “read it, but didn’t fully get it” column. https://darkolorin.com/posts/974.html https://darkolorin.com/posts/974.html Sat, 18 Jan 2025 13:07:11 GMT <p>A “read it, but didn’t fully get it” column. <br> 1/ I’m so damn tired of the whole agents-and-their-omnipotent-power topic, blasting from every iron. Even found a paper like this, (<a href="https://arxiv.org/pdf/2407.01489" rel="noopener">https://arxiv.org/pdf/2407.01489</a>) which is like yeah, this is not a panacea. Also an Anthropic piece (<a href="https://www.anthropic.com/research/building-effective-agents?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/research/building-effective-agents</a>) says agents are needed in very few places and classic automation algorithms actually handle things just fine. <br> 2/ liked this article (<a href="https://rodneybrooks.com/predictions-scorecard-2025-january-01/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://rodneybrooks.com/predictions-scorecard-2025-january-01/</a>) on predicting AI trends. Just entertaining and very clearly written. <br> 3/ was checking out (<a href="https://www.numbeo.com/cost-of-living/rankings.jsp?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.numbeo.com/cost-of-living/rankings.jsp</a>) which of our cities are expensive to live in and which aren’t, didn’t know about this site before, and do you have any cool resources like that—share plz <br> 4/ learned what an invective is (<a href="https://ru.wikipedia.org/wiki/%D0%98%D0%BD%D0%B2%D0%B5%D0%BA%D1%82%D0%B8%D0%B2%D0%B0?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ru.wikipedia.org/wiki/%D0%98%D0%BD%D0%B2%D0%B5%D0%BA%D1%82%D0%B8%D0%B2%D0%B0</a>) <br> 5/ while I was riding I finished Lord of Light by Roger Zelazny, (<a href="https://ru.wikipedia.org/wiki/%D0%9A%D0%BD%D1%8F%D0%B7%D1%8C_%D0%A1%D0%B2%D0%B5%D1%82%D0%B0?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ru.wikipedia.org/wiki/%D0%9A%D0%BD%D1%8F%D0%B7%D1%8C_%D0%A1%D0%B2%D0%B5%D1%82%D0%B0</a>) recommend it to sci-fi fans <br> 6/ watched two interviews, one with Chollet (<a href="https://www.youtube.com/watch?v=w9WE1aOPjHc&amp;t=753s&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=w9WE1aOPjHc&amp;t=753s</a>) who’s a very cool dude in AI and launched the ARC Prize, and my favorite modern physicist Roger Penrose (<a href="https://www.youtube.com/watch?v=VQM0OtxvZ-Y&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=VQM0OtxvZ-Y</a>) <br> 7/ found my first lecture (<a href="https://youtu.be/pg1AMCNcmAs?si=bSx9kIZLuDoCbKUN&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/pg1AMCNcmAs?si=bSx9kIZLuDoCbKUN</a>) at MIPT, back in 2016 I think</p> <p><a href="https://t.me/chillhousetech/974?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Friends, Every day I’m on calls Helping Basically all I do is this… https://darkolorin.com/posts/973.html https://darkolorin.com/posts/973.html Sat, 18 Jan 2025 12:55:11 GMT <p>Friends,<br> Every day I’m on calls<br> Helping<br> Basically all I do is this info-business for you dear folks<br> So I decided to launch for you a course, a closed club, paid consultations, or maybe I’ve just got nothing better to do, and no money, and I gotta somehow earn, I mean I live somewhere in some country, and everything’s new here, and so it is every half a year<br> Yeah fuck that if you’ve got a main business or a job you actually like</p> <p><a href="https://t.me/chillhousetech/973?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The truth about VCs. https://darkolorin.com/posts/972.html https://darkolorin.com/posts/972.html <p>The truth about VCs. A good answer, and especially the comparison to the fact that if you’re sitting in the front row at top team games, that still doesn’t mean you’ll be able to coach it. <br> I also support the idea that the best partners in funds are the ones who’ve built companies themselves.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/972?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “Second-hand information always looks more coherent and convincing than the… https://darkolorin.com/posts/971.html https://darkolorin.com/posts/971.html Thu, 16 Jan 2025 13:10:06 GMT <p>“Second-hand information always looks more coherent and convincing than the full of gaps and ambiguities data a scientist has at their disposal.”</p> <p>Lately I keep running into posts about some kind of processes in companies (usually not corporations, but either startups or small-ish companies around 100 people). I’m a big opponent of bureaucracy at the early stages of development and the same at the early stage of growth. Found a few root causes for myself as to why:</p> <p>1/ any new process is usually one more step in modifying information, whether it’s meeting takeaways or yet another email or comments somewhere. The more steps, the more distorted the original thought is, same as with copying the Bible in the Middle Ages.</p> <p>2/ processes are like taste—everyone’s got their own. The hottest arguments are the endless ones about how we should run tasks and which framework to use, which rarely moves you toward a result, and more toward some people’s personal habit of using certain tools. And yeah, it’s easiest to discuss yet another process, because everyone will have an opinion there, unlike hard questions (where there’s actually an impact on the result, like architecture or design).</p> <p>3/ the less qualified the team, the more processes there are in your company, and that’s logical. Because you want to put everything on rails you’re supposed to ride on and think less, and thinking is hard. All frameworks usually send us back to plants (kanban), factories, and the like for exactly that reason. Working by instructions doesn’t take brains.</p> <p>4/ I don’t believe a process will help a person become a better pro, but it will definitely make the effort of thinking easier, so processes often kill creativity and ideas, because essentially they reduce everything to repeating the same thing over and over.</p> <p>TLDR: like science, tech is basically an industry of mental labor, and automating that will always reduce the chance of getting innovations. At the same time, I’m not going to deny that organizing work is necessary.</p> <p><a href="https://t.me/chillhousetech/971?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Coming back from vacation. https://darkolorin.com/posts/970.html https://darkolorin.com/posts/970.html <p>Coming back from vacation. So here’s the thing: Nikita Bier launched a “new” app. Explode. Same idea as before, just now the invisible-shit blasting goes through iMessage. Previously he sold tbh to Facebook and GAS to Discord. <br> What I noticed:<br> - the app invite sits in activity even if the app is killed, and it lives for a whole hour, wow<br> - after inviting three friends, the app is supposedly meant to start working, but nope — you get a paywall with a one-month free trial — brilliantly scammy<br> - everything else is super nice and they squeeze iOS integrations/features to the max <br> Distribution test runs launched sometime in August. Tuned it for half a year. <br> App link (<a href="https://apps.apple.com/gb/app/explode/id6596779134?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/gb/app/explode/id6596779134</a>) <br> Twitter announce (<a href="https://x.com/nikitabier/status/1879206793118658974?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/nikitabier/status/1879206793118658974?s=46</a>)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/970?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Haha, well yeah, it should’ve been for 2000 USD. https://darkolorin.com/posts/969.html https://darkolorin.com/posts/969.html Mon, 06 Jan 2025 13:10:36 GMT <img src="https://cdn4.telesco.pe/file/FzEp75sd513UCyKNOJSxc73tKdKIwtcxfT3jpjY_e4BG0OY8X11sffxo-8m2kbMnSgSJ9Q0cY5vMUyEcP7nNf3edwqHm7K29NbxG4340mF5EclBTvtgqWT_5zRV6EhE_-yaorilaMDDluYqZsm9gofGulW0cz9uLUnC7xnNiYvDJrtdDqWMR4_0YMJLwXW1dSVgw6uKkt05kT7t3WdYgftyqM3CeMejtHEh6jsSVUF3zefyjWO0nf5ZZh_ROg9syo2q0EZ7c4kObKqtIZQ8XOdP62icz95-CEw5YS_G-cvChLCPCcIXklGIw0625V2PL4moJYXwPvOu92fShebOyLw.jpg" alt="" loading="lazy"> <p>Haha, well yeah, it should’ve been for 2000 USD. Healthy economics for a healthy startup.</p> <p><a href="https://t.me/chillhousetech/969?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I liked this vid with Nvidia’s CEO answering about his leadership team of top… https://darkolorin.com/posts/968.html https://darkolorin.com/posts/968.html <p>I liked this vid with Nvidia’s CEO answering about his leadership team of top execs, which is 60 people (lol). And the explanation is reasonable. The main thing is to shrink the depth of the hierarchy and do fewer 1-on-1s, so there aren’t privileges around access to information, which creates an imbalance in power and management. And yeah, with 60 people you really can’t do that many 1-on-1s. And he gives feedback in front of everyone at all-hands meetings. Short vid. <br> But honestly all this only says one thing: processes and hierarchy are secondary, and what matters is results. And how you get to them is already a private matter.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/968?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Happy upcoming to everyone! https://darkolorin.com/posts/966.html https://darkolorin.com/posts/966.html Tue, 31 Dec 2024 18:41:19 GMT <p>Happy upcoming to everyone! And the main thing — health, and for next year to always be better than the previous one!</p> <p><a href="https://t.me/chillhousetech/966?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “In science there is real knowledge and knowledge that creates spiritual… https://darkolorin.com/posts/965.html https://darkolorin.com/posts/965.html Mon, 30 Dec 2024 16:39:57 GMT <p>“In science there is real knowledge and knowledge that creates spiritual comfort;”<br> The Voice of the Lord<br> Stanisław Lem</p> <p>I think exactly the same about courses, advice, advisors and all that other crap.<br> I’ve identified a certain number of fairly abstract signs of “useful interaction” with a course, a book, a person in the professional field, that increase the probability of success in what you’re doing right now.</p> <p>1/ help takes time, just like in therapy. You won’t get cured in one session, same as you won’t “get” how to make a million or learn how to build sales in a foreign market in one go. And usually you need a lot of time (any practices are hard to переносить given the difference in mindset and team and the current process).</p> <p>2/ knowledge is pain and humiliation, especially after 30. Because nobody wants to study (usually it’s a substitution effect, like: if you watched a short dopamine video then, like, if the info there is interesting then you “learned”). In general I think adult education only works in certain, pretty narrow cases. So learning should be unpleasant and hard and мучительно. And even, at times, useless in the short run.</p> <p>3/ studying other people’s cases is just a pile of general outlook and that’s it. The value of an MBA is in the network of smart people; the value of an online MBA is unclear to me. What’s more, a lot of cases are so tightly tied to the time when it happened (people, culture, historical context, market state, and hundreds of other factors) that I can’t help but understand that the use of that knowledge now is like knowledge of the daily life of a medieval village. So when you’re studying someone else’s experience, try to answer yourself as quickly as possible: can I apply this very fast without serious changes in processes or product.</p> <p>4/ trust, but verify! Ask what specific achievements in numbers your coach/advisor etc. has. TAKE REFERENCES on the person and look for them on the market yourself (usually the coach gives you 2–3 people).</p> <p>5/ and finally — rest more, it’s the most effective mood improver; don’t want to — don’t do it. Working with a calendar jammed 24/7 with useless meetings loads the brain so much there won’t be time left to think, and without “thinking” it’s unlikely anything adequate will come out.</p> <p><a href="https://t.me/chillhousetech/965?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> New Year’s AMA, the week was mostly for vegging out, after all in England… https://darkolorin.com/posts/964.html https://darkolorin.com/posts/964.html Mon, 30 Dec 2024 13:04:26 GMT <p>New Year’s AMA, the week was mostly for vegging out, after all in England Christmas is the main holiday so I decided to take a break too. <br> Anyway here, grab the questionnaire (<a href="https://forms.gle/uX2bH7W8NmF9d48w6?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://forms.gle/uX2bH7W8NmF9d48w6</a>), what would be interesting or just leave your wishes <br> 🙂 <br> And if anyone wants to check how beefy that o3 little model from OpenAI is, you can play those very tasks (<a href="https://arcprize.org/play?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://arcprize.org/play</a>), the ones it’s “called a breakthrough in causal relationships” on.</p> <p><a href="https://t.me/chillhousetech/964?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> AI will kill “insert profession name” or that super popular Reddit post about… https://darkolorin.com/posts/963.html https://darkolorin.com/posts/963.html Mon, 23 Dec 2024 13:28:11 GMT <p>AI will kill “insert profession name” or that super popular Reddit post about the new Coca Cola ad that they made with AI and the marketers chopped themselves a yearly bonus <br> <a href="https://www.reddit.com/r/mildlyinfuriating/s/wcPdBrjuNN?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reddit.com/r/mildlyinfuriating/s/wcPdBrjuNN</a></p> <p><a href="https://t.me/chillhousetech/963?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Section “read, but not fully understood”: 1/ About the neural net that… https://darkolorin.com/posts/962.html https://darkolorin.com/posts/962.html Sun, 22 Dec 2024 01:28:38 GMT <p>Section “read, but not fully understood”:<br> 1/ <br> About the neural net (<a href="https://github.com/official-stockfish/nnue-pytorch/blob/master/docs/nnue.md?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/official-stockfish/nnue-pytorch/blob/master/docs/nnue.md</a>) <br> that underlies the famous chess bot Stockfish. The gimmick of this architecture is that it tries to change its internal states and inputs as little as possible while it’s calculating your next moves. A respectable thing. <br> 2/ Boethius, “The Consolation of Philosophy.” A book, basically. This is for understanding how people thought in the Middle Ages. The guy was so respected that his book got recopied in almost every monastery in Europe. And the dialogue format in the book is interesting too. Wanna get what was going on in rulers’ heads back then, read it without hesitation. <br> 3/ <br> A fairly old article (<a href="https://nplus1.ru/material/2019/12/22/simple-math?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://nplus1.ru/material/2019/12/22/simple-math</a>) <br> on n+1 about some “levels” in physics, like how complicated everything is, and that meme Elon Musk once posted. One of my favorite articles, since I dug into string theory myself (a few years, just for me). <br> 4/ Seems I accidentally unearthed where the term JTBD (job to be done) came from. In the sixties some Theodore Levitt, who worked at Harvard as an economist and a teacher, put out a theory called “marketing myopia.” Put simply, he pushed the idea that a bunch of companies are being dumb defining themselves by their products rather than by what needs they cover (now they’ve taken that topic and repackaged it into the “job to be done” framework). <br> 5/ <br> a pretty neat playbook (<a href="https://caseyaccidental.com/monetization-playbook/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://caseyaccidental.com/monetization-playbook/</a>) <br> on monetization from a guy from Eventbrite. Liked how simply and lightly the decisions are built off basic metrics like virality, retention, and so on. <br> 6/ for anyone who wants to figure out quantum computing <br> I recommend (<a href="https://quantum.country/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://quantum.country/</a>)</p> <p><a href="https://t.me/chillhousetech/962?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Noticed this thing. https://darkolorin.com/posts/961.html https://darkolorin.com/posts/961.html Sat, 21 Dec 2024 17:08:43 GMT <p>Noticed this thing. Often some head of AI is looking for a senior product manager for their team. Like, to join the innovations. I read it like this: so basically there are a couple ideas here, anyone wanna write tasks in Jira, I’ll be dictating on calls? <br> But gotta have ML/AI experience, of course!</p> <p><a href="https://t.me/chillhousetech/961?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> My two cents into the world of great AGI predictors and OpenAI worshippers. https://darkolorin.com/posts/960.html https://darkolorin.com/posts/960.html Fri, 20 Dec 2024 21:49:34 GMT <p>My two cents into the world of great AGI predictors and OpenAI worshippers. <br> I really liked today’s release of a very expensive smart brute-force. Or if you want, an alphaGO analog, but for a certain type of tasks (they’re tuning it for math and logic, seems like with geometry). To better understand where that price per run comes from (solving what is basically a kid’s problem costs $1500, but for serious-level math it’s not all that super cool there, only 25% on the bench, keep in mind it’s a very serious benchmark and demands insanely strong knowledge from a human — I personally had enough for like one and a half problems max), the best analogy is to understand how that architecture that beat everyone at Go worked. And inside it there was the so-called MCTS (Monte Carlo tree search). And a couple nets that generated answers for a position and then this smart search ran over them. So my bet is that the current o3 version is exactly that. Some kind of smart search algorithm plus a little net (most likely not even huge, like ~70B parameters). And those thousands of rollouts in a specific environment cost that much. <br> An even simpler analogy is brute-force, but not straight up, more like with hints. Like the net hints what definitely can’t be there or what definitely has to be there. This cuts the number of variants hard, and that’s why the process at least converges. I think the future is nets plus a few external algorithms like MCTS or Program Synthesis. As a time limiter they use the test time compute idea. <br> By the way, one of the alphaGO authors, Noam Brown, moved to OpenAI about a year ago — well, here’s the result. <br> As for AGI, hell knows what that even specifically is, but like actual mind and all that, it makes new decisions that were previously unthinkable — that’s still a ways off. But not like totally impossible. 10–15 years.</p> <p><a href="https://t.me/chillhousetech/960?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, the new o3 model is delivering some very slick chains of reasoning and… https://darkolorin.com/posts/959.html https://darkolorin.com/posts/959.html Fri, 20 Dec 2024 18:24:45 GMT <p><a href="https://arcprize.org/blog/oai-o3-pub-breakthrough?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://arcprize.org/blog/oai-o3-pub-breakthrough</a> <br> Well, the new o3 model is delivering some very slick chains of reasoning and even solves the benchmark just fine for those who doubted. Long live test time compute. Waiting for the release at the beginning of next year. Or later :) <br> Oh yeah, almost forgot: a single run can cost a couple grand bucks</p> <p><a href="https://t.me/chillhousetech/959?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The picture shows a popularity chart for MidJourney. https://darkolorin.com/posts/958.html https://darkolorin.com/posts/958.html Tue, 17 Dec 2024 16:00:11 GMT <img src="https://cdn4.telesco.pe/file/GPGOc-yrndybCwuvJaswXNwHCaoSUzuBGefc6tCNERaWsbqxjLPQUJeTDEryimpRzuvq26x5-PmM6ljHToZB0G-G_SQsI9aeIO8NqUM665SMYUpWs1jEH550v35eAqlO0YyeQKvuset4zJo8qjKB_suBBNrTgjejJSQpUzSTDpMpFubgqZPUoMFCC5JMTivLFOq5Oh8sn0DkyRTZdGDp-dpEZEu2sbBGUasr-rNtHXmZjUIzX5tKt5k7dwCNORPBdOuCy69vmH83qVBr6aftTpTaJK42pJuy5Ey17DebyHoTKPqK5OqxCqiNuVw6qcx-gpKg3FbPNCkRqO0ogkvUnQ.jpg" alt="" loading="lazy"> <p>The picture shows a popularity chart for MidJourney. This is all about generating images and other graphic stuff. A relative chart, but an important one. It shows really well how much speed and the moment matter in the product life cycle on the internet. How easy it is to become a leader and just as easily lose the leading branch.</p> <p>Nothing is as important as the moment and timing when you’re building your product or startup.</p> <p>Specifically, the guys managed to grow, use relevant community tools, get super important data on preferences so they could iterate fast on the model, BUT also fall behind on shipping a big web product, hiring more people for faster parallel iterations on the model and the product, and losing to the big guys. The founder, nevertheless, claims the project didn’t take venture funding (which of course can be read as “we took money, just not from funds, or we won’t tell you”).</p> <p><a href="https://t.me/chillhousetech/958?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Domestic so-called image generators. https://darkolorin.com/posts/955.html https://darkolorin.com/posts/955.html Sun, 15 Dec 2024 18:57:03 GMT <img src="https://cdn4.telesco.pe/file/A-fODbNPiD5R6-qXMyrkWvUFiLsQdj5CADWDAAWr4SyepUvjTK-PkkwEc0j0MGhXz6fiE4fDpU8mURm6JROL619yOmCPSUtsKz22kwm4HjU0gNcDEMV3Lysu0aU6DbLE_wtrJ1v4G36G2A29w9TIMhy3-9P9P_TSvm_euPmhERT_UL5E0qMhClEqQQmeyPmHfzS0Sujihv_ohGyS_u1Ad0v5fz5gNuxvSiuCzYjLIrgoqTD-l7grKtWzBOVDG--ymClLNFYBJfTf4g_Z692lHISjCEl8FFoQM_g-mvAvcIRI3eIuMbCsFEpFYkdjzQOxvkgdf-ojADaL-emxVa9lzQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/X8Ik7t3yibUSm7v6KRG5SBnQ4YNY49KiyxneGndGPxgkDXELsibK11pvkGzToZOVDYfLpHF-6wHVnbYc6ZetbwJYDDq_zDWRae3Qv9eLrFXI5xBe0I3JPSKiM65kprzyC6u4BG32d9JOm3HoRZXFwxgSvc6vCK_KdFJwGQtKFFs3vlJcgMawW7Q0nLDMPz8w_csVYbxex6ONPb1SPLyRAkwxxl6Z-h0osFnpLqRJyOpPa2esydFRedS_32RZXFITsEXmy-awOkh9Y1kPe9hdwWZzfz5ZGq_PgNBaWzYGxMSx3hDpGSsqTMEUz8uPTGZs8utgEHN3D2UGnCLoEUzEsg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/R7MBtI8jKy7B6U-1Y_0ddMrO905Setzikm0B8lUoGNhk0LyCsM8E77TGSQUJtDZvJazwZKxK6BwtapBMV11q0-N2suLHzuizncXbriRcCIx4Gt-7qyVloBVCvCCp-kgXZhA4dm6QPFuUcD1vY3IusvIY5l213ZaVTdclb2NGtdI6DNS55BAQMDl_DdQH47pvEgC5sobDk3n1pnc7NAXqXa3iKoSKOeqHljBNgmIotXBlJJM2e843Zss-S1SWvKcjbhqimAWCjnqL6x6wqyyZ7Z53ugSZ_-0-c4y_tWOo4BA8aH-4b21WxzESdEqY-sBSY0_Vo3DHC3lKSexvKbV0Pw.jpg" alt="" loading="lazy"> <p>Domestic so-called image generators. <br> 😑</p> <p><a href="https://t.me/chillhousetech/955?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A quick minute of self-promo, by the way, not regular promo. https://darkolorin.com/posts/954.html https://darkolorin.com/posts/954.html Sun, 15 Dec 2024 17:33:41 GMT <p>A quick minute of self-promo, by the way, not regular promo. Me and Anya used to work at mail.ru back at the dawn of our careers. And somehow we keep talking on her channel from time to time.</p> <p>Besides the fact that I’m not saying anything supernatural, my content is often pretty down-to-earth. But still, for Sunday as background noise it might go down fine.</p> <p>Tap tap (<a href="https://youtube.com/live/1d7hRZrJTkM?feature=share&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtube.com/live/1d7hRZrJTkM?feature=share</a>)</p> <p><a href="https://t.me/chillhousetech/954?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Noticed how all the AI app builders (well, all these prompt “experts”) like… https://darkolorin.com/posts/953.html https://darkolorin.com/posts/953.html Tue, 10 Dec 2024 23:43:49 GMT <p>Noticed how all the AI app builders (well, all these prompt “experts”) like horoscope, coach and the rest suddenly started building agents. Like, we’ll help you create what we ourselves couldn’t sell to anyone. Make your own agent that will fill out Excel (btw you can do that without AI too, just write some little formulas there). <br> Just an observation.</p> <p><a href="https://t.me/chillhousetech/953?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Google made some progress on quantum processors there, and while half the guys… https://darkolorin.com/posts/952.html https://darkolorin.com/posts/952.html Tue, 10 Dec 2024 15:32:05 GMT <p>Google made some progress on quantum processors there, and while half the guys just post you a headline like “look this thing computes what a classical computer would take a fuckton of years to compute,” I always advise you to <br> read Reddit (<a href="https://www.reddit.com/r/singularity/comments/1halamr/meet_willow_our_stateoftheart_quantum_chip_google/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reddit.com/r/singularity/comments/1halamr/meet_willow_our_stateoftheart_quantum_chip_google/</a>)<br> or <br> the paper (<a href="https://arxiv.org/pdf/2408.13687" rel="noopener">https://arxiv.org/pdf/2408.13687</a>)<br> and realize the task for the headline was picked on purpose so a quantum chip solves it fast, and a classical one slowly.<br> But overall <br> the progress is tangible (<a href="https://blog.google/technology/research/google-willow-quantum-chip/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.google/technology/research/google-willow-quantum-chip/</a>)<br> , I figure another 10 years until normal applications. Waiting.</p> <p><a href="https://t.me/chillhousetech/952?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sometimes I’ll post more short facts like this, a la TIL (today I learned). https://darkolorin.com/posts/951.html https://darkolorin.com/posts/951.html Mon, 09 Dec 2024 20:36:39 GMT <p>Sometimes I’ll post more short facts like this, a la TIL (today I learned). <br> Today I learned where we are on a really huge scale, so to speak. <br> Laniakea (<a href="https://ru.wikipedia.org/wiki/%D0%9B%D0%B0%D0%BD%D0%B8%D0%B0%D0%BA%D0%B5%D1%8F?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ru.wikipedia.org/wiki/%D0%9B%D0%B0%D0%BD%D0%B8%D0%B0%D0%BA%D0%B5%D1%8F</a>) <br> from Hawaiian (yeah that’s a thing too) — “immeasurable heavens.” Just a pretty name.</p> <p><a href="https://t.me/chillhousetech/951?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> What big funds and YC are looking for. https://darkolorin.com/posts/949.html https://darkolorin.com/posts/949.html Mon, 09 Dec 2024 20:30:09 GMT <img src="https://cdn4.telesco.pe/file/p14l2-5RzFAzW0paAlM3xAWtCvN8BRq2qcbYFM7nS8kB13aLJdlZThdsdDCBEUATyrP_YLSHKRzg3NordMfk3Qjf4qqRV8Hqiglf63154MfHDNlrx6PPjc8KZnN7B8JH_oFnthxyqMPF3QvOjzbiIEjsI5u3YwloDDKM_tTfY9PVBEN3Wt_yadgYiIEgC5ow__H1NwShpd570Otb8KLm8d56W64fpYEgQdkFv1imIOIvqQdmLOqV-CAE-YU3OYGrDjmOa-Z1HIw3jBvK0LWWi8W4XeKdpqTRPpbNfOsYO7184KvPHNX2YoMg5TgyU5zm9ZAOC5epIs-XwmDiexcysg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/R1Vnt3BpYBMfhdp4OzXZJPbICJnOd_Mmp0T9xZos0dbwIpfG_cJHpUIP5u-wuf263nEfPQuWfJXDc5nhUtbtrTQutNjo-R-RQ6veXsQdvt27pix9DltVgoXYOXg6ak74LvC-CXB8iGXMWlV_wpgvZZRKuXS2zpz5_1hGFal5VBTWUH2Mu-8W4kZvOkfGgDPhYsePNacMeC3tSjo4L0xPCbQXbUBcdxPTiZX_8iPEZ1zFIvQAlMwokpV_wnJuppLNLwCT0ePFvk7PCGARM2CsRnuA5vjZDS8FtQNYdVqJHDT2AXVk2YqLapX2WlXaT70ggakffBhaQWUhKPKHPHlG5Q.jpg" alt="" loading="lazy"> <p>What big funds and YC are looking for. A short-ish summary.</p> <p><a href="https://t.me/chillhousetech/949?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In one line — a new llama 3.3 dropped 70B has metrics like 3.2 405B. https://darkolorin.com/posts/948.html https://darkolorin.com/posts/948.html Fri, 06 Dec 2024 16:33:42 GMT <img src="https://cdn4.telesco.pe/file/eq--qYkV65qglNwUGR3PMzLD3W_7D1-WEs18MIVkRTp11GMco2n3O-MQXCi5nlqNMxL3gYB4M-4vBJHKAkCI6ZDoCQquia5jg6L7jknuy2Zn8-tTuK437HQVG8AjF3QBHPZeyREvMZWux9RBo7J-whFI8yzAx8nWX6j5FiFKM0pGlgzdtSTunml39RhTHR4Yw1ZCJbHXGSXiC_CLgVSbuR5lDVDWdE71wOH5j1a3QMx96TihOZAFaAO6wqSTjxcI3dvmNi2gl8n4MB5zxeqsD9FltkAYKIa1YT02pmm6DiUseCZnPY9YvRZ85BsOVgxwwlCFmNNUCkp9KA19zinx9Q.jpg" alt="" loading="lazy"> <p>In one line —<br> a new llama 3.3 dropped (<a href="https://www.llama.com/llama-downloads/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.llama.com/llama-downloads/</a>)<br> 70B has metrics like 3.2 405B. Meta’s moving along just fine.</p> <p><a href="https://t.me/chillhousetech/948?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Do you/does your business need a media strategy or a presence in the media… https://darkolorin.com/posts/947.html https://darkolorin.com/posts/947.html Fri, 06 Dec 2024 15:27:07 GMT <p>&quot;&quot;&quot;<br> Do you/does your business need a media strategy or a presence in the media space?</p> <p>- Of course! I regularly put effort into it<br> - No, it’s never once paid off<br> - 50/50, but the importance is clearly overrated<br> - No idea, never dealt with it<br> - 😑<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/947?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, I also can’t stand the cult of suffering. https://darkolorin.com/posts/946.html https://darkolorin.com/posts/946.html Thu, 05 Dec 2024 15:54:27 GMT <p>By the way, I also can’t stand the cult of suffering. The constant whining about how hard it is, like, to be in a startup, to be in a corporation, how сложно it is to write code when your manager is an idiot or everyone around you is an idiot. Yeah, you should run from people with that kind of message and that whole narrative in general. The problems are usually because of them. You can’t even see the log in your own eye.</p> <p><a href="https://t.me/chillhousetech/946?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The now-hit column “Read, but not fully understood.” 1/ A paper from Duolingo… https://darkolorin.com/posts/945.html https://darkolorin.com/posts/945.html Thu, 05 Dec 2024 14:21:03 GMT <p>The now-hit column “Read, but not fully understood.”<br> 1/ <br> A paper from Duolingo (<a href="https://research.duolingo.com/papers/yancey.kdd20.pdf" rel="noopener">https://research.duolingo.com/papers/yancey.kdd20.pdf</a>), about how the guys built an algorithm for push notifications to keep engagement up and all that (bandits and that sort of thing).<br> 2/ <br> Some old course on ads (<a href="https://minqizhou.wordpress.com/wp-content/uploads/2011/11/lecture-03.pdf" rel="noopener">https://minqizhou.wordpress.com/wp-content/uploads/2011/11/lecture-03.pdf</a>) and the models used there (obviously old, but I like how old courses lay out the basics, skimmed it, seems nothing too hard).<br> 3/ <br> A pretty fundamental work (<a href="https://www.cs.cmu.edu/~sandholm/cs15-892F15/AuctionMechanismOptimallyTradingAMEC03.pdf" rel="noopener">https://www.cs.cmu.edu/~sandholm/cs15-892F15/AuctionMechanismOptimallyTradingAMEC03.pdf</a>) about the point of auction algorithms, or what you pick: money or efficiency (in short, clickbait or actual benefit from the product/target action). Highly recommend for understanding how advertising works in general and where the problems are.<br> 4/ <br> On how to calculate your costs (<a href="https://blog.eleuther.ai/transformer-math/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site#activations-and-batch-size" rel="noopener">https://blog.eleuther.ai/transformer-math/#activations-and-batch-size</a>) when you’re about to host/train your own models.<br> 5/ <br> Something by Kurt Gödel (<a href="https://journals.aps.org/rmp/pdf/10.1103/RevModPhys.21.447" rel="noopener">https://journals.aps.org/rmp/pdf/10.1103/RevModPhys.21.447</a>) on solving the equations of general relativity and the early works on the structure of the universe and time in it, why — I don’t know, but they say the work spawned an entire movement in this area, which one — also won’t say <br> 🙂<br> 6/ <br> An old blog with problems (<a href="https://susy.page/2024/07/01/expectation_of_random_summands_number?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://susy.page/2024/07/01/expectation_of_random_summands_number</a>), this time a popular one: On average, how many random natural numbers, uniformly distributed on the segment from 1 to n, do you need to take for their sum to exceed n?<br> 7/ For now I’m not doing anything, just reading a book about the history of Blizzard (<a href="https://www.amazon.co.uk/Play-Nice-Future-Blizzard-Entertainment/dp/1538725428?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.amazon.co.uk/Play-Nice-Future-Blizzard-Entertainment/dp/1538725428</a>) (game people be like), and you can straight-up feel the stench of trash drama there.<br> 8/ <br> And the classic of French “thinkers” on law (<a href="https://libertarium.ru/lib_law.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://libertarium.ru/lib_law.html</a>), I think it’s one of the strongest works, and there’s a solid book by the same author Frédéric Bastiat too.</p> <p><a href="https://t.me/chillhousetech/945?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, what are your bets: “the new Google” or just another Theranos? https://darkolorin.com/posts/944.html https://darkolorin.com/posts/944.html Wed, 04 Dec 2024 18:36:12 GMT <img src="https://cdn4.telesco.pe/file/grrke6gaMyhHGZf5Gg0VRyCnz2hywdYOk7Q7C8JnMue2_G9N7KSlm0F17-KMTBFi-IcG7LzdhXPJjQMVpmIekJGI2hfI9oZkHsKWLQ6frx6x4ivbWVontu7Epcz4PBN5uUgaZHVq9UCxBC3UvO5p7_hl0dmugxeAAA-mp75BvvJvq-qizmh_fK6Dt9GqjKPMovedUY6fpBu4RpkF9EgHIMQiBwMAIFRWEfc1i29I4mCiIW6qr70lD0ZqqutzFr01tDFHfpyobZwbdqg5m9jX0gGlUvBzw_h-RW9oSMFvbUDT7oGDJ8sA56q9irfH1Li9dCMuf-dteFMXCgd2ULustw.jpg" alt="" loading="lazy"> <p>So, what are your bets: “the new Google” or just another Theranos? Either I’m not getting something, but something’s off here. What do you think?</p> <p><a href="https://t.me/chillhousetech/944?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Be honest. https://darkolorin.com/posts/943.html https://darkolorin.com/posts/943.html Tue, 03 Dec 2024 17:28:43 GMT <p>&quot;&quot;&quot;<br> Be honest. Do you think all these mega-campuses where entrepreneurs come out and tell the story of how they wore torn socks and became successful are motivating? Or is it just a regular form of dopamine?</p> <p>- It’s like TikTok, just paid and VIP<br> - Yeah, I believe it can motivate<br> - No, total bullshit<br> - 😱<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/943?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Just like that, the guy showed what everyone spends hours doing on TikTok and… https://darkolorin.com/posts/942.html https://darkolorin.com/posts/942.html <p>Just like that, the guy showed what everyone spends hours doing on TikTok and so on. Waiting for new products whose features you can understand just like that. Anyone can—no philosophy, just understand and try it.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/942?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Yandex https://darkolorin.com/posts/941.html https://darkolorin.com/posts/941.html Mon, 02 Dec 2024 12:33:22 GMT <p>Yandex</p> <p>Nebius is raising (<a href="https://techcrunch.com/2024/12/02/european-ai-infrastructure-company-nebius-nabs-700m-from-nvidia-accel-others/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://techcrunch.com/2024/12/02/european-ai-infrastructure-company-nebius-nabs-700m-from-nvidia-accel-others/</a>)<br> 700m because they need to get into the US. But overall my expectation is that this is a kinda meh bet from the guys (basically I think it’s a loan).<br> Can only wish them good luck competing with AWS/Azure etc</p> <p><a href="https://t.me/chillhousetech/941?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sunday. https://darkolorin.com/posts/940.html https://darkolorin.com/posts/940.html Mon, 02 Dec 2024 00:44:10 GMT <img src="https://cdn4.telesco.pe/file/b8CBsJJjpMdG3SgjT2d_xFytv7YzEnGYbjCLW7EkENHCe44agS2gR5_1DsryjnnScY6JqkVNWXo85ZMahaM6kgH851zTYSg26_IihQ0_bgElhBngLWljcZrxSUFHboAD5ebCWcWwsSJSP-nDjGkg7UjgZ0tFuSrVEbWJdJ_H4Us5K0S_plfdxSobj1fp-UX8j6G-vvBXhVQTB25kNECE-R3vtmw7E-ysf8dP0FpM1CSJpuQYhfMWNpqCHQj1Yz5EAnjyQfP8d5T8WA9qQLBIkkuhTAO03GKwcquaqRhk5p1gCwJOM8VtDN2xuuzrbaghOLcZpo3sfw68it0Az66EvA.jpg" alt="" loading="lazy"> <p>Sunday. On-theme. In the “best prompt” category, this guy wins.</p> <p><a href="https://t.me/chillhousetech/940?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Thought I’d get back into the groove, but this week I ended up in the hospital… https://darkolorin.com/posts/939.html https://darkolorin.com/posts/939.html Sat, 30 Nov 2024 14:44:03 GMT <p>Thought I’d get back into the groove, but this week I ended up in the hospital by pure dumb chance. Anyway, alive and kicking.</p> <p>Now to the point. There’s this thread (<a href="https://x.com/deedydas/status/1862138506774213061?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/deedydas/status/1862138506774213061?s=46</a>) about motivating engineers in corporations that popped up from one of my favorite investors these days.</p> <p>I’ll allow myself a loose translation plus my two cents:</p> <p>1/ there’s this myth or misunderstanding like engineers (read: programmers) are lazy and basically do nothing and don’t want to. Not quite. More like in modern corporations the incentives are built like a 1950s factory.</p> <p>2/ meaning there are zero rewards for simplifying or deleting unnecessary parts of code/process. If you did something fast — that means the task was easy. Most tasks don’t make any meaningful changes to the product, they just move KPIs forward by a couple percentage points. Remember all those LinkedIns “increased conversion by 1.5%”, and even if the business won big, the task is still boring because you’re probably sawing together a framework for A/B tests.</p> <p>3/ don’t like your current job — switch and find a new one, a competent engineer can afford that, so the risk of getting fired is low. Innovation is swapping out middle managers, who get rotated so they don’t rock the boat.</p> <p>4/ almost all the shares in your option grant or bonuses tied to the company-wide KPI have nothing to do with your work. Your code very rarely affects the company’s stock price growth. You feel like a significant part of your income (and sometimes the bigger part) doesn’t relate to you at all.</p> <p>5/ turnover is weak; if people are weak no one will touch them because hiring is more expensive than keeping mediocre people. Shareholders rarely demand firing anyone, unless it’s really completely hopeless, but then they just cut to prop up the numbers, not to fix the quality situation.</p> <p>6/ all review systems are total trash. The evaluation system gets hacked through friendships, doing only “important” tasks, etc. Goodhart’s law. Any productivity metrics can be hacked and they will be.</p> <p>7/ all these calibrations flatten all teams into one big stream where there are no best and worst. No demotions, meaning a promotion gives you a sense of safety no matter what.</p> <p>Pretty sure this is how it is almost everywhere with 10k+ people, maybe even less. In my experience even smaller companies have all these traits. When you start measuring intellectual work it turns out fixing 2 bugs a week is more “effective” than inventing a new product.</p> <p>TLDR:<br> The only way to seriously increase income is to get promoted. Promo packets and politics matter more than building a quality product.</p> <p><a href="https://t.me/chillhousetech/939?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Getting back to life after a week of birthday. https://darkolorin.com/posts/938.html https://darkolorin.com/posts/938.html Mon, 25 Nov 2024 19:52:30 GMT <p>Getting back to life after a week of birthday. So yeah it’s interesting how things get done and as it seems to me Anthropic little by little is doing more for our beloved AI than OpenAI, namely they’re starting little by little to roll out standards and make them open (Open AI hahaha), this time a protocol for how to feed your product to agents based on LLM.<br> <a href="https://www.anthropic.com/news/model-context-protocol?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/news/model-context-protocol</a></p> <p><a href="https://t.me/chillhousetech/938?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And as for the previous point. https://darkolorin.com/posts/937.html https://darkolorin.com/posts/937.html Wed, 13 Nov 2024 17:51:35 GMT <p>And as for the previous point. <br> Quote: “Ilya Sutskever, co-founder of AI labs Safe Superintelligence (SSI) and OpenAI, told Reuters recently that results from scaling up pre-training — the phase of training an AI model that uses a vast amount of unlabeled data to understand language patterns and structures — have plateaued.” So yeah, finally they’ve started admitting how things actually are. And I think Ilya understands damn well that we need other methods and algorithms for the next step. Transformers have pretty much shown what they’re capable of in full.</p> <p><a href="https://t.me/chillhousetech/937?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One of the most, in my view, significant papers just dropped. https://darkolorin.com/posts/936.html https://darkolorin.com/posts/936.html Wed, 13 Nov 2024 02:27:37 GMT <p>One of the most, in my view, significant papers just dropped. It pretty convincingly shows that we’re hitting the limits of quantization. The paper says: the more tokens you use for training, the more precision you need. Or do you remember all those bitnets (where all the weights are just three digits 1 0 -1) and other stuff (which seems like a total refutation of this idea).</p> <p>Important:<br> - almost always we’ve gotten better performance and all that by reducing dimensionality (even if it’s just the little numbers)<br> - in other words the slowdown is here and it’s real, looks like we need to learn how to improve what we already have and optimize for narrower use cases</p> <p>Anyone who’s seriously into AI should absolutely read it <br> <a href="https://arxiv.org/pdf/2411.04330" rel="noopener">https://arxiv.org/pdf/2411.04330</a></p> <p><a href="https://t.me/chillhousetech/936?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Still agree that unit economics is about way, way down the road, and the team… https://darkolorin.com/posts/935.html https://darkolorin.com/posts/935.html Mon, 11 Nov 2024 21:25:30 GMT <img src="https://cdn4.telesco.pe/file/Y41U7cJ5YPXpy2bSPm6DD-7PvVwybO5qr3TZpQi_enMgy9faeCo-LttFZDbhhF4jM_eazlz3gum1VBVJ2m3AQfbaBf9h8XuI45lftxzFbAMkFJhxkTLUrsHuAY-kttB4eFKkib9_LCeJhy-8ak0M2Q8OYA4sIm0D_0vQnzoQxEeiyr7KgJ7L-zBRFw0z6laeu02OLFsMXalZNh9U7pXdLm6-fGXYXDaLuQGviJr3dgI_HvDzAwLS0Pq6H66HrklK2sPJ_2Cl6Hk0RkZLn3HP3e5PhzP89n-aWyOrkNCVgmm-3D9Z95II4Pb5IOD5ixTF_IEJgdRKTd9WW6lljY9k5A.jpg" alt="" loading="lazy"> <p>Still agree that unit economics is about way, way down the road, and the team and product are about the very beginning. Proven by talking to a large number of different funds and decision makers. <br> P.S. probably worth mentioning this is specifically about venture projects (== the ones that need funding for fast growth and hitting results)</p> <p><a href="https://t.me/chillhousetech/935?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Not everyone might be aware, but there was a time I played pro football for… https://darkolorin.com/posts/934.html https://darkolorin.com/posts/934.html Mon, 11 Nov 2024 13:36:30 GMT <p>Not everyone might be aware, but there was a time I played pro football for what was then Smena, which later became the Zenit Saint Petersburg academy. So I care about sports and it seems like there are a ton of problems there. Especially with scouting talent and the process of how a player goes from youth teams to the adult level. Who’s watched that series “Out of Play”? Yeah, I generally believe that with tech you can seriously improve the process in football schools. There’s a fuckton of corruption there too.</p> <p>For example, I’ve been following JuniStat — the guys taught neural nets to accurately calculate footballers’ data using a phone and collected metrics from thousands of players aged 8 to 18 worldwide. Clubs and federations use this system to track data on promising players through tests validated to FIFA standards.</p> <p>Back in my day that kind of tech wasn’t available — it’s hard to imagine how many kids could’ve shown themselves and gotten on radars just off their data.</p> <p>And for those who aren’t indifferent to sports and all that, here’s the CEO’s channel, and yeah it’s not an ad. One of the few projects you actually want to support, and I’m not even an investor.<br> @junistatsystem (<a href="https://t.me/junistatsystem?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/junistatsystem</a>)</p> <p><a href="https://t.me/chillhousetech/934?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sunday stuff. https://darkolorin.com/posts/933.html https://darkolorin.com/posts/933.html Sun, 10 Nov 2024 20:55:39 GMT <p>Sunday stuff. Or, on the benefits of online games. <br> Maybe you’ve heard, maybe not, but I play a lot of online games, and besides the usual competitive spirit and all that, there’s one thing I really like. <br> Namely, we’re going to talk about socialization. I’m generally a pretty closed-off person, and by 33 my circle is strictly professional plus close people. That, overall, limits your understanding of what’s going on and how for different groups of people and the population. In online games, though, you often group up and go kill bosses together, or you play team vs team. Pretty often you’re sitting in voice chat like Discord. And that’s where you actually meet people. And sometimes it’s insanely interesting to listen to who does what (for example, I’ve had long-distance train attendants, oil rig workers, and so on), and it really sobers you up and gives you at least some kind of view different from your bubble. Personally I don’t have that many chances to regularly hear different stories like that even if I want to. That’s how it is. <br> So how do you get out of your social bubble?</p> <p><a href="https://t.me/chillhousetech/933?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well finally all the fans got served up another benchmark that’s pretty solid… https://darkolorin.com/posts/932.html https://darkolorin.com/posts/932.html Sun, 10 Nov 2024 17:20:12 GMT <p>Well finally all the fans got served up another benchmark that’s pretty solid in my book (<a href="https://epochai.org/frontiermath?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://epochai.org/frontiermath</a>) — this time in math. And there basically nobody solves it, and btw there’s no difference between o1 &amp; gpt4o either. The endless hype around LLMs I think will die down hard, because the current progress just trivially depends on labeling and preference data (one option is better than the other). So I’m expecting the top end to slow down and small models to grow fast in quality. Since they’ll be trained on higher-quality data.</p> <p><a href="https://t.me/chillhousetech/932?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> What’s everyone got in ChatGPT for that prompt: based on what you know about me. https://darkolorin.com/posts/931.html https://darkolorin.com/posts/931.html Sat, 09 Nov 2024 01:52:01 GMT <img src="https://cdn4.telesco.pe/file/JTmjWFQkKVfiGA0zg124B-mLTllW9_-IH3u1YTKlJj0sBpctC7AMpcYB81MFUgPpSwiaRVwO4i2T9_5fUUn1mf3lRKhE5ROudeCJ_kH-G7HadX0NeaDYrSqR3phCQsIXgOSZf5arib1R1RoPMaUZXXyYqyFH8hu3k2lpJxWnJe_ZEHAhONz74YdjL6bU3Y1DHWarM_GYAmqPk_F6u9lQERZN6ziAdypmVj3obBMin12QRQ8KO4iy5OWKlXGmady1dqeq6grVzFncNx2fqTwa7ijCEHNqKuOOjzsuoH7RRVycfqkSI1OOCO7zC6NwGD8UfcX4RYZo132zl9G6PvPEbQ.jpg" alt="" loading="lazy"> <p>What’s everyone got in ChatGPT for that prompt:<br> based on what you know about me. draw a picture of what you think my current life looks like<br> Send me your results :) I apparently asked a lot about wow, because that’s how I check search relevance</p> <p><a href="https://t.me/chillhousetech/931?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I once compared modern tech to the industry of the 1930s–40s. https://darkolorin.com/posts/930.html https://darkolorin.com/posts/930.html Fri, 08 Nov 2024 13:17:57 GMT <img src="https://cdn4.telesco.pe/file/o1p3Gpmktu7mKJ66Z0Gb5DgYmHdsx4SZpg3_kAkCQAwkvQRX36hjMnfBeN9JHUQ9CluwXLfUIpxcFf0z-fQxD6uDrTpr-72UjCmkUxHlqM5wYf6CsBTp6Fmz316YDTPXhyVrtVJ51NsRD2brpNo8VSOKnS-7ix-Mv1giHegsbqestQpcN4MvuRUHMilEOMBD-IOLRyaKEg1Ss5W3W7W5f7_pyu8gUPrwaG5Xo7jMwQezAqzX7XYelW5gymd-Y1hX9KlpoHGhcKjXi3B1vu9iCkRg9e88cvjtG1U2-G7aMogUxRqCAYJMlgrAYWjXrDmJpjpWRw2velezfeKDXrvNLQ.jpg" alt="" loading="lazy"> <p>I once compared modern tech to the industry of the 1930s–40s. Well there you go, it’s exactly that.<br> We go to the factory every morning yeah yeah</p> <p><a href="https://t.me/chillhousetech/930?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Investors used to write: where’s your potential to build a billion-dollar… https://darkolorin.com/posts/929.html https://darkolorin.com/posts/929.html Wed, 06 Nov 2024 01:59:31 GMT <p>Investors used to write: where’s your potential to build a billion-dollar company. Now I’m noticing: where’s your trillion-dollar company here. <br> So yeah, inflation.</p> <p><a href="https://t.me/chillhousetech/929?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Had this thought pop up from classic management. https://darkolorin.com/posts/928.html https://darkolorin.com/posts/928.html Wed, 06 Nov 2024 00:15:52 GMT <p>Had this thought pop up from classic management. Maybe someone more well-read will immediately point out where this has already been written down.</p> <p>So, in any corporate situations you need to account for a parameter in your decisions called “conflict capacity.” That’s when your way of resolving the situation hits a bunch of entities and people at once — then the odds of a head-on clash are higher, so always pick the path of least resistance from the participants, meaning the smallest number of them.</p> <p>By this modest theory, going to the CEO is always the safest and fastest decision. Only you often can’t make it there :)</p> <p>Just thinking out loud.</p> <p><a href="https://t.me/chillhousetech/928?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Claude is rolling out more and more genuinely convenient stuff. https://darkolorin.com/posts/927.html https://darkolorin.com/posts/927.html <p>Claude is rolling out more and more genuinely convenient stuff. Still, looks like that CPO from Insta is having an effect.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/927?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “Read, but maybe not fully understood” column. https://darkolorin.com/posts/926.html https://darkolorin.com/posts/926.html Fri, 01 Nov 2024 02:22:33 GMT <p>“Read, but maybe not fully understood” column. Issue 2. From my phone. <br> 1/ <br> A classic paper (<a href="https://arxiv.org/pdf/1911.01547" rel="noopener">https://arxiv.org/pdf/1911.01547</a>) <br> on what AGI is and how you can even make sense of it. <br> 2/ I rarely use frameworks when working on a product. But <br> this one (<a href="https://www.lennysnewsletter.com/p/how-duolingo-reignited-user-growth?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.lennysnewsletter.com/p/how-duolingo-reignited-user-growth</a>) <br> is probably the best I know for working on retention. The whole idea that you should pick metrics to match the use case and do pattern matching across a bunch of benchmarks is just chef’s kiss. And overall the article is great. <br> 3/ I just like reading Quanta Magazine. Don’t know why. This <br> time (<a href="https://www.quantamagazine.org/the-s-matrix-is-the-oracle-physicists-turn-to-in-times-of-crisis-20240523/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.quantamagazine.org/the-s-matrix-is-the-oracle-physicists-turn-to-in-times-of-crisis-20240523/</a>) <br> it’s about s-matrices of probabilities. <br> 4/ a lot of scientists I respect believe that AGI or just the next-level lil model will show up when they combine deep learning and <br> this stuff (<a href="https://people.csail.mit.edu/asolar/SynthesisCourse/index.htm?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://people.csail.mit.edu/asolar/SynthesisCourse/index.htm</a>) <br> 5/ <br> some thoughts (<a href="http://www.incompleteideas.net/IncIdeas/BitterLesson.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://www.incompleteideas.net/IncIdeas/BitterLesson.html</a>) <br> from a smart dude about how, like, search and learning will always be the methods <br> 6/ <br> an old little article, (<a href="https://www.wheresyoured.at/the-men-who-killed-google/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.wheresyoured.at/the-men-who-killed-google/</a>) <br> but I love coming back to how easy it is to add a couple billion in profit at Google</p> <p><a href="https://t.me/chillhousetech/926?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> How did it land for you? https://darkolorin.com/posts/925.html https://darkolorin.com/posts/925.html Thu, 31 Oct 2024 18:56:29 GMT <p>How did it land for you? Share your feedback <br> <a href="https://openai.com/index/introducing-chatgpt-search/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/introducing-chatgpt-search/</a> <br> I haven’t had a chance to try it yet</p> <p><a href="https://t.me/chillhousetech/925?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Surprisingly, nothing really changed. https://darkolorin.com/posts/924.html https://darkolorin.com/posts/924.html Thu, 31 Oct 2024 18:05:19 GMT <img src="https://cdn4.telesco.pe/file/Jv9yxkwgsrRuZ53rcpQeqkMrg0pov1CmqYSJTS8OLCwPuAfyyq4CfqU4nLlhABCVc8uNMAhqvfnQjqmwufEbcwBFqXu8yGClIz14OlsBQcdHVLEZrVtNUPCMFrR3HWLbvdMQ5KvXPNfT72gBLilgZBQCAD57QIXwF1sb7826AsZ88c8tQetpMfJNTobQMr4bopmrT3OK4bWsaxKpLXAhyNrvpnfFvepUwDWws7_sFG7SpJjHYc_oJE5A3lml5sGGZ-mboPxaAlZCphcV3IxOQcgmXbzpFlsx0D0OuIvWzu-08TVb4sEm30zqgwkp2xLp2T2KRI2kvmH7_cxj6-w-5Q.jpg" alt="" loading="lazy"> <p>Surprisingly, nothing really changed. So the market is more or less in the same place for 5–6 years. Well, in terms of the adoption rate of new apps and device turnover.</p> <p><a href="https://t.me/chillhousetech/924?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A set of courses on using Anthropic’s LLMs. https://darkolorin.com/posts/923.html https://darkolorin.com/posts/923.html Thu, 31 Oct 2024 10:10:11 GMT <p>A set of courses (<a href="https://github.com/anthropics/courses?tab=readme-ov-file&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/anthropics/courses?tab=readme-ov-file</a>) on using Anthropic’s LLMs. Recommend it.</p> <p><a href="https://t.me/chillhousetech/923?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Be honest. https://darkolorin.com/posts/922.html https://darkolorin.com/posts/922.html Wed, 30 Oct 2024 17:33:03 GMT <p>&quot;&quot;&quot;<br> Be honest. Have you ever officially worked two jobs at the same time? Like two paychecks and all that :)</p> <p>- Yes<br> - No<br> - No, unless there were side projects without much money<br> - No, but in parallel there were contracts or freelance gigs<br> - 😏<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/922?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Stole it. https://darkolorin.com/posts/921.html https://darkolorin.com/posts/921.html <p>Stole it. Well, this is the best use case for AI.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/921?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Checking the status. https://darkolorin.com/posts/920.html https://darkolorin.com/posts/920.html Mon, 21 Oct 2024 12:28:36 GMT <p>Checking the status. Are you using some LLM provider for search?</p> <ul> <li>Yeah, I forgot about Google altogether</li> <li>Yeah, but only a small chunk of queries moved away from Google</li> <li>Yeah, but the use cases don’t overlap with Google</li> <li>Very rarely use it, like a couple times a week</li> <li>I don’t get why the hell anyone needs it at all</li> <li>😏</li> </ul> <p><a href="https://t.me/chillhousetech/920?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So here’s the ultimate guide to cold outreach tactics, no “drop a + in the… https://darkolorin.com/posts/919.html https://darkolorin.com/posts/919.html Thu, 17 Oct 2024 14:45:31 GMT <p>So here’s the ultimate guide to cold outreach tactics, no “drop a + in the comments” to harvest your contacts, webinar subs and all that. <br> I’ll write it using an investor email as an example. You can map it over yourselves, I think.</p> <p>1/ subject line — what you’re doing (we are doing commerce solution for gaming influencers) and if you think they’ll recognize you then add your first + last name (downplay how sure you are they know you, people get hundreds of emails a day and if you’re not personally talking once every three months with friends of friends then they don’t know you).</p> <p>2/ opening is hi &lt;name&gt; then introduce yourself (first last) and add a 5-word track record: did Prisma, raised from, acquired by etc etc. No details.</p> <p>3/ then show you’re following the dude or his work. Read your twitter and appreciate efforts towards bla bla bla, saw your last deal in bla bla bla. Then say like yeah that’s respect and it’s relevant to us.</p> <p>4/ what you do in two sentences. We are building ai for dog seaters, this is a special device that helps bla bla bla. Got 3k users, 500 MRR some numbers. We Believe in this 1T market. Ai will change it.</p> <p>5/ what you want from the person. We started fundraising. Wonder if you are interested. Please lmk if you wanna chat.</p> <p>And zero links. Converts best for me. Passes any filters, I send from my personal email. Beforehand I try to find the dude on twi and comment at him there, but it’s optional. The whole email should fit on a phone screen with no scrolling. If you have substance like strong numbers and results always drop it into the “what we do” block, if there are request details add max one sentence.</p> <p>Best response is a short yes “will connect with the team, when do we call etc”, short “no, not interested”. No in-between.</p> <p>👍</p> <p><a href="https://t.me/chillhousetech/919?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Wait, so do all successful entrepreneurs and startup folks inevitably end up… https://darkolorin.com/posts/918.html https://darkolorin.com/posts/918.html Thu, 17 Oct 2024 13:21:41 GMT <p>Wait, so do all successful entrepreneurs and startup folks inevitably end up selling their courses or, like, webinars? Because it’s either you’ve got a lot of money or you’ve got a lot of work, and if you’ve got time for courses then something’s not adding up.</p> <p><a href="https://t.me/chillhousetech/918?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One-liner for my dear subscription businesses: In ~6 months FTC will force… https://darkolorin.com/posts/917.html https://darkolorin.com/posts/917.html Thu, 17 Oct 2024 11:48:12 GMT <p>One-liner for my dear subscription businesses: <br> In ~6 months FTC will force (<a href="https://abcnews.go.com/US/ftc-adopts-click-cancel-rule-aiming-make-easier/story?id=114869338&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://abcnews.go.com/US/ftc-adopts-click-cancel-rule-aiming-make-easier/story?id=114869338</a>) <br> one-click unsubscribes without a pile of confirmations, and in general everything’s moving toward signing up and canceling being equally easy. Well, fucking finally—end of all these scammers, I’ve been waiting for this for years.</p> <p><a href="https://t.me/chillhousetech/917?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Since my channel is slowly growing, and mostly thanks to more collabs with… https://darkolorin.com/posts/916.html https://darkolorin.com/posts/916.html Wed, 16 Oct 2024 18:23:38 GMT <p>Since my channel is slowly growing, and mostly thanks to more collabs with folders. This time it’s a folder from the folks in the r founders community. (<a href="https://t.me/addlist/72XGG5CmczlmODM6?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/addlist/72XGG5CmczlmODM6</a>)<br> It’s mostly about specific people’s experiences, so take a look, maybe someone will click with you.<br> Sneak peak:<br> - Misha Lyapin has an interesting take on Africa; I’ve literally never done anything for that region, and you can see the cultural dynamics there especially clearly<br> - proventure is the classic from Denis Efremov, who’s been doing investments for ages<br> - and in general I recommend the job board :)</p> <p>Special thanks to the community lead Ekaterina Kurasheva (by the way she also has a channel (<a href="https://t.me/katyagrows?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/katyagrows</a>) if anyone’s interested).</p> <p>Feels like the audience-swap format works up to a certain scale. We made a lot of communities in Discord and now in Telegram. And it seems like every community, while it’s narrow topic-wise, has a ceiling. And then it slides into generic topics: like how to make money or how not to fall into depression. Which is obviously too general for me personally.</p> <p>Who’s got their own channel? Who wants a collab?</p> <p>P.S. Don’t know about you. I spend 5–7 minutes a day on the channel, sometimes less. Just because I keep writing what I think. Probably that’s also why it doesn’t grow faster.</p> <p><a href="https://t.me/chillhousetech/916?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Since a topic about small teams came up here overall, and for me that’s… https://darkolorin.com/posts/915.html https://darkolorin.com/posts/915.html Wed, 16 Oct 2024 00:19:11 GMT <p>Since a topic about small teams came up here overall, and for me that’s basically an analogue of a startup or “startup mode” inside a corp.<br> And I’ve kind of built those teams, more than once. I’ll write down a few important signs.<br> 1/ You’re looking for someone who shares your idea. Or it’s called product obsession. It’s rarely a feature, it’s usually a certain development vector for the domain. Like in consumer it can be gamification or messaging. You’re looking for someone who wants to work in a certain area. And don’t look for someone who wants to build a specific feature, because we all know that most often the core idea goes through massive changes (pivots), while often still staying in the same area (that’s why moving from B2C to B2B feels so painful, because you need completely different people).<br> 2/ You’re looking for group dynamics. In short, all members of your team interact with each other to an equal (well, at least non-zero) degree. It’s important for everyone to see that their colleague is also trying and contributing, even if not always valuable, but the effort matters. That’s why it’s important to create an atmosphere where everyone feels the movement and you all invest into it together.<br> 3/ You’re looking for growth potential, not potential for “applying existing knowledge.” The ideal team is one that has somewhere to grow in hard skills (and not only soft skills). That’s why it can be extremely hard to build something working and fast with a team where everyone has 20+ years of experience. They don’t understand where to grow in skills and accordingly how to develop them.<br> 4/ You’re looking for maximum self-management skill. Because processes kill small teams.<br> 5/ And finally, what people often forget: you’re looking for honesty. If someone suddenly doesn’t need any of this anymore — talk, and only talk. So you agree right away on the shore that if you’re sick of it — say it. There’s always a solution.<br> Gathering a high-performance team is probably the most important skill for a CEO or a manager who wants to get results.<br> The topic is big and of course the above isn’t all the signs and there’s a lot more that’s specific to the domain. I guess once I phrase it, I’ll write it.<br> P.S. You’ve probably noticed that on my channel I have no formatting and there are mistakes and typos often. All because I write posts live and without prep. This post was written in 3 minutes of pure time. Just the thought matured.</p> <p><a href="https://t.me/chillhousetech/915?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Remember: best price guaranteed:) 500k for a little server box to train your… https://darkolorin.com/posts/914.html https://darkolorin.com/posts/914.html Tue, 15 Oct 2024 17:41:32 GMT <img src="https://cdn4.telesco.pe/file/jtu9jbo2rO7V_BMMtd9-KZscFNTyOyEbmEQTyacMXMCIIEGPeaaj5MFonCsKYiQDaCatpYdaZzu1v_8jiA-_o3J_3aPr_XlXY7X8_A2hWa4gM54Gg_0rKzyumIUC0P_QREp-uxOl9GZM6MESEPdTufSty5r-a5vG15QYZ7c3Ds-xkse3oJKqJGwj6NwaB9TKlN3goqm69oYqaiTWcNggQs7LjW4BbMONCooW4e3N16JrM1vxxcfZKTvqDc5gDvPJc0uDyh-u_InULPjho3jsqzWnhg9S0bbefskixmQtwQYdqM4oWcNiErB3JgmyT6vJx7Y1c9r-yGohdq9TP8YbYw.jpg" alt="" loading="lazy"> <p>Remember: best price guaranteed :)<br> 500k for a little server box to train your models on</p> <p><a href="https://t.me/chillhousetech/914?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Decided to post a weird column like “read this week, but maybe not fully… https://darkolorin.com/posts/913.html https://darkolorin.com/posts/913.html Tue, 15 Oct 2024 15:22:22 GMT <p>Decided to post a weird column like “read this week, but maybe not fully understood” with a short description of what’s what. Maybe it’ll be useful to someone.</p> <p>1/ <br> The collector’s problem (<a href="https://www.mathnet.ru/links/e36a9d79c107c571b5113dbd78e7aecc/mp978.pdf" rel="noopener">https://www.mathnet.ru/links/e36a9d79c107c571b5113dbd78e7aecc/mp978.pdf</a>) <br> or how many purchases you need to collect a set of n hippos from Kinder. Tons of analogs of this problem.</p> <p>2/ Sometimes I stumble on <br> a forgotten tab with an explanation of rating systems (<a href="https://logic.pdmi.ras.ru/~sergey/papers/NSS12_RatingsAdditional.pdf" rel="noopener">https://logic.pdmi.ras.ru/~sergey/papers/NSS12_RatingsAdditional.pdf</a>) <br> in online games and whatnot. But I’m always too lazy to read it. Maybe you’ll find it handy, you can skip the math and just get that it’s not that simple and “rigging” in matchmaking is “bullshit”.</p> <p>4/ I always read <br> Stratchery. (<a href="https://stratechery.com/2024/enterprise-philosophy-and-the-first-wave-of-ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://stratechery.com/2024/enterprise-philosophy-and-the-first-wave-of-ai/</a>)</p> <p>5/ They showed <br> a neat case (<a href="https://talktastic.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://talktastic.com/</a>) <br> of AI dictation</p> <p>6/ They say (<a href="https://mlcommons.org/2024/08/mlc-algoperf-benchmark-competition/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://mlcommons.org/2024/08/mlc-algoperf-benchmark-competition/</a>) <br> you can start believing in second order optimizers</p> <p>If this kind of random stuff was useful, drop pluses/hearts, I’ll share open tabs more often.</p> <p><a href="https://t.me/chillhousetech/913?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Assuming the company is roughly split between Starship, Starlink and… https://darkolorin.com/posts/912.html https://darkolorin.com/posts/912.html Tue, 15 Oct 2024 08:40:10 GMT <img src="https://cdn4.telesco.pe/file/eX869tCpIXr6-AwE2LDcf-YQ47MQ9xLbmeGJ0p5N_HmkyCRRlVrsaYgdK4eFYFRxvDYyeks4o2F0hwp5Yp-8ANG5Zum_A_68kVp16m04ogZvsfOPCLb6JrYEgesFyLQw3WBSMOVsqssLUc9Tj3IUoE5nac8oLgYNXEV7Q8YjT2hacalE8zg6_QJhZl0yyj_EyC0iK4tCrfsv0QqQ93ZjYfwMbaC0zVbWqdrVUKzEq2lAnEXHATV6vkKbNGO2Y15PQ8gE_UOAZikZJN7ERfp_PxCxci5PMATiJGXHlBZJw3euxDaL5VJR4V06Gstkh440uCakmIBTmjqL9YTx4qr3tA.jpg" alt="" loading="lazy"> <p>Assuming the company is roughly split between Starship, Starlink and F9/Dragon/etc. <br> But honestly, even back when I was working at Yandex I didn’t get why you need thousands of people to make search. Then for a while I kinda got it, then I lost the thread again. <br> And yeah, I think huge companies with tons of clients and millions of users can be done with a much smaller number of people.</p> <p><a href="https://t.me/chillhousetech/912?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ll toss you an essay to read over the weekend — by Anthropic’s CEO. https://darkolorin.com/posts/911.html https://darkolorin.com/posts/911.html Sat, 12 Oct 2024 23:12:34 GMT <p>I’ll toss you an essay to read over the weekend — by Anthropic’s CEO. A big essay. Mostly about what AGI is gonna give us; he calls it powerful AI. About different areas where the effect will be most noticeable (nothing particularly surprising overall), and also he’s extremely optimistic that such a…</p> <p><a href="https://t.me/chillhousetech/911?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And also OpenAI showed an experimental lib for orchestrating agents and… https://darkolorin.com/posts/910.html https://darkolorin.com/posts/910.html Sat, 12 Oct 2024 16:46:27 GMT <p>And also OpenAI showed an experimental lib for orchestrating agents and assembling them (<a href="https://github.com/openai/swarm?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/openai/swarm</a>)<br> and so on. Anyone tried it? Let me know how it is :)</p> <p><a href="https://t.me/chillhousetech/910?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ll toss you something to read — an essay by Anthropic’s CEO for the weekend. https://darkolorin.com/posts/909.html https://darkolorin.com/posts/909.html Sat, 12 Oct 2024 16:19:57 GMT <p>I’ll toss you something to read — an essay by Anthropic’s CEO (<a href="https://darioamodei.com/machines-of-loving-grace?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://darioamodei.com/machines-of-loving-grace</a>) for the weekend. Big essay. Mostly about what AGI will give us, he calls it powerful AI. About different areas where the effect will be most noticeable (nothing particularly surprising overall), and also he’s extremely optimistic that such an AI (one model, many, or some combo of a whole bunch of stuff, possibly with a new architecture) will show up in 2026 (apparently they need to raise some more money).</p> <p><a href="https://t.me/chillhousetech/909?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Do you believe in those IQ-ish tests / number puzzles during interviews / the… https://darkolorin.com/posts/908.html https://darkolorin.com/posts/908.html Wed, 09 Oct 2024 13:25:42 GMT <p>Do you believe in those IQ-ish tests / number puzzles during interviews / the hiring process?</p> <ul> <li>Yeah, it helps to see if the brain works at all </li> <li>Yeah, but more like just to take a quick look, see what’s what </li> <li>No, total bullshit </li> <li>😏</li> </ul> <p><a href="https://t.me/chillhousetech/908?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Tip of the day from Tobi Lütke (I respect the hell out of Shopify’s CEO for… https://darkolorin.com/posts/907.html https://darkolorin.com/posts/907.html Tue, 08 Oct 2024 12:59:18 GMT <p>Tip of the day from Tobi Lütke (<a href="https://x.com/tobi/status/1843318517900493132?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/tobi/status/1843318517900493132?s=46</a>)<br> (I respect the hell out of Shopify’s CEO for the culture he broadcasts): every so often you need to nuke all recurring meetings so the company is healthier. Then for any new ones you’ll have to come up with the justification and reasons all over again. Helps you focus more on actual work.<br> I guess I suffer through one team weekly, which I sometimes cancel. But all those recurring 1:1s and the rest are just a curse, if you ask me.</p> <p><a href="https://t.me/chillhousetech/907?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile the Nobel in physics (why in physics — I’ve got no clue) is being… https://darkolorin.com/posts/906.html https://darkolorin.com/posts/906.html Tue, 08 Oct 2024 11:16:36 GMT <p>Meanwhile the Nobel in physics (why in physics — I’ve got no clue) is being awarded to the modern founding fathers of neural nets in computer science — Hinton and Hopfield. Recurrent networks and backpropagation. Kinda surprised, but I guess even the Nobel is signing up for the hype now.</p> <p><a href="https://t.me/chillhousetech/906?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> On Sundays, besides dumb grinding in games, I often pick something to read. https://darkolorin.com/posts/905.html https://darkolorin.com/posts/905.html Sun, 06 Oct 2024 13:58:01 GMT <p>On Sundays, besides dumb grinding in games, I often pick something to read. Today I’m finishing up what I’d forgotten in ML. And the best option in Russian imo is this one (<a href="https://education.yandex.ru/handbook/ml?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://education.yandex.ru/handbook/ml</a>)</p> <p><a href="https://t.me/chillhousetech/905?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m still firmly convinced that product managers in a startup aren’t needed. https://darkolorin.com/posts/904.html https://darkolorin.com/posts/904.html Sun, 06 Oct 2024 08:10:13 GMT <p>I’m still firmly convinced that product managers in a startup aren’t needed. <br> I still think the founders should play the role of product lead, since they own the strategy and the budget (if they managed to raise it or earn it). <br> I still think the product manager function shows up and becomes valuable when a company needs optimizations in an already clear direction (a ready business, product growth, PMF exists). <br> I still think the product owner can be a developer, a designer, a QA, whoever — as long as he or she can prove it needs to be done and drive it to results. <br> And with all that, I keep hearing people sing odes to the product manager profession.</p> <p><a href="https://t.me/chillhousetech/904?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Wanna raffle off my free consult for like 30 minutes, you interested? https://darkolorin.com/posts/903.html https://darkolorin.com/posts/903.html Fri, 04 Oct 2024 23:13:59 GMT <p>&quot;&quot;&quot;<br> Wanna raffle off my free consult for like 30 minutes, you interested?</p> <p>- Yessss<br> - Nah, what’s the point<br> - So I can just write lol<br> - 😏<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/903?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sometimes I jump into random dialogues on Twitter. https://darkolorin.com/posts/902.html https://darkolorin.com/posts/902.html Thu, 03 Oct 2024 00:04:21 GMT <p>Sometimes I jump into random dialogues on Twitter. This time I ended up with GP at A16Z Martin Casado. <br> He posted a really interesting take that’s very close to me and that I instinctively lean toward, but I can’t explain why it’s like that. <br> It all hangs on this hypothesis. There are two types of problems solved with AI: problems with soft decision boundaries (soft-decision boundary problems) and problems with hard decision boundaries (hard-decision boundary problems). In soft-boundary problems (all sorts of assistants, image/effect generation, etc.) you can allow small mistakes without affecting the outcome. Hard-boundary problems (agents) are the ones where even a single mistake will affect whether you succeed. An agent makes a sequence of decisions and performs actions to reach a result. One wrong step leads to failure. That’s why they “empirically” feel unreliable. And that’s why a bunch of questions pop up around their alignment. <br> The “fix” for this mismatch is that you verify the solution (either with another smarter agent, or via verification against facts or search), and if it doesn’t fit, you do another 100500 different prompts for the same step. And keep going until you hit the target. Which, for me, basically contradicts the idea of an LLM as a compressed representation of data with some probabilistic structure. <br> TLDR: why try to eliminate probabilities if they’re baked into the very essence of the tech. Haven’t gotten it yet. And it seems like for soft-boundary problems an LLM is a great tool, and where the cost of an error is high — not so much.</p> <p><a href="https://t.me/chillhousetech/902?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And once again — a double down bet from investors. https://darkolorin.com/posts/901.html https://darkolorin.com/posts/901.html Wed, 02 Oct 2024 20:39:51 GMT <p>And once again — a double down bet (<a href="https://techcrunch.com/2024/10/02/openai-raises-6-6b-and-is-now-valued-at-157b/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://techcrunch.com/2024/10/02/openai-raises-6-6b-and-is-now-valued-at-157b/</a>) from investors. Personally, it seems to me the risk is high that a new architecture shows up that wipes out their success and makes a level of quality like the guys have available to everyone. But time will tell. For now I’m digging into it.</p> <p>What do you think — will OpenAI win, or are we waiting for a new wave of companies?</p> <p><a href="https://t.me/chillhousetech/901?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And once again, for our own analogs to show up, I guess, yeah? https://darkolorin.com/posts/900.html https://darkolorin.com/posts/900.html Tue, 01 Oct 2024 19:23:09 GMT <img src="https://cdn4.telesco.pe/file/fVEsM415WUcNrvwFFpj7W5HAiM5F8b_E9SRaeqHvCyZjWJ-3wGhXdiaZyEsXsLWE3UvesIsTy5n3tiaT8lizgZxrqwWm1dQmafgp5GWDgkX4VGhPQKRFAm73DE4i66cnG9BaKtAhf3ixBHzVlKNfITkARvlhujHBbDW6ltki2X_SS5vdGCYX4y_iCgmbWuWls7jdwdd4SH_QryeVJ6G8IZwE8q3N_Z689KS7HgIRPCz9EuuP-vIR9NAM3t3ErOVEgUsj6w2eA_Wk2THX2l8eSctg0IqXRUdwKcCz3JDBItKs6wGLGslW6yBe4vwhtb-KlB1sYRUP6P5PIJsqAKx9Kg.jpg" alt="" loading="lazy"> <p>And once again, for our own analogs to show up, I guess, yeah?<br> 1. Progress is unstoppable<br> 2. It still won’t work, nothing will<br> (C)</p> <p><a href="https://t.me/chillhousetech/900?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> How do you take honest but unpleasant feedback about your work? https://darkolorin.com/posts/899.html https://darkolorin.com/posts/899.html Tue, 01 Oct 2024 12:55:17 GMT <p>&quot;&quot;&quot;<br> How do you take honest but unpleasant feedback about your work? No rudeness. Like, for example, they tell you: you’re not managing to do this, you need to learn.</p> <p>- Yeah, fine<br> - It’s hard for me, brings up mixed feelings<br> - It stings and feels like it should be phrased softer<br> - Don’t care<br> - 😏<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/899?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> AI startups are barging into the market. https://darkolorin.com/posts/898.html https://darkolorin.com/posts/898.html <p>AI startups are barging into the market. With a bang. Vasily :)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/898?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Almost the entire end of last week I was driving around Scotland. https://darkolorin.com/posts/888.html https://darkolorin.com/posts/888.html <img src="https://cdn4.telesco.pe/file/Qf3uBNgqILOG_N-lUvcBHZhbFnGTH9eRg689zEflGFAYoPTQcU5gh_ugKXlMSjtvLhLvjw2Puns6AjShu1xykIXh-g86_Tqs6t9qoKeRNnzj5QheH51fu8OMA5V5Nu2WaixMZuekVroYPbrqkW0oYm3zO6Pa5MQTSaRmXs_4aJLerySvZjGbtW-bCByp1ybs1ODXeVBDAtsbopSyrcU95oo8LyxsaWyrdgQz6t4UHeDWU6NBOfaTpCs2NzvCOLlBsOjVNJGdWZOTCc3sEJX47A1iaBmB-utDx2QZIbHGdHNL12XlIIskEDs9tiL5OXcXaLxXmFYVmdTcLFMJ7WFNMg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/ihQgqVxk4uhV_6A8YrSdKSImiXFrPHfS2FT0koy2B0fbb8mh8IFp-CSMLxEmQA387Te27u1BNUw7iosTABHIXZpGZTXwCAnzh5rB0fJIRdLcNp8XImrqHklq2GsXZoAJ4JcRQMiMf95NQbBIQHi93MqJ92ahfMxusaVWA6X8NVk2wYe4CaWpqzsZL-bufztyqk4o9qG5AVkxEmyvvfTkOwbY0G12gFn-hbIG0hMxcGxtd-pxFR3tYrTMKIG1ow6WMAKelemiIi6J41m-FWfgansYPP7IyE9XHG9GlS4LYVUnZyRANy2cNO-NeAiDvY8WIcpDxxpJb4iwBR4_yGs8zQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/LAFsEB_dkVPNE0QqkIm2HBRqRYOlgj2XgINaqNvoYgRjijpS07QSIyE5ZTWEDMb-V875L0jxF4xR6omCn-82ZePB2vja_lonnYd-6O4vUJyobdf4iOslHKzdCg9prbYFCIg4JNC5LIAwhiREpT794cPRx55lo9ff7UYNG9rBURDD58JC61nmHTmQWh_SUW9X_FczpmV8TMcW3ZWTmSKhcSRNNRR8HwNYFPQ1gqFmiKb1M533dnI9dS1DkMRle8t-CWqhUZO5CONKhrO2kXXMk8JN15q_jsn65_9NM_iUH6YFPD2TBs5LrDx3T892ikiSVlcY6MBv_24CuckW1tTJeQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/OH38XbSDBqgfqYyugSeuuLlI81hQxaxHJxMYQxPgLpzzyPXryIC7xeNVFIWoxdvT-qHMUuhQBLDPOUZzplfs5emMR2jf22WYQifh9orQoGjMLfPYI_gZIMLNfbf6MX2hgn_fsD_8oOEZ1NNYJHGCcFo26oATEEsdny6azgul_SqMz-NAcgYewJr5YmvK6_8tVVD9IgoH98dSoSAsoFaxz3NzpxzMGdGX7JKfioZtu66tWx3NJQ505hLMq7YQJR4Hfz-ROVGneOYDUutstp2KB_t-mNVldUjPLYCD1qwMMpCiipSQBysl9TrZmrmKY3bEHWGTH_Mc4j45hXG5xtYJtQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/ozUduQoso-RiHkSaKdHCxrJPe6lFIXXT22kHkhyEep4en-yfcduEtwJKpkgR2mAJebNhTVW3YUTwEngG2NYJoVJ_5Fy6EAQLcHvSJRITRRH3UQBzS9r5IZxk4fpKU7ODgJo0JwczW36IsFK4IpxoTPaKDDs0bKhKE95hsLom8v1dEFJXIX09gJUFakhrO53wajM8NRmP4sVvm2PqWhHl-XJnxCl1fWdTcRjpyc24fuiPMpiC7b9qhS-xqAR76_Kvh_meWbbX0uLaDj3B0TGbCteavZl-5z5aY_E_QIflLRxtLXabfY_TvxCiH6ojaHNRbLpu3Lp2_8-UKl_Jqp3xZw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/uS6UC0ZYRzHVne32X66aaFm5Gj4mStLHDK2auFoUM44RhATMO2-ii1EsDxTnyHOCQh7FuVSj237K-R4GmfDyhz_klsz-Wa-e4U2lN2GsOEBJs3ly2uaq1T-UXCfKNspDDDYfk6IqjrHgfw6t92fBHZk3JcSnW7V29DpRIb4I2rav6gVSfLnO2DVhcFA-WdQWy72Rm6jhogxkuYuu5yLuZHMaWLZYgl2bWkneW_nXSPn5utlNuWogxpcQ-VweZKN76zpIV18cMI3_EIxaHcJRxmBY4N-NmlKgCAmQwgBoOwL1pFOMmdTyw561xYMP-9LN7IfjYDPTFeOk424Tzw1T5g.jpg" alt="" loading="lazy"> <p>Almost the entire end of last week I was driving around Scotland. Usually I don’t have much lifestyle content here, but sometimes it happens<br> (for example (<a href="https://t.me/chillhousetech/657?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/657</a>)<br> from my last trip to Korea). I’ll add a few pics — this time I was in the main part of the Highlands, the Isle of Skye, and a couple Highlander castles.<br> And while I was out in the woods and mountains, the governor of California vetoed the infamous SB1047, which could’ve seriously messed with the work of AI companies and model building in California. The veto text is<br> here. (<a href="https://www.gov.ca.gov/wp-content/uploads/2024/09/SB-1047-Veto-Message.pdf" rel="noopener">https://www.gov.ca.gov/wp-content/uploads/2024/09/SB-1047-Veto-Message.pdf</a>)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/888?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A few memes about OpenAI CTO Mira Murati leaving https://darkolorin.com/posts/886.html https://darkolorin.com/posts/886.html Wed, 25 Sep 2024 21:03:21 GMT <img src="https://cdn4.telesco.pe/file/iHof0zJNJD_Lou1VuLstoQJb10ENIXsN9YqvA4TpEjOLOTOfdeNGchw4DYD0BtgfA_4YdfKWVH_9mVRX22lD07pE3fy5kKkQZD52VDSqtaTNVVpc-buFSCiCBKXTceU1pmMTb6m9vQKDIjthcDcb5jHsvnOLT7dnGuInmWRT76utdtAx2PAZ59w7SPAxxf_7SK5-XzlWyQ0rTAynecM7UUdcdz0J00HZ7GZM31ePYEStV8HVYdWo0ORD_QmPlcwj9x1EhqoAV1U_rcmc1PxkJp-0mmlldOvPpoOBAMnjKqbGUypi2Kw5rqf264pYuXQwUOB-DUQoHAqDO7f8jYEduw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/oOm_ZEmU7C9BRxu5vQYxoT2PDU9-KiovOM_JhvM_-Jps8kI9m1GH_Xhhhed6mrXaK831bghX-bJ2u83IiQ0ngo0izLtN615bhb_ljIWIUFa4NzVmfl2DS_-JeOGBtvjAQjdXbFGnkFKl0KmMgKCKwzS7f2wAmu35gdZP4wfs9dcSbBdfcaXqhG1ZTshG24D7xsF83jXbdQMg2q_x7XRj8a7i_KLYMK8_VY1RUsNoFtU6M35UKPGCaB-8jdAE335BgXhdBwnSZMwDxfAH60jJt-IPPqhULIUH9OcgH-tD5h6RvuvDOvDkTc9-ZwqPworCR7SmZvhxm_MipGwe7GpN4Q.jpg" alt="" loading="lazy"> <p>A few memes about OpenAI CTO Mira Murati leaving</p> <p><a href="https://t.me/chillhousetech/886?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, do we believe this? https://darkolorin.com/posts/885.html https://darkolorin.com/posts/885.html <p>So, do we believe this? Orion glasses from Zuck.</p> <p>I’ve got this opinion that a new device shouldn’t fit into any of the existing form factors of ready-made accessories. In short, Neuralink is our everything.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/885?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I use it regularly https://darkolorin.com/posts/884.html https://darkolorin.com/posts/884.html Wed, 25 Sep 2024 08:20:07 GMT <img src="https://cdn4.telesco.pe/file/DaXyaR7-FUggoBi0u_Ncn4lZo4ISPJxOg8QUHm49BYASWaDmd-31ncMoysthF9EhgDwOuJFOvOPHg3aKRob7xtWCsBP4VlPtwCKAZGd5nv3a_m2u1MjoHR2vlSyxtB4GfFxYJcPRtPphSiScEMU1FDe3hwjl20idGCYBJyUSaXxUwVpTFrkHzQiFYeNbEJH7O4fEqM7Q-TsNu2s0uBZkJnAh9rmt026LBI1J7p-FOv0uP0eaeoaSs1s9B0ZYtCrOwrzWyThy0wOw-cgZvvADPITsV0JwTrCG0gtkaSDZXMvoM-4coMEmue8wWwVEsuF2draXiSEMSZETc2wc26fGBA.jpg" alt="" loading="lazy"> <p>I use it regularly</p> <p><a href="https://t.me/chillhousetech/884?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m looking at yet another deck from yet another startup. https://darkolorin.com/posts/883.html https://darkolorin.com/posts/883.html Sat, 21 Sep 2024 08:45:09 GMT <img src="https://cdn4.telesco.pe/file/TWv9JXaFasSYLIwlh3KsnCH4g03y5fVvD43ScRoDj85G2RKjYW4YQM7Pk-bsyYof-wsOTnXUuwdMfQ6tqM1ZpBto1I67_zIlgsQeIIN7MndL2JGR9guV2-Hro5kYMpY6NO3EFY9fbJrGdTc93SawmRWjNR9b-Ja1Mn32Dq-hsc3s9oAOCxSUIlmBalhsQer_4VQCgNhLgHPaymTwFuicuTuTGh87VVCnwsMCvLKACxJwgGvbrw_32qFl3W1Ia0Twq-iY3mKALukJn7kzaGIS11UksIRyOQwCXNZYbrlFjJFgM00hLwSM91VQQqN2g9p40Tn8qKXMOwooR8jOa1yqIA.jpg" alt="" loading="lazy"> <p>I’m looking at yet another deck from yet another startup.</p> <p><a href="https://t.me/chillhousetech/883?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Devotion. https://darkolorin.com/posts/882.html https://darkolorin.com/posts/882.html Fri, 20 Sep 2024 15:03:59 GMT <img src="https://cdn4.telesco.pe/file/PLEjx2wyWztfqlz2JQSzUvlqF7-wDvW2ycd0sYVPIh9Ezfam9EkOPUzRJ_ehTp0iNohbTVB3W-DCWnTJDfVSztFlUwnvluWEU7SDdF0yqViVhWDUK_VrFcxqW5z8qTVyiAnQs9iTb9JzaZD8lltSeKGxTO1Y0gRapRXSeQwVjBEVgp4UcqcGPFgjr70i9TA1BQWTWJYHA18kI7NXDnLexUWBwm260Y7UNYe83Lolt4xhykwt_V3vYgAgNeNNNNTBaNndRZ3_8l0_Mi8eOM2ZGQ1srrZDqd9jHIsC1471EbQ4HgyHXmuvv05fYlD5ODu0pqHTH3UafValzOBrYN7WHQ.jpg" alt="" loading="lazy"> <p>Devotion. That’s the path.</p> <p><a href="https://t.me/chillhousetech/882?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There’s this strong feeling that the more people watch all these YouTube… https://darkolorin.com/posts/881.html https://darkolorin.com/posts/881.html Fri, 20 Sep 2024 11:40:13 GMT <p>There’s this strong feeling that the more people watch all these YouTube podcasts about the magic of “insert trend name here,” the less they actually understand anything about the market at all. Probably first and foremost because the video authors themselves have never even tried multiplying matrices.</p> <p>Like, AI agents are the current hot thing. Ever since university I’ve been super careful with pop-sci and all those “simple explanations” of complex stuff. Getting an absolutely wrong concept lodged in your head is way too easy. How about you? Do you watch any podcasts or interviews? Drop your favorite channel in the comments.</p> <p><a href="https://t.me/chillhousetech/881?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For the benefit of users of regulators in Europe https://darkolorin.com/posts/880.html https://darkolorin.com/posts/880.html Thu, 19 Sep 2024 21:35:33 GMT <img src="https://cdn4.telesco.pe/file/by7ljtYFHPsNnrs3TlI_Yvp37856cggQx3UiNRTNHFNqJI6Lj8Sqyhu2bPOw6EZ8pGLtYxoXZ7TGjKG2PH1QmNUARXCQgcaDave5FpKiJS_m3MyFSvrZpChjJ2QF6IIVQWBceYNxzAJBrjVsM0NtsUJTXvhDIEOe-FmyJKIIjMSkQYH052G7Im34zmOM28eB_XFEVIrvPLcRe7y0qmoGzNMIK_YorWqZohT830Lg2PYCud-TuuJ5emIlB25zf8h3l9ijpIZVt69-87TnCgNzATRwpCFPfSWS45bFt0bF-mMZw8587OVFGeTckgPJajAoSuvM8B0CGZDX9gxogwB-Vw.jpg" alt="" loading="lazy"> <p>For the benefit <br> of users <br> of regulators in Europe</p> <p><a href="https://t.me/chillhousetech/880?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Gonna toss in from one discussion about “well we’ll totally do solid… https://darkolorin.com/posts/879.html https://darkolorin.com/posts/879.html Thu, 19 Sep 2024 16:31:03 GMT <p>Gonna toss in from one discussion about “well we’ll totally do solid performance on photo editor budgets”<br> In general my tldr: don’t do anything where:<br> - organic is under 50% — like, as a principle (organic is search, invites, wom)<br> - churn &gt; 5% for the paying cohort per month (for subscriptions)<br> - LTV &lt; 180 days, i.e. your user drops off in under half a year<br> - in the ideas bucket — “we’re strong at performance traffic and that’s it” (trust me, I had a popular photo editor)<br> If you don’t have answers to these problems and you think you’re smarter than the whole market, then save your time.<br> There are some exceptions:<br> - AAA games<br> - casinos<br> - various apps where the whole scheme is “let’s test virality”</p> <p><a href="https://t.me/chillhousetech/879?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One of the few write-ups that’s short and to the point. https://darkolorin.com/posts/878.html https://darkolorin.com/posts/878.html Thu, 19 Sep 2024 00:45:09 GMT <p>One of the few write-ups that’s short and to the point. (<a href="https://open.substack.com/pub/andrewchen/p/unfortunate-startup-pivots-a-short?r=3p9xs&amp;utm_campaign=post&amp;utm_medium=web" rel="noopener">https://open.substack.com/pub/andrewchen/p/unfortunate-startup-pivots-a-short?r=3p9xs&amp;utm_campaign=post&amp;utm_medium=web</a>)<br> Since Andrew fell out of the real product world quite a while ago. But his investor insights are fine.<br> The article lists typical failed startup pivots. Of course we’ll add social features, chats, pivot to B2B (lol, I pivoted into that myself), we’ll build for a broader audience, and so on. Anyway, highly recommend.<br> Meanwhile, be understanding: this doesn’t mean such pivots are impossible, it means think three times what’s what before you start a new product with the old team.</p> <p><a href="https://t.me/chillhousetech/878?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Have you ever had (or do you have) a mentor? https://darkolorin.com/posts/877.html https://darkolorin.com/posts/877.html Mon, 16 Sep 2024 14:21:33 GMT <p>Have you ever had (or do you have) a mentor?</p> <ul> <li>Yes</li> <li>No</li> <li>No, but I want to find one</li> <li>No, and I don’t get why I’d need one</li> </ul> <p><a href="https://t.me/chillhousetech/877?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Kong did a study on what models their enterprise customers are using. https://darkolorin.com/posts/876.html https://darkolorin.com/posts/876.html Sun, 15 Sep 2024 07:08:06 GMT <img src="https://cdn4.telesco.pe/file/QRbAouAHjYhKjmIcATsLsPnwJJ479QveAtOANMPqH5My4-15yigPEDesL1_-Ht0nhKvmc5v5sNo3ZQ1fqH6tofgYPwoUxWTs-XBfzpn6Vw3gDfNYYjGCfy-Olqkc_ZQa8Ro3-jnhDl6WNOSQjNRbMF5d3ud-3Sq6p0nu71yugO-CWx0yJjfKKtpF1UFgu2Hag_dFfvA5MYy5KlYlvtMjd2qpIsjh-2KicfTaTYecrOupWws0ierwDDwP6byFJQq5-osZdVnSdpenmkGuLkMK6JAHvZj6MuS9GNjs-Os01UkWaOGj4vYMhxs8xx2-vi5WytviSamB1N4NqXdfQfQevA.jpg" alt="" loading="lazy"> <p>Kong (<a href="https://konghq.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://konghq.com/</a>) <br> did a study on what models their enterprise customers are using.</p> <p><a href="https://t.me/chillhousetech/876?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Saturday stuff. https://darkolorin.com/posts/875.html https://darkolorin.com/posts/875.html Sat, 14 Sep 2024 16:20:27 GMT <p>Saturday stuff. <br> They’re watching the Geoguessr World Championship (<a href="https://www.youtube.com/live/Llh4_h-vvKQ?si=ksN25qj6qy23GQou&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/live/Llh4_h-vvKQ?si=ksN25qj6qy23GQou</a>) <br> ~25k live. Honestly kinda interesting. <br> Nice.</p> <p><a href="https://t.me/chillhousetech/875?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I love the term “pattern recognition.” A ton of people sell themselves and… https://darkolorin.com/posts/874.html https://darkolorin.com/posts/874.html Sat, 14 Sep 2024 12:56:36 GMT <p>I love the term “pattern recognition.” A ton of people sell themselves and their services like, yeah I’ve seen how thousands of companies make mistakes and blah blah all that.</p> <p>I still can’t wrap my head around how a consultant’s pattern recognition turns into experience. You can watch a lot of stuff, but doing it is a completely different thing. I know firsthand how wildly different the position of a consultant/helper is from the position of a founder/manager who does it with their own team and carries responsibility, not empty words.</p> <p>What do you think? Is pattern recognition (MBA) really that necessary?</p> <p><a href="https://t.me/chillhousetech/874?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, you could say they straight-up said it out loud: like, you posted to… https://darkolorin.com/posts/873.html https://darkolorin.com/posts/873.html Sat, 14 Sep 2024 07:44:06 GMT <img src="https://cdn4.telesco.pe/file/VGN809tWDQZzkpbnKZ1HK1nolkjSTJ8bWsEare5iG0XRAaWYTfEfFIHdzQ57_WMd-SvPpkU0ZeGtODT9SaLQldoQgsHuG42kII5OKGiu6CF3Q5nSJ41S1giGotU6bAyF3cLTOX--bIB33XB9UWKFwrx1tUJTkq68ISBuyJVFDPY_eIzoemdLxLgzjgK-2v3ejQe0j9fMSuA-utjOVzLkVwMaVvJc80c14mqEpU7XVm3FmDScT6P4Yu6udNE0wShBbv4tUjtH_p1dZnSYibgn6GdP8GXSmAulFJWEnK5oMINwrfNe-GMieHLmWABsFunP8xKQu6oDTnaH7-kcp578MQ.jpg" alt="" loading="lazy"> <p>Well, you could say they straight-up said it out loud: like, you posted to Insta, congrats, now your content gets used for training, and on top of that they monetize you via ads. Double value for the company, zero value for the user. Hope they at least ask me whether I even want to hand my content over to train the nets.</p> <p><a href="https://t.me/chillhousetech/873?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile, in the not-so-distant future https://darkolorin.com/posts/872.html https://darkolorin.com/posts/872.html Fri, 13 Sep 2024 21:29:45 GMT <img src="https://cdn4.telesco.pe/file/OBzYhDBpyIgiK0GfusNmVWXrOu-pXcEH8w3qyrra2oOK8gK5W_GbOqEe3aeDH_canjUemdVJyZKfZukYErStzMxWB0dpMmzuVBi7KiLpDw25rQFGK17xz_hA3xQFNOv8IeYZ3nIjdZ3j-seftsTIPKAxIjwTZsbkJzOO_P4tK7KlxcODzJ7I_3ttXqXZR658_HfreNVV1ADfOLQMRS8VffD0kDxWWo5WVcoACa5dMETMrFfMmavZiotjk-Iqe4uPaR6Wz_1f1WlFftHtNTg9NA6HlRY5Dla6OU1RWDUh3eA6wX4fTPkjRCIyULv0NLoK7Akt1CzMB1kzUspR2VJvzg.jpg" alt="" loading="lazy"> <p>Meanwhile, in the not-so-distant future</p> <p><a href="https://t.me/chillhousetech/872?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And here comes strawberry. https://darkolorin.com/posts/871.html https://darkolorin.com/posts/871.html Thu, 12 Sep 2024 17:04:20 GMT <img src="https://cdn4.telesco.pe/file/ra-fH7jG-awRDB57PrbnsqU5JNQBZfrardq7m9fjRoKkgXQNrvZd4QNUG_gj6FwI-vLx-PQo6_mlPcftREaEUGd8VGJLt62MLm6c70LD_NOFdjxTVoYpUAJZ9ly-slCtYVvc14G6MVZb4FVTqTTHFQCJn6sAco9Fr710H37tU0BxhJk7gx92upE8ZSwUQUGwWuX5hSI19uN5yDBm5g_xpZHrAIFSUJhoBvWBv0PEZtWdhWLtdHatRDpEE6NU7HqYExfdfXJkohaKm-ZGZUGETueD6s9LfT8c4mN03PvJ9kX5Z7PJMyyrII0myq-Afb8ZfmyjdXTYTptNpMQQxGDEXQ.jpg" alt="" loading="lazy"> <p>And here comes strawberry. Or rather the o1 little models from OpenAI. Of course, just like I thought, they were done by Noam Brown, the guy who did various things with game theory and was into all sorts of poker strategies and bots. A very respected person. <br> Inside, before it answers, it thinks via chain of thoughts. Also, by the way, they don’t beat 4o on every benchmark. <br> More details (<a href="https://x.com/polynoamial/status/1834280155730043108?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/polynoamial/status/1834280155730043108?s=46</a>) <br> and <br> here (<a href="https://openai.com/index/learning-to-reason-with-llms/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/learning-to-reason-with-llms/</a>)</p> <p><a href="https://t.me/chillhousetech/871?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Out of Apple’s whole presentation today I’d highlight a few components that… https://darkolorin.com/posts/870.html https://darkolorin.com/posts/870.html <p>Out of Apple’s whole presentation today I’d highlight a few components that matter to me:<br> - formally, a presentation of new hardware is increasingly turning into a presentation of new software. A ton of focus specifically on the software, constant mentions of ML/AI and all sorts of optimizations<br> - most of the new features are about health. Apple is doing everything cleanly and going straight into the most expensive area (your time is worth the most, and accordingly your health status) — health tech. We’re hearing about datasets and research and new cool features (which maybe most people don’t really need, but they’re pumping the market like pros)<br> - otherwise it’s boring, and it feels like for quite a while now Apple has been a one-device company and its friends (the iPhone and headphones with a watch). And it looks like there’s nothing new to expect anytime soon. I’m expecting more releases of service products and more cash poured into software, especially AI</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/870?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> GenMoji — Apple’s new level of intelligence. https://darkolorin.com/posts/869.html https://darkolorin.com/posts/869.html Mon, 09 Sep 2024 18:11:08 GMT <p>GenMoji — Apple’s new level of intelligence. Review later.</p> <p><a href="https://t.me/chillhousetech/869?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Second question from the AMA. https://darkolorin.com/posts/868.html https://darkolorin.com/posts/868.html Mon, 09 Sep 2024 13:54:02 GMT <p>Second question from the AMA.<br> It’s super simple and I think it’s important to say out loud: a potential startup buyer popped up (M&amp;A), asked for P&amp;L, revenue composition and cap table, but they’ve got a kinda sketchy reputation. There were successful deals and there were “scams”. What would you do?</p> <p>In any deal you always need to drive it to results, usually there are buying stages and each stage needs to be closed:<br> 1/ You started showing interest / the buyer started showing interest. Often it’s important to voice your plans/forms of cooperation and do it in an email. That email should include not just corp devs, but stakeholders too. Because corp devs are always interested in buying a company (their KPI is to do deals).<br> 2/ Compress the timeline from first interest to LOI (Letter of Intent — a formal letter of intent to buy you). Clearly state what’s missing to issue an LOI and when it will be there. Put the question bluntly: “are you buying us — then LOI/price tag — or we’re out.” Any deal implies hard pressure and it’s important to hold your position and always get an answer. The faster you can get to an ultimatum, the better. Of course this assumes all the other questions are solved, and fast.<br> 3/ Any big corporation has screwed people over, and many times, in M&amp;A processes. So always push it to documents and an offer. Even if that process doesn’t end well, it can totally help you find another buyer. And use the first one as leverage in negotiations.</p> <p>TLDR: always take what you consider reasonable to results in the form of documents or money in the account; if you think you need an exit — work it, reputation and that kind of stuff won’t suffer in this framing. And the time spent is experience and new opportunities.</p> <p><a href="https://t.me/chillhousetech/868?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> That very same cold outreach, just aimed at you:) works, I guess? https://darkolorin.com/posts/867.html https://darkolorin.com/posts/867.html Thu, 05 Sep 2024 21:36:46 GMT <img src="https://cdn4.telesco.pe/file/RmhNGYtIqzokvr7Yq9vJefmOf0mARCTDl0whgeu9YzKa0LiTgugV4jJRASc9KZd3sr6N4cUBbvLCZv52vg0kytPp4EL1cZuYeQBT17pxnJeLqe95Y_or_zVOolizafqkopxEUt-dLBUM-ePRj17K99Su1EvK1M3UEp3-op55AHt8INCaRzn6o8IO22D5RG9_lAZPOSyYAWt2LdI0AQfo6eOfaAtZn9L9rWISE8NYjQKpjQdzTTWLEu9blC9n6LxmjDRUoIHa0w3H_yXzdegg4TR2oxYKfEhAas58papvGIMZYC4-ATN2ksJxStLt-oias2_V8htZPbP3eoNWSLgEqg.jpg" alt="" loading="lazy"> <p>That very same cold outreach, just aimed at you :) works, I guess?</p> <p><a href="https://t.me/chillhousetech/867?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Another trend from the past (remember Facebook lite?) Snapchat is testing a… https://darkolorin.com/posts/866.html https://darkolorin.com/posts/866.html Wed, 04 Sep 2024 21:30:51 GMT <img src="https://cdn4.telesco.pe/file/oiezuincmC1kV__ySBNKcUkDj5cqsW__QXrWZn95sM0K0deIawDroolG_huPz0_oB0YI9l6zBc9vYxWW_87mZXCJ3ljeHs8yt9SCDe0N5ak7zKbvQsH00VSG4I3BxOI7b__ppAGLrgg4hfOaDqgwCKvyap8xKmlRLI5U8ZGiPuWlbS5v79Zkg-DDD4MSiWCE1IfLqeM_U3CEiWZ-_Gb-G1MVQ_KEmJXB1SCgxWEK2ZfJZiZrXzt5sZrJJdHL0Ws-rWdSpcaYfhX1QtfiSv7e1lscQqeswuWXEZryZp_wXXiZQdlbQ1qhMcAmU1AzGMFz4q3PFdPwINzhKn85kmaWDA.jpg" alt="" loading="lazy"> <p>Another <br> trend (<a href="https://techcrunch.com/2024/09/03/snap-ceo-says-the-company-is-testing-a-simplified-snapchat/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://techcrunch.com/2024/09/03/snap-ceo-says-the-company-is-testing-a-simplified-snapchat/</a>) <br> from the past (remember Facebook lite?) <br> Snapchat is testing a simplified version of its app :)</p> <p><a href="https://t.me/chillhousetech/866?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile a normal kind of presed Sutskever’s got. https://darkolorin.com/posts/865.html https://darkolorin.com/posts/865.html Wed, 04 Sep 2024 17:55:04 GMT <p>Meanwhile <br> a normal (<a href="https://www.reuters.com/technology/artificial-intelligence/openai-co-founder-sutskevers-new-safety-focused-ai-startup-ssi-raises-1-billion-2024-09-04/?fbclid=IwZXh0bgNhZW0CMTEAAR0inrpIdvpV5sZHZKNnZ29ZBocNuJVUH9A7ZchK_ErYzzyQPdldFXVC6yg_aem_pjl6x7iPJelf3ltiaBHy0g&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reuters.com/technology/artificial-intelligence/openai-co-founder-sutskevers-new-safety-focused-ai-startup-ssi-raises-1-billion-2024-09-04/?fbclid=IwZXh0bgNhZW0CMTEAAR0inrpIdvpV5sZHZKNnZ29ZBocNuJVUH9A7ZchK_ErYzzyQPdldFXVC6yg_aem_pjl6x7iPJelf3ltiaBHy0g</a>) <br> kind of presed Sutskever’s got. $1B at a valuation of 5. <br> I’m sure it’s a bad investment. And it’ll only make the market situation worse.</p> <p><a href="https://t.me/chillhousetech/865?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> First post from the AMA series. https://darkolorin.com/posts/864.html https://darkolorin.com/posts/864.html Tue, 03 Sep 2024 11:52:54 GMT <p>First post from the AMA series. Pretty simply about how we started and built the first versions of LFG.</p> <p>Can you please talk about the first steps of launching LF.Group: from picking the idea to PMF?<br> Really interested to hear about the very beginning: what you did, and how you understood you were on the right track and should keep going.</p> <p>To make it easier, I’ll throw in topics, you can say a couple words on each:<br> 1. What other ideas you had back then, and how you chose between them<br> 2. Did you raise money, or bootstrap<br> 3. First users — where did you get them?<br> 4. Did you build the MVP before users, or users first and then MVP?<br> 5. What acquisition channels worked at the very beginning? did you try anything else?<br> 6. How did you understand metrics were good enough and you could scale? at that stage was monetization more important, or retention?</p> <p>Alright, let’s go point by point, I’ll try to answer short and to the point.</p> <p>1. The idea was simple: I played a ton of World of Warcraft and finally decided to make something for myself. And the problem was eternal: “how do you find a decent person to play with at your rating and time.” Everyone was sitting in Discord, obviously. So we decided to start with a bot so we could build an audience faster and also less dev. We kicked off with custom lobbies for Hearthstone (threw out a call and played with whoever joined) and parties for key runs in World of Warcraft (picked your class params and rating and threw out a call). It worked like a simple “looking for” post; the main feature we built was the ability to see a search request not only on the server where you left it but on other servers connected to our bot too.</p> <p>2. We raised. Almost right after the first demo and the first 10k users. $1M USD and the lead was Grishin Robotics.</p> <p>3. Like I said, Discord, since we made a bot. There you go manually to a server admin (local community) and make a deal, plus even back then we started working with streamers, since a lot of them have communities in Discord too.</p> <p>4. Honestly it’s hard to say what “users before or after MVP” even means here, I don’t fixate that hard on that concept. Probably we just started acquiring users with the first working version.</p> <p>5. We onboarded communities into the bot manually. Manually came to streamers, manually posted wherever we could (Reddit, different communities).</p> <p>6. No way. You just had to grow and see that growth wouldn’t stop the moment something happened. No monetization. The main task was to understand whether the use case works and whether it can grow at the needed pace. DAU/MAU growth ahead of Retention growth in this case. We started making money already much closer to the exit.</p> <p>Overall it’s hard for me to say there was some kind of PMF as such; what was proven was the idea that with bloggers you can sell in-game items &amp; currency and it works and it got bought. And we managed to grow it to decent numbers and very fast (in under half a year from zero to $10k USD GMV per day — in our case basically revenue since the goods were bought out).</p> <p><a href="https://t.me/chillhousetech/864?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Must read for everyone who’s in startups or in product. https://darkolorin.com/posts/863.html https://darkolorin.com/posts/863.html Mon, 02 Sep 2024 18:48:09 GMT <p>Must read (<a href="https://paulgraham.com/foundermode.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://paulgraham.com/foundermode.html</a>)<br> for everyone who’s in startups or in product. A very clear idea is conveyed about the modes of a product manager and a founder. Where the latter is:<br> - responsibility for what the company does and “owning the product” in the sense of striving for a bigger mission<br> - the ability to listen<br> - and being consistent in feedback and in grounding it</p> <p>Manager mode is a separate topic. And arguably not unpacked that deeply.</p> <p><a href="https://t.me/chillhousetech/863?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Ha https://darkolorin.com/posts/861.html https://darkolorin.com/posts/861.html Mon, 02 Sep 2024 08:12:58 GMT <img src="https://cdn4.telesco.pe/file/Sd3ufvfngeCen0ErNiyB5wyFbGaFYU_gTtCD4QhXvig8CwildkI4RfVpvpKgr2VsIyiFw0hflFqjquzcyK0jujznIJ_ToPUPkgYD8zrRjy0WvtvdAqTRJtHUg5RWtq06cOcL-M3msbL_PonutSpJ8rlMkgPHoXXJFm9eOUaVjUzKGEm8lU4AlPwyumnjyjtU1DG8dQqIrPLccJXxGH_ygvHs3C-DzMy1JLJoP7AA1ZcIonKtzaePOc9ny__tYF4FV4INhJ-UaWu9hmmLo2os-y5NOe5RkHyp25tDK1TvjwuOYk8lg7FFNot586JNdQ3aOWvp34wk6wIYck0ABS9tig.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/WbSMpHBhjYb4_UAc8PVGwGTxShCOQwybPk6kiRCiN3DirAE84Pv9qijFzfi3M4-IifadCEinj2t1IL63U_IHkss-rpZbguKoRAj8IWV3Xs-Z11dA8y7jQ_zasH35oagoeLyCiLGKuLCT-nJethkBn3ou8z7cNx61zLa2362r2WcYp2n-1A4kfGm6fefKMdbxIYwNhDxW9X0-KomtAfVfhT5u4jLqMXtaF_ZMFELcbWv__1Se9-a4UaHc4mzFJJ2XoRyO8KCdz3FxwJ8YYQs52qo_UC1qNMKG1R4Ameon-vmwOKMwQiu-87AaqIcj0-Ca6jTPUfSUIgNHM8Ran7nQAA.jpg" alt="" loading="lazy"> <p>Ha</p> <p><a href="https://t.me/chillhousetech/861?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A Valley engineer is getting ready to log into Jira and knock out a couple… https://darkolorin.com/posts/860.html https://darkolorin.com/posts/860.html <p>A Valley engineer is getting ready to log into Jira and knock out a couple tickets for the year</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/860?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Like I said, a regular AMA. https://darkolorin.com/posts/859.html https://darkolorin.com/posts/859.html Tue, 27 Aug 2024 08:27:57 GMT <p>Like I said, a regular AMA. I’ll be happy to get your questions — usually I try to collect them for about a week and then answer when I can; if there’s something really interesting and I immediately know the answer — I’ll write back almost right away. Usually I pick 3–5 best questions. <br> Questions go here (<a href="https://forms.gle/i9Zh4rEUdkzuiQ1Y9?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://forms.gle/i9Zh4rEUdkzuiQ1Y9</a>)</p> <p><a href="https://t.me/chillhousetech/859?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I love those first posts about hiring from what’ll eventually become huge… https://darkolorin.com/posts/858.html https://darkolorin.com/posts/858.html Mon, 26 Aug 2024 19:29:18 GMT <img src="https://cdn4.telesco.pe/file/pt8JBR6ygM6dZd15cKnBwbPB9es7O-4iXoElCvisF2EGaEhQQVdJiJjhxZW4CcliW6xq1Q0Ewc25rHsE8igh-ZVkpadve1Wab_jaRMy6WY6rXS-MJlITRtK76r8iv63DGZTdu6xpTHN9cJ7ZssbH9QGFnQRSIvRVRzQOh4-ppftj_-ueMW_hgvnMwDokXJKT81fnL4q634uvbnQ9WxQ7Epo0yoiV_a_Ro_eFHCF0hWvtNdf2IoOrbrzFPOc4nh3S8PnzRyt8THksJLIdgKN2XtI5Y9OCVOQX7alcHC_bTai2ysGLgdMrZC4cx5CeMZaXIvfYOy9A68ruscIREzsTNA.jpg" alt="" loading="lazy"> <p>I love those first posts about hiring from what’ll eventually become huge companies <br> Coinbase first post (<a href="https://news.ycombinator.com/item?id=3754664&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://news.ycombinator.com/item?id=3754664</a>) <br> And in the picture Jeff is looking for developers <br> P.S. I’m prepping a quarterly AMA for you, as usual your questions — my answers.</p> <p><a href="https://t.me/chillhousetech/858?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Can’t not write about Pavel. https://darkolorin.com/posts/857.html https://darkolorin.com/posts/857.html Sun, 25 Aug 2024 17:41:11 GMT <p>Can’t not write about Pavel. I’m not going to spin up theories, more like underline the important stuff about the internet and products there. <br> 1/ The age of global consumer products is over, and yeah I think B2B too. <br> 2/ When you’re big, it’s impossible to stay neutral. “Big” is from a billion valuation and ~100M monthly users, somewhere around there. <br> 3/ The internet will be regulated by states just as strictly as regular offline life, maybe even harsher, since it’s formally easier to keep tabs. <br> 4/ That kind of split will create a huge number of opportunities for new products and startups across different markets. <br> And there are already plenty of theories about Pavel Valeryevich. Not my thing to stir up that muck.</p> <p><a href="https://t.me/chillhousetech/857?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Since we’ve got a bunch of new subscribers here again. https://darkolorin.com/posts/856.html https://darkolorin.com/posts/856.html Fri, 23 Aug 2024 07:52:22 GMT <p>Since we’ve got a bunch of new subscribers here again. I’m updating the post with my “most useful” posts.<br> I think these are more or less worth your attention:<br> -<br> about networking on Twitter (<a href="https://t.me/chillhousetech/10?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/10</a>)<br> -<br> about TikToks for dogs and cats (<a href="https://t.me/chillhousetech/17?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/17</a>)<br> -<br> about how I was the first official visitor of Apple Park (<a href="https://t.me/chillhousetech/76?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/76</a>)<br> -<br> about push notifications in apps (<a href="https://t.me/chillhousetech/172?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/172</a>)<br> -<br> about mute (<a href="https://t.me/chillhousetech/174?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/174</a>)<br> -<br> my draft deck for an early-stage consumer product (<a href="https://t.me/chillhousetech/214?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/214</a>)<br> -<br> about early-stage startup analytics (<a href="https://t.me/chillhousetech/395?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/395</a>)<br> -<br> what an investor update looks like — email template (<a href="https://t.me/chillhousetech/449?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/449</a>)<br> -<br> about patterns of fast-growing apps (<a href="https://t.me/chillhousetech/482?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/482</a>)<br> -<br> a bit about community (<a href="https://t.me/chillhousetech/503?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/503</a>)</p> <p>-<br> why I ditched being public (<a href="https://t.me/chillhousetech/543?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/543</a>)<br> -<br> about what I would do differently if I started Prisma now (<a href="https://t.me/chillhousetech/544?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/544</a>)<br> -<br> who I’m subscribed to with the bell on Twitter (<a href="https://t.me/chillhousetech/556?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/556</a>)<br> -<br> about the deal, and startup acquisition deals in general (<a href="https://t.me/chillhousetech/824?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/824</a>)</p> <p>-<br> about various interviews (<a href="https://t.me/chillhousetech/797?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/797</a>)</p> <p>-<br> a tiny bit about retention (<a href="https://t.me/chillhousetech/782?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/782</a>)</p> <p>-<br> about sabotage as project work from the CIA (<a href="https://t.me/chillhousetech/728?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/728</a>)</p> <p><a href="https://t.me/chillhousetech/856?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Google’s pre-IPO financials https://darkolorin.com/posts/855.html https://darkolorin.com/posts/855.html Wed, 21 Aug 2024 07:39:06 GMT <img src="https://cdn4.telesco.pe/file/nCaWhHfGG3u2Y7vkD-irPStyEsQHDsPMtA0LRmfh8TsfmFMxwii5tmqPd4o0efirE5laZ2c89r07xyGU-cYue4hMGeBH9YJbgSXNCrqtuS6mGGMWWF8KjS2rs9J1NipDDR1ZqQqKxh0LzLS0CSXdFwL_ZdXSM1bitn7K_UZM3IvfZHPSG770NZHlJyX9xSHZUYTTvIPg0r-g-xZMAvHGULPNn9o8o-kAuFhMqTvvXwL975sXiJolpo1g27VRfel56vpwThE6m1wEjm1eaSJeo-jtEIPR-yvDZPxtKeDfnHNIrVgm1G1x4dr_LddYhW6ZPoTp44MiVNoLY0nwsdlXrw.jpg" alt="" loading="lazy"> <p>Google’s pre-IPO financials</p> <p><a href="https://t.me/chillhousetech/855?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Found my deck from like late 2017, early 2018, or somewhere around there. https://darkolorin.com/posts/854.html https://darkolorin.com/posts/854.html Tue, 20 Aug 2024 23:32:05 GMT <p>Found my deck from like late 2017, early 2018, or somewhere around there. Maybe someone’s interested. Raised this deck like a mil right away, but just because it was after Prisma. In the Valley it landed fine :)<br> What would you change in it? Or add? The slide with the team wasn’t needed btw.</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/854?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I, dear friends, jumped into a small folder with various kinds of authors… https://darkolorin.com/posts/853.html https://darkolorin.com/posts/853.html Tue, 20 Aug 2024 16:08:55 GMT <img src="https://cdn4.telesco.pe/file/fCwo53e4vFD9PckJ42eX02gUoVEUNllFzZRX4aC2IzuBRVrHvIGsDXfwab6DMw-aI3J05i-Mf-ynjoWHVsm1X9FLYIN78ZDETTiui5rnPZXE1Zqk8TTHdzl4UPmMEN3YvAwSPvzP7ztXmQsfdl_wbPm-tsoW6TLUjlCOcYlREj-uTZLId3Pi47Wtn5GCd4mrxtIvcwvZAJT0S_6GwthBM-hpsQIpJgjUbDqVgXWQiS080SX27BZax1i4erLW5Z_jqfSGUpK-N1ygJCUDKB5VWLVrx9-0vTF3yVa4vj66U73GzaM4eOryrmzaMuorwSA3F6b5nVSd37C2sQWLj_yRNw.jpg" alt="" loading="lazy"> <p>I, dear friends, jumped into a small folder with various kinds of authors about product and whatnot, testing, so to speak, traffic in Telegram and how it works. Obviously all folders are about cross-posting whoever else is in there.</p> <p>So yeah, catch it — click (<a href="https://t.me/addlist/qaZc3jagz08wMDZi?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/addlist/qaZc3jagz08wMDZi</a>)<br> (out of everyone there I’ve known and read Kostya Gorsky for ages)</p> <p>Wondering, if I make my own folder myself, will the effect be cooler?</p> <p>Overall, I haven’t really tested almost anything that worked effectively as “incoming subscribers”; I broke my first thousand just with a repost, then caught a couple more reposts. I’m not putting in effort, I write content that actually comes to mind, try to keep it relevant (most of my posts about startups and product folks in general are dated from 2018 practice and I think there’s not that much useful in them, especially regarding the market and strategy). Money invested: zero, except for an absolutely disgraceful efficiency-wise test of Telegram’s ads cabinet, where I dumped 10 TON and got something like 20 subscribers. Not setting foot in this brilliant mechanism again (though if someone’s in the know on how to pump it, I’ll be glad to spend the remaining 10 TON better).</p> <p><a href="https://t.me/chillhousetech/853?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> To all fans of “small” business dedicated. https://darkolorin.com/posts/852.html https://darkolorin.com/posts/852.html Tue, 20 Aug 2024 09:44:06 GMT <img src="https://cdn4.telesco.pe/file/NPu2qvgkB2TvS7QMI5jUl9iqETNCf396S1URu_hu91Lpkm2fZEveFmpbJdkTvtP99aV0Wg17w1v9FYLBY0POBdeOAP5Pb2_cAMev09krDaSr0SfytJB0SRcHSlGNXWVgysYZPIWH0QuMnVDQPPqk9nbsX4XNDsYtgKeaekwQpsloTK3l4IpGEC6f-PdfVVOx7LA5T_TE9wP74faNKXOrAycJADHQsnLXQUDABkHnbRckvUPVcZvsM5a3LJ1FHkDQozkZXQs11h2EbpkaT1JADcRdsBKKI8s9tvdC3vV-dEhJSdYjWPHYDOZuCpk7em3V6Rj_FDZFzsjII_GYiANt5A.jpg" alt="" loading="lazy"> <p>To all fans of “small” business <br> dedicated. (<a href="https://www.reddit.com/r/Entrepreneur/comments/18ckcw2/who_knows_someone_with_a_boring_business_that_is/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reddit.com/r/Entrepreneur/comments/18ckcw2/who_knows_someone_with_a_boring_business_that_is/</a>)<br> Reddit is basically a treasure trove of this kind of stuff. And there’s probably no reason not to trust the users, I guess.</p> <p><a href="https://t.me/chillhousetech/852?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, looks like the videos are already getting better with controlnets for flux https://darkolorin.com/posts/851.html https://darkolorin.com/posts/851.html <p>Well, looks like the videos are already getting better with controlnets for <br> flux (<a href="https://blackforestlabs.ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blackforestlabs.ai/</a>)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/851?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There’s this thing at Netflix (as a company) called the “keeper test” — I’ll… https://darkolorin.com/posts/850.html https://darkolorin.com/posts/850.html Mon, 19 Aug 2024 15:47:15 GMT <img src="https://cdn4.telesco.pe/file/VQgXI_BIc7ubFBrW4h_LMF_-1Iv5cu6VziL_q8E7cXC45us6nzURaBQmVCxu8Vie7FmTMazugto5A_0g1NovnDpxBNnbbpEt8SHDo-6c_k2h1agKkG8Q76ZMQCOuGsT2OFVIe3zy_xO0xT6T3fdOA3PPSYAzDhfjqm6J6wBePxUlG7lQ3m0CxOqnbON9-5x6hsrY4NbHgymerCbJCZGzRXMHQeHkd2CoUQfHAtmsft22J9bmS4B1cEzJaD0PC7Nl1w3Iv_1rfWf-MRPYCL2TCoN9NS8QjNaNW-NFMqK293Q3KD07cXkE70nzFa3Nb0VEAlHqJS64qf-LP7WE9Rw10A.jpg" alt="" loading="lazy"> <p>There’s this thing at Netflix (as a company) called the “keeper test” — I’ll leave it untranslated because I’m not sure what the correct option is, but the point is this: <br> Managers are supposed to constantly ask themselves: “would I fight if my team member decided to leave?”. And if the answer is no, then it’s time to fire them. And this exercise is daily. <br> They say it’s harsh and builds a culture of honesty in the company. To which there’s an excerpt from the company’s culture code. Here’s a Russian translation with the formal style preserved: <br> A dream team is one in which all of your colleagues are extraordinarily talented at what they do and highly effective at working together.... We model ourselves after a professional sports team, not a family. A family is built on unconditional love. A dream team, on the other hand, is focused on having everyone strive to be the best possible team member, care deeply about their team, and understand that they may not be on it forever. Dream teams are oriented toward performance, not seniority or length of service. <br> In my practice, I try right at the start to honestly lay out what you’ll be working on, who you’ll be working with, and point out the most important thing (I don’t have much of a safety guarantee) about intense work and fast-changing conditions.</p> <p><a href="https://t.me/chillhousetech/850?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> An experienced Russian-speaking—or, as it seems to me, even European—investor… https://darkolorin.com/posts/849.html https://darkolorin.com/posts/849.html <p>An experienced Russian-speaking—or, as it seems to me, even European—investor and founder reading the YC blog and American Twitter</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/849?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sunday stuff https://darkolorin.com/posts/848.html https://darkolorin.com/posts/848.html Sun, 18 Aug 2024 21:09:28 GMT <img src="https://cdn4.telesco.pe/file/LvrbkNQPGbeccdPb_lugvCEBCNmTYee5cSlByCuQ_sS-lfz7p_wZ1FxJVO80uOUNvbLaw1dIZpbLUe21AildZln242itvyVV4mdyeFhVlGviykhBYvixJpQqbwXucd7Hd_pULmMFfYIdLdYR9Ga-bRmoZIoXzPFL6ugf-lCdOtYjwL4e5FV5xqd-YuVVXo9_uzKz4btGM1mmgfWzGfSx4AvpzbkgUkArgb0bTXA0gJJOSw_Zgsza81S5ERUZLPBoiWdd1GBhsljLlAbTrL_SuJt2X6d6Ma0rgc-cKWjrwvt9Y-PWiMk9e_sRLTHkTEPgKlHgl3Rksno8hlJwKk0Wuw.jpg" alt="" loading="lazy"> <p>Sunday stuff (<a href="https://x.com/sashakaletsky/status/1825258632310894716?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/sashakaletsky/status/1825258632310894716?s=46</a>)</p> <p><a href="https://t.me/chillhousetech/848?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Once again, field reports from the venture market. https://darkolorin.com/posts/847.html https://darkolorin.com/posts/847.html Sat, 17 Aug 2024 09:27:41 GMT <img src="https://cdn4.telesco.pe/file/q6QgGWW0lHckBFhOOb5ycWL2nrA5QOAbE0NeXOQngjqUJ7cIgrHoTmrogOJu_NHIEOo3Nq6ROpacG5L_zJZPWikAgQ9DNxvyT5GMGavkpirP-Pp9AXbegqJOLJqSkX5Ryb5V2TTMLgsFgN9wK7QnrB647lTNHrvwIKxMEO1sfLQLP2BKFHDZfhhbMcEXaybMTAIpJBYXSiC2Ws8R7iuQbVjyzpyR83MErpPZ-oMVY2UKGHwUGp3TCF72YfYYP7kiKtC-JbFrL8EGtP2d6Y2u1TeHvsFqQj9yq_PxoJanBbnu89d-TululNGXTb7SGt1qa1jiJrAUVzUYsT16zdfU8g.jpg" alt="" loading="lazy"> <p>Once again, field reports from the venture market. As you can see, fund efficiency is dropping fast (obviously the later it is, the fewer returns so far), and the reason here isn’t only, as it seems to me, the overheated market during covid or the venture winter (rates in the US), but rather that the skill of venture investors is falling.</p> <p>I’ve been in the industry for a long time and helped people raise and raised myself, and imo the quality of rounds has dropped. More and more money goes into hype, less into building value (don’t know how to translate value building properly). Trying to fly in fast has become more important than making something that works. And you can see it even at the level of small companies.</p> <p>One of the reasons is mass B2B and crypto (then came it and AI). The level of people in B2B and crypto was lower than in the age of building the consumer internet. And no, this isn’t about intelligence, it’s about understanding the market and the audience. Most current AI companies are just another wrapper around an API or an omg-cool model. And there are also pictures that nobody’s figured out who needs and how to build a business. In general, searching for a business model has become a problem.</p> <p>/thinking out loud/</p> <p><a href="https://t.me/chillhousetech/847?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So I turned on some little stars in the channel as reactions. https://darkolorin.com/posts/846.html https://darkolorin.com/posts/846.html Fri, 16 Aug 2024 13:00:06 GMT <p>So I turned on some little stars in the channel as reactions. I absolutely don’t get why any of this is needed, we’ll see. Maybe someone will drop one and I’ll find out.</p> <p>P.S. Thanks to those who did. Turns out you can’t “cash out” less than 1000 stars. Wonder what the exchange rate is? Anyone got info?</p> <p><a href="https://t.me/chillhousetech/846?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh, yeah? https://darkolorin.com/posts/845.html https://darkolorin.com/posts/845.html <p>Oh, yeah?</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/845?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Quick tip sparked by a recent call. https://darkolorin.com/posts/844.html https://darkolorin.com/posts/844.html Wed, 14 Aug 2024 17:52:39 GMT <p>Quick tip sparked by a recent call. A few tips for cold messages/emails:<br> - Keep it to 280 characters<br> - Avoid extra pleasantries at the start because they take up a lot of space<br> - Phrase a specific ask<br> - Make the message relevant to the recipient (turns out you have to figure out who the addressee is)<br> - Avoid excessive bragging because it takes up a lot of space haha<br> - Don’t push from the get-go​​​​​​​​​​​​​​​​ (gimme 200k pleeease)<br> Your captain (c)</p> <p><a href="https://t.me/chillhousetech/844?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A few bits of news from the AI fronts: - Anthropic dropped context caching… https://darkolorin.com/posts/843.html https://darkolorin.com/posts/843.html Wed, 14 Aug 2024 17:41:31 GMT <p>A few bits of news from the AI fronts:<br> -<br> Anthropic dropped context (<a href="https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching</a>)<br> caching that really helps save money<br> - Elon’s XAI are testing<br> version 2 (<a href="https://x.ai/blog/grok-2?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.ai/blog/grok-2</a>)<br> of their model<br> - Grok version 3 is expected by the end of the year<br> - imo soon all models will be the same in quality and the differences will be in the API and the extra stuff around it</p> <p><a href="https://t.me/chillhousetech/843?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Quarterly update of roundup reports based on U.S. https://darkolorin.com/posts/842.html https://darkolorin.com/posts/842.html Sun, 11 Aug 2024 19:42:35 GMT <img src="https://cdn4.telesco.pe/file/VRbhEs0kmn8wT4855IRngkNbse3s8WMxqrPqcqAiz1aEQ0FWTrM0mMH_poaWJEgxCWzyJgLN74jUJ6V3uttX19OXChl7P3gO7Nbu2SDWpn3Amf9ujdSR4iJRqKf8ndf7HjTJXClPuvjDCW8L8ISY58uOScnE8tDrEDB7VXsuNVHb3HY3grLIfJur19Ds0y54xjl2b2HUJs5JE0wpuUzyPhK48RN-ig9pRb4isd76Bh68zDFEgQuGQsJG9imQfpOVJldhIHZ5GPW6mz0PeGIgTs_Vj2lfm6RnZApnlZFgvRBy9zIWky_tzqj5V7SbaaUfmrzF8UjsocdUdhxNezTbnw.jpg" alt="" loading="lazy"> <p>Quarterly update of roundup reports based on U.S. startup round data</p> <p><a href="https://t.me/chillhousetech/842?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Bezos-style hiring system. https://darkolorin.com/posts/841.html https://darkolorin.com/posts/841.html Sat, 10 Aug 2024 22:13:06 GMT <img src="https://cdn4.telesco.pe/file/iol41mw4VsIadUHcQdW3ZE6SNDNHadheNcHH1fJSrkbgaIrecdnlis0sU6jX2pgNyDsVlGDdSgQrgXPJ8-BJZHw6YGcQWZ2Oqo2cYIZeNIf6aEHI8R9mo29qUMPF-49W9Kzn3Ba4taiWijuEnfJI7DNlVvJGChPrAetHBT6avIpZ97MWsZ47RQq31cGb0FW5MOh49shvx979T_nbfFtaVXFTuDhUFRfbnhXIT9IW-E33nH0Qs1E-gs5k3gVS_t9jpaP2mm2YWSCbVM0hFV63pS__6TObcplJ255XKIRgsV25LuH4Nf5TjIlZK9B4HHIy-h_YZfUJuXkmn-8fKxNQtw.jpg" alt="" loading="lazy"> <p>Bezos-style hiring system. In many ways I do the same, having never even seen this system. But it’s also important to add my two cents: I try to find in a person I’ll be working with an insurmountable desire to build the product they’ll have to grind on (usually even in a pivot case).</p> <p><a href="https://t.me/chillhousetech/841?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A freaking awesome article came out (don’t take the epithets and metaphors to… https://darkolorin.com/posts/840.html https://darkolorin.com/posts/840.html Fri, 09 Aug 2024 21:45:03 GMT <p>A freaking awesome article came out (don’t take the epithets and metaphors to heart) about how Bumble, bought from Andrey Andreev (Badoo, Begun, etc.) for 3 yards, collapsed.</p> <p>And of course the main takeaway for any B2C initiative/direction/company is that the people running the show have to “skin-in-the-game” understand the use case and understand their target audience really well. Yeah yeah, swapping them out for “pro managers” won’t work.</p> <p>Put together a thread-from-Twi as a post for you.<br> <a href="https://telegra.ph/Bumble-Crash-by-signull-08-09?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://telegra.ph/Bumble-Crash-by-signull-08-09</a></p> <p><a href="https://t.me/chillhousetech/840?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> New Product Manager on the project https://darkolorin.com/posts/839.html https://darkolorin.com/posts/839.html <p>New Product Manager on the project</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/839?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A question came up. https://darkolorin.com/posts/838.html https://darkolorin.com/posts/838.html Thu, 08 Aug 2024 12:32:40 GMT <p>A question came up. Have you ever bought consulting, and if yes, drop the hourly rate (even if you haven’t, but would like to — put a rate too).</p> <ul> <li>Never bought it</li> <li>$100/hour</li> <li>$200+/hour</li> <li>$500+/hour</li> <li>$1000+/hour</li> <li>I think it’s a waste of money and time and I don’t buy it</li> </ul> <p><a href="https://t.me/chillhousetech/838?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Couldn't resist https://darkolorin.com/posts/837.html https://darkolorin.com/posts/837.html Wed, 07 Aug 2024 20:13:45 GMT <img src="https://cdn4.telesco.pe/file/YxT2on8F4FIee1VYOu6fjaNOdCwhQxwv77KSiN6VN0YXCLna5yv5mG4Dh__ZlPHH_uyHM9nlW-IlNv9dX8tha2jwXpxv8PbwcRLy5L2qASYKqjPkyRs74RG14HGb-6eAu-FBXXSxvwqWw-d5nr-8bao4eX4mU3J-l9TWShoJw-XQeK8oM_lXGecxYt7qpHsWypfFlJz9_-CES0N3V8_I9gDLNCl5mURHlI-7o5OAsXe6OC8aJCKO5zukhvM8SrDc2quMqk_jczd9skw5EO3HHbKFxX7d8r14TtiSG9O-g7ZIBEb-fTEsBa6PnNweH5XPqW8yNaXoYa7RYXLddxcjGw.jpg" alt="" loading="lazy"> <p>Couldn&#x27;t resist</p> <p><a href="https://t.me/chillhousetech/837?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Basically, when you see “experience” on an entrepreneur’s or product person’s… https://darkolorin.com/posts/835.html https://darkolorin.com/posts/835.html Tue, 06 Aug 2024 23:02:12 GMT <p>Basically, when you see “experience” on an entrepreneur’s or product person’s profile with numbers (like I often run into — “grew the product by a whole 10x”), the best thing is to ask: what product, and from what little number to what, and even better — the name of the metric they “grew.” And if there’s no answer, or there’s a long story first, just run. Usually the results are extremely sad, and all those “achievements” are for “selling my hour for more.”</p> <p><a href="https://t.me/chillhousetech/835?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Getting back into the groove. https://darkolorin.com/posts/834.html https://darkolorin.com/posts/834.html Mon, 05 Aug 2024 22:03:21 GMT <p>Getting back into the groove. Slowly but surely, and with a very interesting piece of news,<br> (<a href="https://www.theverge.com/2024/8/5/24155520/judge-rules-on-us-doj-v-google-antitrust-search-suit?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/2024/8/5/24155520/judge-rules-on-us-doj-v-google-antitrust-search-suit</a>)<br> about how from now on Google might get banned from deals like “make us the default search, here are our billions or a rev share.” Curious.</p> <p><a href="https://t.me/chillhousetech/834?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A few insights from Apple’s ML team. https://darkolorin.com/posts/833.html https://darkolorin.com/posts/833.html Mon, 29 Jul 2024 19:06:36 GMT <p>A few insights (<a href="https://machinelearning.apple.com/papers/apple_intelligence_foundation_language_models.pdf" rel="noopener">https://machinelearning.apple.com/papers/apple_intelligence_foundation_language_models.pdf</a>) from Apple’s ML team. <br> The input data includes various sensors and breathing and audio, apparently. There’s a local little model (still unclear how good it is, the benches are ok). I think this will make it to production next year and only on the last 2 generations of devices (because memory, as I understand it).</p> <p><a href="https://t.me/chillhousetech/833?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A post about working in tech. https://darkolorin.com/posts/832.html https://darkolorin.com/posts/832.html Thu, 25 Jul 2024 10:33:51 GMT <p>A post about working in tech. About vision, building, and all that. <br> One of the most common problems in startups and corp — we’re building features, but we’re not moving toward the end goal. And the чистейшая reason is the lack of a goal or vision.</p> <p>What I usually observe (when I come in to talk about a startup or into a corporation):<br> - in a startup there’s such volatility of ideas and what actually needs to be done that a lot of founders start, after a round or even during fundraising, doing social media posts, pumping their personal brand (ew-ew-ew, because a brand is built by results way more effectively than by content), thinking that visibility (being seen) will help solve product problems through network (and building that network via posts is harder than it seems)<br> - in a corp a top manager/founder/cool leader, not understanding why who is doing what, starts throwing in an endless amount of genuinely interesting and potentially high-upside ideas. And chaos begins, because the top-down request is “important” and you basically never manage to push back<br> - the startup simply, physically doesn’t understand how to get closer to the future it’s selling in the deck, and then a persistent feeling is created (with the help of leading questions from investors and the team) that the founder just won’t pull off what they’re pitching</p> <p>Run into this — drop reactions so I can see how interesting the topic is. If it lands — I’ll break down in detail how I solved this stuff in my company/team.</p> <p><a href="https://t.me/chillhousetech/832?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Ha, the race goes on. https://darkolorin.com/posts/831.html https://darkolorin.com/posts/831.html Wed, 24 Jul 2024 17:05:13 GMT <p>Ha, the race goes on. Welcome <br> Mistral 2. (<a href="https://mistral.ai/news/mistral-large-2407/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://mistral.ai/news/mistral-large-2407/</a>) <br> Apparently it’s good at coding and reasoning. Waiting for tests with users.</p> <p><a href="https://t.me/chillhousetech/831?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> OpenAI is spending on a staff of ~1500 people $1.5B USD a year. https://darkolorin.com/posts/830.html https://darkolorin.com/posts/830.html Wed, 24 Jul 2024 14:58:15 GMT <p>OpenAI is spending (<a href="https://www.theinformation.com/articles/why-openai-could-lose-5-billion-this-year?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theinformation.com/articles/why-openai-could-lose-5-billion-this-year</a>)<br> on a staff of ~1500 people $1.5B USD a year. Well. That’s probably the most expensive team in the world I’ve seen. And then they burn another $3.5B USD on servers every year. Waiting for the next round.</p> <p><a href="https://t.me/chillhousetech/830?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One-liner: release of the new llama 3.1 version from Meta. https://darkolorin.com/posts/829.html https://darkolorin.com/posts/829.html Tue, 23 Jul 2024 15:36:48 GMT <p>One-liner: release of the new llama 3.1 version from Meta. They say it’s on the level of chatgpt &amp; co <br> <a href="https://llama.meta.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://llama.meta.com/</a></p> <p><a href="https://t.me/chillhousetech/829?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Mark is, of course, one of the best CEOs. https://darkolorin.com/posts/825.html https://darkolorin.com/posts/825.html Mon, 22 Jul 2024 18:23:49 GMT <img src="https://cdn4.telesco.pe/file/X4X8iKXq7HKI2tFgZ1vhzBT8AdCE31RkbCl4kHQ5xeVS0I95CYIviO5W5-WJSBCCOdWeARII6OU7gE5oct3QL0MxUYwypdhjvAkl1p3mbqCXp_ouwnDqSlgCv19R2Q7iGrzyl1CbZRA7hjMYdlOWYJqClwcn6eABmdIios3yoMS4F2TztSBK4h2tD2lgsA6YB9CsTZFNwWWxLHDuBDNpZfYWi0rcnqAqkMNHyUPZjewmSw2oN5_f9SP8xy2rOkZSBBDxEPQgG-B5OO6D-OvVZeR0L-_UpSp2ADVZb7Hr10vhWaY1qK4LbtAfSlJbGtM-0tDpGfDQnBFYFlO6uj3q7g.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/IA2QDw7gDnn00WhKSOxPjzRnWWhKipZ7h4g6mlAI5H8YrcaezZJObEHmS0Luz7223YdTTwtWYj0OZC7kBHMjDfMPV9AtIBC1MqKWvDQx9Ju3ndmRvA_0dP09ar1QiV7qHoj-0XtfmYl6jFCFwa9jk0zTHxfJgNoFbrmXquyx3bp4idXi3XZxROtWNq_WbxYMZ8BgdJ9uKDHLBnJBzXIKu6BkiaMYws29sUZguAl246SPM7dbgptN_F5cdAVSPBZ5drvtaRR4fzunvkgPk_IG0aXV1sb6gYI7FxY6YMg2809EwBUftRQfkOqUthHT_ep-Zgh3nVfhntz8xzk-AsV6lg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/DSoBQy-lqZyDoqwNxh1v01cZs3VTS_m-FHtleKuTu4NhJjfuTSjBM6wGHRtsEr7vpa7zYaSHwsRL1g8lZ9DLANpPNlj96wzPIlF-OVdgkSR-7C67RBC0UTmyLE9tDbeG9UqSawBLtheixtAIRZqFgmXJbSkPxQDYQzW5sq1WbF1kjtSpA5xaYSGh_bmuzrpVkEW_ckF1Xlx6oRb0HcO7PetbEt8YSi81SbrwgdsS-SaNeYJgaKeZY-XBhNFliW-uXzM5ZD_LK0W2bhAxnf7STNFF08LoLYW6sm5MFZs5H9NAOCV4WWpJjz49Mdtmrityf_MHxxFRz-YRoel46dh3jw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/n78xArAZXhOlwLmVsPeOkphjRXJ_6erGkth_3oRA5gxDnVj4JM5Yqn1NJkECuym3RquORZeBawMGIzG-Kz1dd9a6nGcpQTtNoy89ymfthb--0p0N43u518xHNL9Wlbzb_7gYPGF6FB7R5yB39_syeHMA5hB5mVgUgL2ZTzmaHjczB8eps4gS1D4UbNUjXE0imAAnjX9cxurlF4yJGCDuSP0HLh6-5IwIg5OFGBtybRKTnR9XrYjgvKYFNb7OQZgW-szHwritqi82d_gZenSi2TGFf-qlYra7UvQZIMxvOI7yQ4_SCTrquDSiCG5S3eyF2rvh1XdtT3lV6zse5jeePQ.jpg" alt="" loading="lazy"> <p>Mark is, of course, one of the best CEOs. Whatever anyone says. Look at the company’s goals in 2015 and what they look like now. A very correct and realistic feel for the market and the numbers you can actually hit.</p> <p><a href="https://t.me/chillhousetech/825?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Post one about deals, M&A, startups. https://darkolorin.com/posts/824.html https://darkolorin.com/posts/824.html Sun, 21 Jul 2024 15:30:19 GMT <p>Post one about deals, M&amp;A, startups. <br> There are two types of deal: <br> - the startup gets bought <br> - the startup gets sold</p> <p>Most deals on the market are actually the second type and it’s important to understand that. Let’s unpack each a bit.</p> <p>“The startup gets bought” is when people come to you on their own, without you putting in much effort, and start offering M&amp;A talks (merge &amp; acquisition, merger and acquisition), like, they want to buy you. Usually there are a few tiers of what exactly they’re buying:</p> <ul> <li>deals up to $100M USD (from here on everything will be in dollars): this is when they mostly buy the team and its expertise in building something (product, sales, marketing, etc.). For example, the acquisitions of Nikita Bier’s startups fall into this category. Same for me.</li> </ul> <ul> <li>deals from $100M to $500M: this is already buying users/contracts/something else that’s super important for your company, they’re buying the so-called North Star Metric, roughly speaking, and its growth rate. Don’t forget the team often comes as part of the package. The brightest recent example is BeReal.</li> </ul> <ul> <li>deals above $500M, with rare exceptions (Instagram &amp; messengers): this is buying a business. In those so-called exceptions it’s usually a very clear understanding that they’re buying a market leader (Twitch, YouTube, etc.) and an understanding of how to use that leading position for the buyer’s business.</li> </ul> <p>They come, they buy (how the deal gets run will be separate), then KPI &amp; 10–20% of the deal amount paid upfront. Rarely more, even rarer the whole deal immediately in cash (Twitch).</p> <p>But in reality most deals are when you need to find an exit for your startup and sell it. Here it matters more who your investor is or who you have on the founding team. Because it’s exactly the network and the ability to find where your team and expertise are needed and to execute that move fast — that’s the whole point. Selling is way harder than haggling over the price when someone is buying you. So this is where all the qualities of the team show up, and especially the funds that gave you money. In my case in the previous story, when I was doing Capture, the partner from GC very quickly realized we needed to sell and did something like 10–20 different meetings where, in his opinion, we could strengthen a team or expertise (by the way, I think we could’ve sold, I just really didn’t want to go to that company back then and I decided to shut it down).</p> <p>My advice to those of you building something of your own — think about how you’re going to do an exit and what you need for it. This isn’t for your deck, where you sell the future. This is so you understand which investors and co-founders, besides product-tech ones, you need.</p> <p>And that’s exactly why more gets invested into experienced founders.</p> <p><a href="https://t.me/chillhousetech/824?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Basically I’d be too lazy to post all this and go like, look what feature I… https://darkolorin.com/posts/823.html https://darkolorin.com/posts/823.html Sat, 20 Jul 2024 11:57:47 GMT <p>Basically I’d be too lazy to post all this and go like, look what feature I dug up that they’re testing. Found this pretty damn nice resource, (<a href="https://www.socialmediatoday.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.socialmediatoday.com/</a>) for anyone who’s into keeping an eye on what the big social networks are testing.</p> <p><a href="https://t.me/chillhousetech/823?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> All you need to know about the power and brains of OpenAI models — and… https://darkolorin.com/posts/822.html https://darkolorin.com/posts/822.html Fri, 19 Jul 2024 16:33:49 GMT <img src="https://cdn4.telesco.pe/file/HANeSxZj-swSmPj0oCGBxpFHkfa-wjFwKS2580WRyf_LDuEhHcArM778IG2jLEmH9jEfJ2iQs01SAf1ACY7Y7wWm9EgvIAGJciPH5p4inRx-LYJVTgPWrvy9Tb5lSQEevCfuaO0U5rOxgKlynTN5ey5steq-keXI0mgzxsWJa5Dqo5LtN2bwhRSWBhEl882VwjmiQFbm2-GyIXsucbgVvPBoWXYhGy7LrJZkgHrR1unYPwsrQMK96ucmbFgFkCDqbYjuUpbAVkXfOyh64QUE4SkEhBjUHHRYavSx3QJGzvvhtecjTd_IkE-QjFrGA-NrvSjy9y0uSMHTSY5-OmMCxA.jpg" alt="" loading="lazy"> <p>All you need to know about the power and brains of OpenAI models — and especially about math benchmarks</p> <p><a href="https://t.me/chillhousetech/822?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Thursday, time to share some news. https://darkolorin.com/posts/821.html https://darkolorin.com/posts/821.html Thu, 18 Jul 2024 19:20:28 GMT <p>Thursday, time to share some news. (<a href="https://venturebeat.com/games/xsolla-acquires-lfg-to-build-gaming-connectivity-and-commerce/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://venturebeat.com/games/xsolla-acquires-lfg-to-build-gaming-connectivity-and-commerce/</a>)<br> Proud of this deal and my second exit.<br> Worked on it for a long time and put in the grind.<br> Thanks to the team and investors for the trust, and to my close ones — for the support!<br> A lot of work ahead together with Xsolla!<br> P.S. For everyone who wants to share the joy and spread the news, please don’t forget to include a link to my channel. Thanks.</p> <p><a href="https://t.me/chillhousetech/821?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile Andrej Karpathy founded an edutech I like the site. https://darkolorin.com/posts/820.html https://darkolorin.com/posts/820.html Tue, 16 Jul 2024 18:59:20 GMT <p>Meanwhile Andrej Karpathy founded an edutech <br> <a href="https://eurekalabs.ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://eurekalabs.ai/</a> <br> I like the site. Pretty sure it’ll be useful. And I think he’s also building another company doing AI for embedded. Gonna run neural nets in fridges. <br> Curious how fast they’ll start copying his courses and the whole approach.</p> <p><a href="https://t.me/chillhousetech/820?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Got it:) https://darkolorin.com/posts/819.html https://darkolorin.com/posts/819.html Mon, 15 Jul 2024 23:10:52 GMT <img src="https://cdn4.telesco.pe/file/kK8JEcHeKDYMrevPKBk64usG62exgCHOvHeG3QOiyDFTnExtblSZcSOswSg9JIhQfvtKnRb7E-byYvbOv24nmSAlJ3U71hJ66d-CCiqjh_ptntCqD1rdanq9scbTj-Ef5L14cX8cmE25kF0kM2iLjW9ldjw9_BH37ViUiJTyJn2neWRd3xXSipfiYI2IYoiTQkwqiyuVak6eseBvpRJ7uIEi8g4S2B2kMmIXETqacaU3gvQtTrOxEfejmb8WjSRaYeVbP09dW0xmhWmKE2e6Rf1YafIyIX_vPb5m72Ko6JVwzceQ1l35WUdQrfpAOJLa9vq18f4OOBM7whlcPQep2A.jpg" alt="" loading="lazy"> <p>Got it :)</p> <p><a href="https://t.me/chillhousetech/819?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Reminds me of a crypto pump and that whole thing with BTC’s price https://darkolorin.com/posts/818.html https://darkolorin.com/posts/818.html Mon, 15 Jul 2024 16:21:26 GMT <img src="https://cdn4.telesco.pe/file/dzW6xVQ0q8whJPqTu-XQRR1oHTn2OxWl8nNnxxhD0Rp8h-qEjp8QsXa7BL7Y7BgEU7qX9--F73ouMS0zuqv91uilKx8TgXG0Q3HVgsrpgzdDnxxhA7NO2n4gSQ8VqHelXLNrMiwhJ-ZM29O-wNz-vQ8gv7T4V60aO8OBSvr8yIT04XOMgzai16rhp8lw_Qi1euZgg29XsVh9ZmGjNONS2FrBkowPL0IB5ORWKpeu6cqtlkZtHLECf2taN6ISizWeZjAxpSHYLn_4NrYWsi8xmY9GJPCUr-D7uh76rrJXjfILTFZ4hCs47pZhlGNnfBsC1DH4wNhJSz8bw7o7L9JKeA.jpg" alt="" loading="lazy"> <p>Reminds me of a crypto pump and that whole thing with BTC’s price</p> <p><a href="https://t.me/chillhousetech/818?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Your guesses for what the annualized revenue will be in a year (that thing… https://darkolorin.com/posts/817.html https://darkolorin.com/posts/817.html Sat, 13 Jul 2024 12:42:32 GMT <img src="https://cdn4.telesco.pe/file/B4CtQqgE_DzgudZr0vtVIjZIq_tp5VIIj5H8yhEFrEuJDbCLGY4A8-aYWef7y06FIqqL3y52F3TuIbPZAdjSz-mEweJOWdF12y0-_5OkXqvmfDy9gPpm1FQSRJukE9-S9rki5xKuNH8u-jgD-pHkcs9j_xcWrjQG-YBslHtR2m4_KTyiHiJfq2VT5Mi4qberz5bC1Zxk2RkUEE1ZGIj4TVsRn8pRyOu9JaeKEDEcT0fbpbS-E2ZMbQqEcrZ6gpwN5dDa165aIIx3UQ9pxLbs0_Ednqq-AyyaR3RGhGASXi59sqvJd3y_E4wh4pOMysmkUx2NCAH1_rmK2fQJYXAPaA.jpg" alt="" loading="lazy"> <p>Your guesses for what the annualized revenue will be in a year (that thing where you take the current month and multiply by 12, like churn is definitely less than growth, sure).</p> <p><a href="https://t.me/chillhousetech/817?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> While I’m writing a useful post, I decided to write a useless one. https://darkolorin.com/posts/816.html https://darkolorin.com/posts/816.html Tue, 09 Jul 2024 18:35:30 GMT <p>While I’m writing a useful post, I decided to write a useless one. <br> I started noticing that a lot of people from the industry (Russian-speaking tech) start channels for themselves and at the very beginning run it openly, honestly, post often, barely prepare any content. Usually those channels are interesting for me to read. Then the guys gain an audience and it turns into yet another channel “about this” and just template posts from SMM managers. And at that exact moment a second channel appears, like, where the person is again writing honestly and easily what they think. <br> And why the hell, sorry? What’s the point of doing it like that at all? I’m genuinely interested in your opinion, especially since there are a lot of people with big channels here. <br> One of the reasons (probably not for everyone) is that for many a channel becomes a significant source of income.</p> <p><a href="https://t.me/chillhousetech/816?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Here’s a fresh launch for you. https://darkolorin.com/posts/814.html https://darkolorin.com/posts/814.html Tue, 09 Jul 2024 08:57:37 GMT <img src="https://cdn4.telesco.pe/file/FqI43K9SWysMSY2jVlMh7Oz_kdEPthvYzHfHX-PvRZONYhIseoa_tosnv4pyflHar1mdG0i3gjY93HZdoWZ_xOMISY4kYmTPHyCwbmIVf87Kf6ekpunAVDuwgftjFOTvUSyW0DuM4gHsEWPICDgdzpNeMsrEBhu1v5WSH71kjh8dcSTOErZUmnxXjp96LK06ewF5okz6MH0lAJWC9BCoV5NSYHSMly-mkfe6RQLlMqKGhdUj_91CQW-TtjofqQzou-vUIfLW5v16J-5kEf9viuVrmrnTdOTHTFOCaZzZC7cvuRQaxUffq9DxgOzFrOCp1blIylQn8S55bJx6l8hETQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/o1HgtE3-GzAYDf9nHpTLcVlR5vvJDpps7ydcS2pQveel_ik-ShpwLacKc6Dxug3a_6ESVrgg7pzX4T3WFDRO4_ytt9winYPUAoBj4CAnx4c2JO6nY51xweXcN5aTIQvJ6INl-RVdneXgXsiqnAG9xx2Qz56Ypo7NoM67ccgIy7JGAY8Of0TUXKfBQ62dUEg3ZZfx9MtwusQ6A1zi6bLWo98S9zM5w7Imi921uc2cG_vRMXAq_Uz8qjiDVFWe4YRco-Tf0UVszmeqgj0H0ckme237WWXaw5jbAtuI5oXzzOVeA-kbvlJ2vJv-FfjTfNlie0_hZ9P2uvKqRoFWY9E_4Q.jpg" alt="" loading="lazy"> <p>Here’s a fresh launch for you. The guys at <br> <a href="https://runway.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://runway.com/</a> <br> clearly decided to chase some hype. They mailed some of the most well-known authors on Twitter bags with combination locks that would open when the platform launches (in less than 24 hours it worked for everyone who got one). And of course they posted the whole thing to their accounts, and now everyone’s discussing what kind of platform this even is. Nice move.</p> <p><a href="https://t.me/chillhousetech/814?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I came up with the perfect coworking space. https://darkolorin.com/posts/813.html https://darkolorin.com/posts/813.html Mon, 08 Jul 2024 22:46:06 GMT <p>I came up with the perfect coworking space. Meaning this kind of place where you can book meeting rooms that are enclosed in a Faraday cage (<a href="https://ru.wikipedia.org/wiki/%D0%9A%D0%BB%D0%B5%D1%82%D0%BA%D0%B0_%D0%A4%D0%B0%D1%80%D0%B0%D0%B4%D0%B5%D1%8F?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ru.wikipedia.org/wiki/%D0%9A%D0%BB%D0%B5%D1%82%D0%BA%D0%B0_%D0%A4%D0%B0%D1%80%D0%B0%D0%B4%D0%B5%D1%8F</a>). I’m sure your brainstorms will be super effective and everyone will be very attentive. You can’t leave the meeting before it ends.</p> <p><a href="https://t.me/chillhousetech/813?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> From the world of amazing statistics. https://darkolorin.com/posts/812.html https://darkolorin.com/posts/812.html Mon, 08 Jul 2024 13:05:51 GMT <img src="https://cdn4.telesco.pe/file/TF5rtLDCcU_bCDo_BOnD2brJTWNnpGw_n2tnWk8W79zrBEU7Ty5RFGitXkYVYxMvvHeRIr-868g6S4rsY3E9XwIPnCCRZs-zeYq4-LVyvake3HAh9UcAKufsL1v3Ge57jI_Q3ps-icsQdI5kAucr9I8tRRJAaw8uNw7XVpiin19JauAMZ30GtJ_YsNZ0tRFKCwuk19XFKfZ9ccWnqAxQpal5rhUGk1zh_eGJXOZyAM1ZtpKZRA7ltcELWt-P1ISBnfP8ntG-RCdMXzPSUYRtSj3O76HXT-T35PjXxdZdA0I4xIzpS4EBO5hFG2wDvi2p2g6mk72yBfqIapDilUaXTA.jpg" alt="" loading="lazy"> <p>From the world of amazing statistics. <br> So shorter means better? Some interesting stats on startup names:<br> - Short names = higher chances of getting funded<br> - Seed stage: names &lt;6 letters make it to Series A twice as often as 15+ letters<br> - By Series B the gap is even bigger<br> Only 25% of startups change their name in the first 2 years. Maybe it’s worth renaming your startup? :)</p> <p><a href="https://t.me/chillhousetech/812?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If you suddenly need to pick a model for your tasks, check out prices, speed… https://darkolorin.com/posts/811.html https://darkolorin.com/posts/811.html Sun, 07 Jul 2024 13:52:26 GMT <p>If you suddenly need to pick a model for your tasks, check out prices, speed, providers there. I recommend <br> this resource. (<a href="https://artificialanalysis.ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://artificialanalysis.ai/</a>)</p> <p><a href="https://t.me/chillhousetech/811?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ve been saying for a while that Google’s strongest competitor in the search… https://darkolorin.com/posts/809.html https://darkolorin.com/posts/809.html Sat, 06 Jul 2024 22:57:54 GMT <img src="https://cdn4.telesco.pe/file/FNAHaQw6gzl2n1-Gch_H1wzBHXrLKLKeK9F8TYLPYLtJmdAJ7JTaU1oj3fdvDT-DwtuwKvByuiUZgnCjU6kV5T4mRrAk7LDO_A-Sz8Vf_XtvqrvBWChMKwwkqdTiWO1mF4E_By2s473kGrl_AMKq0ixBeUrue8_p-HD_noMY_Pg-kr__6sHdu_smKk_0Yx_L5inXcd10zlI6oIDYOX2dZnzequgeXFUVHeZqEZvRKT9HpW0PcVYdlJnmyqf71wKgU54NuClaRw0ZYK2_m9BcVLr6-1oy7azr8LufWuYn5IDtTEe4KeNaNWgwXTv16Mla_6xfqOW3dBvuNKM6CxQkMw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/nIbu-zuS4LrKHucLnqmokB8SamNa6aeltfXV6Y0Ohx5S6xZDQ_EkEM2vQwtcaTitxloWI3Jiwutatii8B2yrHKlZ5hFhChinUgv_j7AJ0geMPGqnsNUHCi5Y2PasLy-omdAtug0ptWTefyb6gy-a1bqHARRJ6FLGhXh7gdHvDS4PAp8QLHBvNu_ZDn_vaLaI4mYXZqyel3PK5abSC6271qmdqETRnwh6Et8xz7b5L95i2SME9ic13SUOoGj4UgkO4DjqkB1WFh_f8lhRib2qc6-OmEY1fu9XF2DARjRkTV5KqbB2jHqwCjkksT0-2Jm3IhFKHnXbX3uiQGV5gJTz6Q.jpg" alt="" loading="lazy"> <p>I’ve been saying for a while that Google’s strongest competitor in the search market is TikTok. And finally, we’ve lived to see the release of their product.</p> <p>Now in search it summarizes videos right in the results and highlights the key moments. Exactly what you need! I already pretty often look for restaurants or cafés through TikTok.</p> <p><a href="https://t.me/chillhousetech/809?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Though. https://darkolorin.com/posts/808.html https://darkolorin.com/posts/808.html Thu, 04 Jul 2024 12:08:20 GMT <img src="https://cdn4.telesco.pe/file/a8JEckxjr8vcKvxvnwO5jM9a49leUff6HC8Sc2PTv_eTVkQfejMT9CUaqMQiSio6sAJAa824Hum1OhZfsZ1GGeYj9weYGkiHtwDFGi9hhxR5Teu9pyyq_E82h9q4PPyr9ksjc_n_ZA7z0OnrC2vmcjFVCrOzM_XTCL4JBeSra_ICj0MU0vMnYWWEKiE_JvAvRd_qJZGTWf9pVasxhbwHSte1Y1Pf_y4Lf7I6CyWsvlhFQr5Yxt_rsQwiOnVH5owTftCcdIbsD1pU_5R5qpQGsOcSjwIV7ImEQzlorXpyes7WNHbPzVZSadOr-6K6oLvF4q7KfeHDkq-H0yk_v7BG3g.jpg" alt="" loading="lazy"> <p>Though.</p> <p><a href="https://t.me/chillhousetech/808?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The reality of video generation. https://darkolorin.com/posts/807.html https://darkolorin.com/posts/807.html Wed, 03 Jul 2024 16:19:00 GMT <p>The reality of video generation. 1 minute of waiting, 5 seconds of video, $1.5 price tag, 10 artifacts and bugs inside. <br> And it’s still not clear why. But by my rough math, to put together a successful video right now you probably need a couple thousand dollars for a ~30-second clip. I think that in 3–4 years a clip like that will already cost about a buck. And it’s still not clear why, other than automatic creative testing at the exchange/DSP level. What other big markets are there?</p> <p><a href="https://t.me/chillhousetech/807?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, yet another social network for friends and statuses. https://darkolorin.com/posts/806.html https://darkolorin.com/posts/806.html Wed, 03 Jul 2024 14:26:27 GMT <p>Well, yet another social network for friends and statuses. Top 1 in the US. <br> <a href="https://apps.apple.com/app/id6478601246?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/app/id6478601246</a></p> <p><a href="https://t.me/chillhousetech/806?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The authors of that very benchmark have rolled out routing for your LLM… https://darkolorin.com/posts/805.html https://darkolorin.com/posts/805.html Mon, 01 Jul 2024 19:51:35 GMT <p>The authors of that very benchmark have rolled out routing (<a href="https://lmsys.org/blog/2024-07-01-routellm/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://lmsys.org/blog/2024-07-01-routellm/</a>) for your LLM requests, so you can figure out which one is better to hit with your prompt.</p> <p><a href="https://t.me/chillhousetech/805?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, weekend thought. https://darkolorin.com/posts/804.html https://darkolorin.com/posts/804.html Sun, 30 Jun 2024 22:14:37 GMT <p>So, weekend thought. For some reason all these big LLM providers keep looking at this benchmark. (<a href="https://chat.lmsys.org/?leaderboard&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://chat.lmsys.org/?leaderboard</a>)<br> Called ChatArena. That thing where you pick an answer from two LLMs, and like you “rank” everything that way through pairwise comparisons, and then all all all the models get sorted out.<br> It’s considered this super honest leaderboard. But like, who’s actually sitting on that site? I seriously doubt it’s, well, let’s say “regular guys from the factory floor.” So isn’t there a bias in the scoring relative to the audience? Aren’t we assuming this sampling method is, well, kinda meh?<br> Or why do we trust this bench so much? Maybe there’s someone in this channel who can explain it to me.</p> <p><a href="https://t.me/chillhousetech/804?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If your product has more features (buttons and actions) than DAU, I’d suggest… https://darkolorin.com/posts/803.html https://darkolorin.com/posts/803.html Sat, 29 Jun 2024 13:27:17 GMT <p>If your product has more features (buttons and actions) than DAU, I’d suggest you think about it.</p> <p><a href="https://t.me/chillhousetech/803?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So far the hardest question for everyone that I ask in an interview is “What… https://darkolorin.com/posts/802.html https://darkolorin.com/posts/802.html Fri, 28 Jun 2024 11:48:50 GMT <p>So far the hardest question for everyone that I ask in an interview is “What happens when you multiply a row vector by a column vector”</p> <p><a href="https://t.me/chillhousetech/802?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Expect it in 2024, oh 2025, well whenever it’ll be cheap to do inference and… https://darkolorin.com/posts/801.html https://darkolorin.com/posts/801.html <p>Expect it in <br> 2024 <br> , <br> oh 2025 <br> , well whenever it’ll be cheap to do inference and everything won’t be crashing, and in 2026</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/801?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Alright, so for the first time someone came to me asking to buy an ad spot in… https://darkolorin.com/posts/800.html https://darkolorin.com/posts/800.html Thu, 27 Jun 2024 11:11:45 GMT <p>Alright, so for the first time someone came to me asking to buy an ad spot in the channel. Let’s be honest, how do you feel about it? P.S. I personally don’t want to do any ads at all for now, but I’m curious what you think and how appropriate it is.</p> <ul> <li>Yeah, fine, it’s ok because money (but not once a day pls like posts)</li> <li>Rarely rarely is ok, I won’t even unsub</li> <li>Horrible, ads in a 1k-person channel already</li> <li>Your option in the comments</li> </ul> <p><a href="https://t.me/chillhousetech/800?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There was a time https://darkolorin.com/posts/799.html https://darkolorin.com/posts/799.html Wed, 26 Jun 2024 11:13:58 GMT <img src="https://cdn4.telesco.pe/file/i8miw3345Zfofl6za0D64x1dOwBgTXsFgTAQ2DMA-k0WChLrH54NQKUVgHxy4UrjftYLL0-5_BvKWRHCHB1VtQ1zbRSXQ_M6fV_NyG_mPO9Q_inszyhDdijniO4zCZsr1FqlDea8e3pU0xkSrAZun1Kv9dTNfrLE5FPs1lyZPry-nYr_snU7fWmt7_hxekiv--XXQhASDhYZIYZZCCnKQQjSODKXjohUrCb4jGlT_IKgCJzij85D-LzDybeCi50PscKCLuR1GJyMoB0HlhRC36ozNJTJb67H_LTcSA-QGMk2KkYfYQ7GPKjlqsyKPz3NTONh7wMV67wMcjIiTxsPZQ.jpg" alt="" loading="lazy"> <p>There was a time</p> <p><a href="https://t.me/chillhousetech/799?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One-liner. https://darkolorin.com/posts/798.html https://darkolorin.com/posts/798.html Tue, 25 Jun 2024 13:55:46 GMT <p>One-liner. <br> Spotify is launching its platform (<a href="https://confidence.spotify.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://confidence.spotify.com/</a>) <br> for experiments as a service.</p> <p><a href="https://t.me/chillhousetech/798?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A post based on the results of my interviews at Google & Microsoft. https://darkolorin.com/posts/797.html https://darkolorin.com/posts/797.html Mon, 24 Jun 2024 21:36:32 GMT <p>A post based on the results of my interviews at Google &amp; Microsoft. Only 10 years later. Exactly 10 years ago Google rejected me, and later Microsoft took me (Yammer team).</p> <p>Back then there wasn’t this mass swarm of courses and other crap about Product Management, or maybe I was 23 and just wasn’t really in the know. But I worked at Yandex and thought I was hot shit and deserved to get into the very top of the top and to the top of the hit parade — Google. I didn’t really prepare, just sent my resume to a Switzerland opening with no specifics (Product Manager). Surprisingly, I got a response and the process kicked off. By the way the interview was dumbly over the phone and, what’s more, all 3 out of like 4–5 were over the phone — they didn’t take me onsite. 90% of the questions were so brutally technical that I realized: you’ve gotta prep and review everything Google does and understand how each of their products works. And also you had to storm all those algorithms in your head. The famous “how many balls” references and all sorts of other brainteasers. And I was just burning from it like crazy (a lot of people still burn over it now, by the way).</p> <p>Half a year passed, I realized I needed to prepare and went to Microsoft, same kind of role but the team was known, and this time I read Cracking PM Interview. And this time after 4 rounds I got an offer. But I declined for personal reasons — like, Microsoft isn’t that cool.</p> <p>But what am I getting at. Back then I’d rage and think why the hell these puzzles and all that, because you’re supposed to research hypotheses, build A/B tests or whatever else they say is important. And 10 years later I realized that actually the world of Software Development is about people (like everything else), and we work with developers, and it’s with them you need to build, and if you, buddy, can’t and don’t understand that, no knowledge of frameworks will help, but knowledge of algorithms will help — and how. I’d even say you can toss all knowledge of product management approaches in the trash, but if you can say what you want and build it with the team — that’s a win, the rest is details.</p> <p>I recommend this little book —<br> How Google Works (<a href="https://www.amazon.co.uk/How-Google-Works-Eric-Schmidt/dp/1444792490?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.amazon.co.uk/How-Google-Works-Eric-Schmidt/dp/1444792490</a>)<br> Yours truly, captain yeah yeah (c)</p> <p><a href="https://t.me/chillhousetech/797?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well I’d start by planning 7 right away, like what level of stuff can you bake… https://darkolorin.com/posts/796.html https://darkolorin.com/posts/796.html Mon, 24 Jun 2024 21:11:43 GMT <img src="https://cdn4.telesco.pe/file/kQy-6FlVaqHZwWaQKVAKqhZ8prOTWXLqby3WALJgoI8qPb7VGnPfAxA5M-BEeHhodmuWsovrZagVS8WTwUgoxT-YKe3epgtZu9yZY45wSBUOZAvezN0ViN8ZTSXP7nWzr7XdAuhU74cKhviYEpownXLqRhSDSw5JFlo2BSDdi7T8XwCo3tsfC7uFjdWAkkMvbrkLHVdQwlYalexIyohyi1VAG847ilwTkEU5q5r3gd0CA1s3IKniQBWmGhLPKkAzIhl9giEpn7m0NyA5TL8vYNrm2tJ-zDf1Y1ekm1uwZpqRDDobQOhkl0onxvl2BDaDFdk8M5EJ7lEGUzQUk47Clg.jpg" alt="" loading="lazy"> <p>Well I’d start by planning 7 right away, like what level of stuff can you bake in from the start?<br> Recognizing the Microsoft approach <br> 😏</p> <p><a href="https://t.me/chillhousetech/796?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh, now *this* is a legit thing. https://darkolorin.com/posts/795.html https://darkolorin.com/posts/795.html Mon, 24 Jun 2024 14:11:40 GMT <p>Oh, now *this* is a legit thing. (<a href="https://huggingface.co/spaces/TTS-AGI/TTS-Arena?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://huggingface.co/spaces/TTS-AGI/TTS-Arena</a>)<br> People keep asking which model/tech does the voicing like that and how much it costs. Well, here’s basically a table that lets you more or less gauge the quality. Would be nice to add prices too, but you can just go look for yourself and check that the top-2 model is way cheaper than Eleven Labs.</p> <p><a href="https://t.me/chillhousetech/795?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile Andrey Karpathy looks like he’ll be launching a course soon write… https://darkolorin.com/posts/794.html https://darkolorin.com/posts/794.html Mon, 24 Jun 2024 07:02:32 GMT <p>Meanwhile Andrey Karpathy <br> looks like he’ll be launching a course soon (<a href="https://github.com/karpathy/LLM101n?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/karpathy/LLM101n</a>) <br> write your own LLM from scratch and learn everything you can do with it. Highly recommend it, his videos seem like the clearest and most informative to me. That’s exactly what training courses should be like.</p> <p><a href="https://t.me/chillhousetech/794?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One of the early prototypes of Hamster Combat https://darkolorin.com/posts/793.html https://darkolorin.com/posts/793.html <p>One of the early prototypes of Hamster Combat</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/793?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Friday data on salaries in the US (per year, commas where they’re supposed to… https://darkolorin.com/posts/792.html https://darkolorin.com/posts/792.html Fri, 21 Jun 2024 20:35:01 GMT <img src="https://cdn4.telesco.pe/file/U_iW_R1o5ej3RGp70f3A1GOyBu9NkMZ6jxklkhWNd9tlno1TdArl7GRqD497hd9EtTjcuAV0EX_Q1KqKvEh34Ur5dLQpEH8KMY5yfDLI6KiyJvrEe4qEdIce05_b6Kf5tS-fBIKgTB6J5rA0DOqew7UqmCYLgOxOBFzYlrsxTHYgJBWggZKyWNSuwdQ4-y6ek_-d7GmhCxBQwS3SPIy-JOt9WdXCZRrsaoMttADOMGcGLaPPBUD7V9eIGrT1QGzBLEO0e4vfVEXo-7iPKHP6dJ8GFraQIOGvzCXLsLEtPXIR1CGc6ejQEh1DC-F0lPLPab273FIaId0hw-8XkAwJ3g.jpg" alt="" loading="lazy"> <p>Friday data on salaries in the US (per year, commas where they’re supposed to be, yes yes you didn’t misread that)</p> <p><a href="https://t.me/chillhousetech/792?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh right. https://darkolorin.com/posts/791.html https://darkolorin.com/posts/791.html Fri, 21 Jun 2024 17:18:53 GMT <img src="https://cdn4.telesco.pe/file/X3iw74Dhi-CnWAf1_3hMnsQv217tEk45HcMFGNfeTJfu5uJYk3iyHcoTOLL92jlBMHTWqGeLqEaTUI1fctT0q5c-Upib4cMGyKbIe1SCVxfKuGGbjsFSqBLUR3ucbLPzY0nXU4AO_YZ2uMKirAWqTudSLF8ydxNmvyAQbxg1HFHIstYVre01AmJDRAwGKrQsKJg40-cwAOoFLfxT8ATuOy09Cv-lnaFfI_yePBoGAroRdbQmWAmH8Gzi-D6rnBaqFjB7DSLijrXZi7NZph0hQq0XKXz1dqFT3nQ6UH5EEGTAnZo0nlo-5tBfLUJUeXW3xYacCl2Hz_G5MGb1ZhAfJA.jpg" alt="" loading="lazy"> <p>Oh right. And everyone else in 2026</p> <p><a href="https://t.me/chillhousetech/791?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And there go the acquisitions. https://darkolorin.com/posts/790.html https://darkolorin.com/posts/790.html Fri, 21 Jun 2024 15:51:29 GMT <p>And there go the acquisitions. You can tell right away the corporation’s growing. <br> <a href="https://openai.com/index/openai-acquires-rockset/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/openai-acquires-rockset/</a></p> <p><a href="https://t.me/chillhousetech/790?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The race continues. https://darkolorin.com/posts/789.html https://darkolorin.com/posts/789.html Thu, 20 Jun 2024 14:35:07 GMT <img src="https://cdn4.telesco.pe/file/T6jR29wbsvQpBHpUqPSVW0ABuY1nB6ck_H1m7Ivw36mEhB8IKlRlzEw_uXjlIzzQajKTD4gAt84QOmqZ8Pks-bQMcLyOCQTKgHqlNxTrmTiJLLm9BXFBsnOdnqfCKssPzBwx7xTuabvCV4wI-zo8mdktmRPwiCVUfkTd-jlbDw6HROQCLIxuN04ASgEtzj4FRrL6a2V2esdlDDHNDZAsE3SRRXXLGQuCMu6a8ZxNqtK8KtJMeYm6A7Kz5R_ojSk1b6hXnFPcF8fdD3soX6KzLLC-byFvVC_3zspYZJI7thZjRlToFAJpKSK0ueH7rFWImtAje6-EsKzy7IZLCqlzkQ.jpg" alt="" loading="lazy"> <p>The race continues. <br> Anthropic is pulling ahead. (<a href="https://www.anthropic.com/news/claude-3-5-sonnet?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/news/claude-3-5-sonnet</a>)</p> <p><a href="https://t.me/chillhousetech/789?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, meanwhile Ilya, who bailed on OpenAI, has founded a new company and is… https://darkolorin.com/posts/788.html https://darkolorin.com/posts/788.html Wed, 19 Jun 2024 18:35:38 GMT <p>Well, meanwhile Ilya, who bailed on OpenAI, has founded <br> a new company (<a href="https://www.bloomberg.com/news/articles/2024-06-19/openai-co-founder-plans-new-ai-focused-research-lab?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bloomberg.com/news/articles/2024-06-19/openai-co-founder-plans-new-ai-focused-research-lab</a>) <br> and is planning to build super-reliable AI. <br> Also, separately, I’m asking you to check out the site design <br> <a href="http://ssi.inc/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://ssi.inc/</a></p> <p><a href="https://t.me/chillhousetech/788?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/787.html https://darkolorin.com/posts/787.html Tue, 18 Jun 2024 17:18:49 GMT <img src="https://cdn4.telesco.pe/file/Y7FaJNMg_CD_6imLibFpij-uKt-DXheAm29hu-sVQl1iiAFwBrkMdSeSOWYlMOB1Qo-9f-aP_b2FBfHXqMVjhACbpsKDoWnIwhtZtwICPMaqA2UvJVK929lA1ciiBGatQiF0nGti1oW_RRxXoaaDu5wcn0OLxM2_qQ44H83ybwxbcUxtnD5KRhfuQU2FPOMyI0InXbNUYdcy_c1fpeVVAC64BFPcMbmS3g271gT11LmvB13EL6_WwL1HSDrYpp2Yy9M2UBPvvBAZHHT2W6uT9AtpyEbdPPnuw0dHB6hvrLP2IA2XNmk-jecbm5CMY2QOq5ex1Z46zjG0IhFGP3DLkw.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/787?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Need more gamification https://darkolorin.com/posts/786.html https://darkolorin.com/posts/786.html Sat, 15 Jun 2024 14:15:28 GMT <img src="https://cdn4.telesco.pe/file/poaBr3J5oHNejk98FfLho7zDBSWvDYipBAWiLlz9-bKnkl2HN6OquAcpOuKTE1osJJyg6_Oz9J-CJuyI9jX9vJh36_cEOAwEEwx40q7DhSju814Xt_XouS-2_Ecc9l0Prg6TlQcxf2jnW7B2UNIXvbKWsXkSAzgJbVo3yg6Q9yx0kd_5sN5-XaPdUI_yiDMjFSugMBqHbxFQEfZBRnyAiCwTCCTeg87YBBeNVbdyS1ikyUodDPKz9FwF-4xp1osT8otIs-4BQ-UKC1iL7wHIddGX_sM6GMZj4IY-J_kQX7qAkRMfIl_iibPCqh2kiUzZArbwDZQ5NDIU6hkqVCVRYQ.jpg" alt="" loading="lazy"> <p>Need more gamification</p> <p><a href="https://t.me/chillhousetech/786?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So what do you actually know about the culture and motivation in your team? https://darkolorin.com/posts/785.html https://darkolorin.com/posts/785.html Thu, 13 Jun 2024 16:39:01 GMT <img src="https://cdn4.telesco.pe/file/C0mP1plEsuhMgbFkT4CDIWib3fza7QMKS19uXvVKGAl60_CmByechjEor-CTdFQhiWMVea-r6cfIMyhymCpj7y9AaEes6O4NWnWktxc5wZvU9B54GxWErOw-lpDOTtTVjLuWHzltEeCUQQVEJJr5NirS7R_HWkJpFYsKeimpErb6z32ap5S30fF2L_powVzolB7T0eYKSD_PeRPc4ohngCGwr4bWOQFt5MWVnL9wn8YrC0rt59hDEeT0TqPGaMgeEEiIGWcVJw4-IptGa9y8O2QR1RT2oXJEDVzucMkHhdkYWQ4U9g_-z6UcWKqfsxroAmYHNQxN2JAgIi_lPacBYg.jpg" alt="" loading="lazy"> <p>So what do you actually know about the culture and motivation in your team?</p> <p><a href="https://t.me/chillhousetech/785?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> From the beautiful https://darkolorin.com/posts/784.html https://darkolorin.com/posts/784.html Thu, 13 Jun 2024 16:08:00 GMT <p>From the beautiful <br> <a href="https://petapixel.com/2024/06/12/photographer-disqualified-from-ai-image-contest-after-winning-with-real-photo/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://petapixel.com/2024/06/12/photographer-disqualified-from-ai-image-contest-after-winning-with-real-photo/</a></p> <p><a href="https://t.me/chillhousetech/784?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A confident founder pitches his project https://darkolorin.com/posts/783.html https://darkolorin.com/posts/783.html Wed, 12 Jun 2024 23:36:58 GMT <img src="https://cdn4.telesco.pe/file/OSys_-sfyhRnjQgMTLxh-LY5iYuXpS0W7znERIizEXWyu0qN7BITolyVqJTR4cfc2ndPDhxNNnzLziNQKLGNdZSKVIadBGuE9CFQIMvDrUOn8yflCwquwsAl-DOAQaaK_aOVTT7C95bz2EstBIrZoD3T9AciFEeNWtpFeDAH6_6nqetFj3kBThmZXfoWAtEIJK3gVfqx4yx5JPeVcSIKPqap8M_QKOCWYe5AnzD_N3yO0Pgn39mUdT_VOhUHQTYFxHSZiIrP32Di0ZtiejfAvP1J1Qcm3QDZ9ksbO3TWBFt4etWPLPYuJU1t2blNQ7Njauk8FuBGLPyML5_UGJP7gw.jpg" alt="" loading="lazy"> <p>A confident founder pitches his project</p> <p><a href="https://t.me/chillhousetech/783?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Posting you a draft of my series. https://darkolorin.com/posts/782.html https://darkolorin.com/posts/782.html Tue, 11 Jun 2024 21:44:10 GMT <p>Posting you a draft of my series. Starting with retention. Throw stones. Gotta write a lot. Started little by little.</p> <p>What do you think, should I translate it into English?<br> <a href="https://telegra.ph/Retention-is-all-you-need-Not-really-06-11?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://telegra.ph/Retention-is-all-you-need-Not-really-06-11</a></p> <p><a href="https://t.me/chillhousetech/782?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Voodoo bought the social app BeReal Apparently it didn’t grow for a year. https://darkolorin.com/posts/781.html https://darkolorin.com/posts/781.html Tue, 11 Jun 2024 15:37:31 GMT <p>Voodoo bought (<a href="https://x.com/yazdialexandre/status/1800545937346957377?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/yazdialexandre/status/1800545937346957377?s=46</a>)<br> the social app BeReal <br> Apparently it didn’t grow for a year. I think it’s a good deal. They were paying for users. 500M deal.</p> <p><a href="https://t.me/chillhousetech/781?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> My two cents on the announcement: - the first hour of the presentation about… https://darkolorin.com/posts/780.html https://darkolorin.com/posts/780.html Mon, 10 Jun 2024 21:18:21 GMT <img src="https://cdn4.telesco.pe/file/eqWiZRqyLM0-otyi38sRLBA17ymioY525BQHHDL1ayk_XQrGDxKAl_Ya8nFOPkih9ZtbTmwvjVdRdXjEXWchMGmoxxG6gC_zPwxzwsCH5xXRMqYrsWrocsvqNC7yqvFM7bHrX_QDGE8SvJoa_h42PB1_801fSmVsEQQ5vSx3WXn8q_ClTT1sD5sXa0qvBXAgp2WlOIm2FfEg7OF7g6HEkFPOGwQwRjjoHiVHWDUnhOtd1uQXGkpyU8NVl8L-1GXf2mzPnUqJWcknsUke4CamyTZt2P11FxtfCe581bdjsrcygxHo9rrKDUZeE6PkZEBN7u_fxueeEfByAVsv3vVWqg.jpg" alt="" loading="lazy"> <p>My two cents on the announcement:<br> - the first hour of the presentation about announcing sidebars, recoloring icon colors, and all sorts of bullshit you can skip, they brought Android over<br> - liked the gestures in AirPods, nodded yes, shook my head no, been waiting for that forever<br> - finally my Mac will start accepting push notifications from the phone, and testing will be more convenient</p> <p>Then AI. Musk got toasted by a privacy-first company :) good. But overall I still don’t really get how all of this is going to work quality-wise. What matters here is how much the system will adapt and whether it will at all. Or did we just freeze the weights in memory and that’s it? What am I going to be sending to OpenAI?</p> <p>It’ll only work on the latest devices. And that’s fair. The company is still a hardware seller and needs to boost sales. That’s why they don’t give a damn about old devices.</p> <p><a href="https://t.me/chillhousetech/780?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Gotta celebrate the round at Burger King https://darkolorin.com/posts/779.html https://darkolorin.com/posts/779.html <p>Gotta celebrate the round at Burger King</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/779?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> KWAI KLING video generation model https://darkolorin.com/posts/778.html https://darkolorin.com/posts/778.html <p>KWAI KLING (<a href="https://kling.kuaishou.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://kling.kuaishou.com/</a>)<br> video generation model</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/778?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Some awesome news is gonna roll in soon, so I kinda dropped off:) there’ll be… https://darkolorin.com/posts/777.html https://darkolorin.com/posts/777.html Thu, 06 Jun 2024 15:20:18 GMT <img src="https://cdn4.telesco.pe/file/SvMypLwTMP84jBqRfG5kvrgc3IwEJPH329fomGmFfQIz66eGwtNPDG87UiCFX_gOmiFpCAc4s86t8hMPgw4PDI6r-0D4ZWjD41O5_uhd3gu8RRm04NnUd9lNiogrYy4Fv0_thk6ZfLKmIKsXUA6WdlFi3uhjqY4TVerTCLVeCYDax8HdPl8cq9mbFTIUBjqKcIbjydE5rEtWqxNIFODtR1D8ZGzUwu4bg2yVzbVlqLc_LOrUJlzKCTGj_qCSpnTqEY54LCe1g64IfCPJUbIFCnZtKmb3xCsi38CkENTb2zJPlBR_oNj7zjUCacP3XpR0Am-EAbQajHVVNUucMTvX_A.jpg" alt="" loading="lazy"> <p>Some awesome news is gonna roll in soon, so I kinda dropped off :) there’ll be a lot of content related to it <br> In the meantime, a lil meme</p> <p><a href="https://t.me/chillhousetech/777?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Is this a joke? https://darkolorin.com/posts/776.html https://darkolorin.com/posts/776.html <p>Is this a joke? Or are the guys actually serious?</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/776?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A new crisp leaderboard for evaluating models https://darkolorin.com/posts/775.html https://darkolorin.com/posts/775.html Wed, 29 May 2024 17:47:55 GMT <p>A new crisp leaderboard (<a href="https://scale.com/leaderboard?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://scale.com/leaderboard</a>)<br> for evaluating models</p> <p><a href="https://t.me/chillhousetech/775?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In one line — a new Mistral for code just rolled up, the competition’s growing. https://darkolorin.com/posts/774.html https://darkolorin.com/posts/774.html Wed, 29 May 2024 14:42:16 GMT <p>In one line —<br> a new Mistral for code just rolled up, (<a href="https://mistral.ai/news/codestral/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://mistral.ai/news/codestral/</a>)<br> the competition’s growing.</p> <p><a href="https://t.me/chillhousetech/774?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For anyone who was curious — well anyway they’ve started training a new OpenAI… https://darkolorin.com/posts/773.html https://darkolorin.com/posts/773.html Tue, 28 May 2024 17:18:45 GMT <p>For anyone who was curious — well anyway they’ve started training a new OpenAI model — probably GPT5Q or something around that. It became known from the news about forming a safety committee, (<a href="https://openai.com/index/openai-board-forms-safety-and-security-committee/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/openai-board-forms-safety-and-security-committee/</a>) — well yeah, they’re rightly afraid they’ll train some nonsense again on all sorts of sensitive topics.</p> <p>Also it finally reached me why Igor Babushkin at xAI is so badass and apparently it’s because he wrote this kind of paper, (<a href="https://arxiv.org/pdf/2201.02177" rel="noopener">https://arxiv.org/pdf/2201.02177</a>) which says there’s this effect called Grokking (so that’s where Elon’s model name comes from, I guess), supposedly it’s this advanced generalization even if you didn’t directly feed that kind of data into the net (Petya is Vanya’s friend. Vanya was born in 1999. Then we can say that Vanya, born in 1999, is Petya’s friend.). The topic is interesting because it kind of looks like you can get some logical chains by feeding, well, a *lot* of this kind of data. You feed both the data and some dependencies between them, and the model, with very long training on such data, starts to somehow connect these dependencies and infer new rules (implication and other stuff).</p> <p>And then there’s<br> a new paper (<a href="https://arxiv.org/abs/2405.15071?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://arxiv.org/abs/2405.15071</a>)<br> that develops this theme really well.<br> The authors study how it sometimes works out to get grokking (a-la magic), and sometimes it doesn’t. In short, the effect basically boils down to the fact that we can get OOD (out of distribution) links if we have lots of examples and links between them and then train on them for a very long time.<br> Good overview. (<a href="https://t.me/seeallochnaya/1473?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/seeallochnaya/1473</a>)</p> <p>Another cool topic is Super-alignment, but about that later.</p> <p><a href="https://t.me/chillhousetech/773?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Pretty solid job interview, if you ask me:) https://darkolorin.com/posts/772.html https://darkolorin.com/posts/772.html Tue, 28 May 2024 01:34:41 GMT <img src="https://cdn4.telesco.pe/file/reJjaBv4ao34k4b2MO9r3dV-lJGWQDXiricf9w5jJPXjESawpQgKdCu20Ba6puSNcKVx5OtH_1FpB3YzkKkEVmaiqHUx7l20IVEMD5N9l9NK4sPiTQZGRC4sWsPclJi5G1Z8Bg97oQeG_JLAKfB7vOqK9Z96ovcMjBKvrm8KSXaZV2EfoI9TBiMZfDWnhmj0triNpesKJXFLWvxLkR1f76zFsTmfHK8oxYm8iTEgMfE5Oddi2tWPmPbBsOzM4ueE5JSxVxlEVqKAlpy2GOB2MCNGuzVQkVMPW2pXq77FIkyXF3t520e38gRhHvcsDyNOx4XMP7dCKsIQWJe01eskkg.jpg" alt="" loading="lazy"> <p>Pretty solid job interview, if you ask me :)</p> <p><a href="https://t.me/chillhousetech/772?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sunday one. https://darkolorin.com/posts/771.html https://darkolorin.com/posts/771.html Mon, 27 May 2024 17:35:47 GMT <p>Sunday one. Just be honest. Have you ever actually gotten any benefit (and even remembered it) from some product/founder/narrowly-focused and specifically professional communities?</p> <ul> <li>Yes, yes, and yes again </li> <li>No </li> <li>It’s just hope, maybe there’ll be something there </li> <li>I look at it as a place to spam-chat</li> </ul> <p><a href="https://t.me/chillhousetech/771?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Not a bad metric for average check in Series B https://darkolorin.com/posts/770.html https://darkolorin.com/posts/770.html Mon, 27 May 2024 10:43:50 GMT <p>Not a bad metric for average check in Series B <br> <a href="https://x.ai/blog/series-b?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.ai/blog/series-b</a></p> <p><a href="https://t.me/chillhousetech/770?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Friday mood https://darkolorin.com/posts/769.html https://darkolorin.com/posts/769.html Fri, 24 May 2024 15:11:36 GMT <img src="https://cdn4.telesco.pe/file/K5klAWVbg85XHIrCPmUhWk1utli45oDnr4B2yUbXTBMtAClpqSFwO7FxPXWV9AjapVt0k0ROfrP8JRMk5gtWxmdpRjjfZVbA25VenfqfIYLtQb7PRuQBMUQuf4_JYydO0zMsxtdpWTe2BMvyAMvt_O7wRoH0JKEckIMgvGTRuFTG1Gz139jhao1VIvDr-MdKos_PXYcy2vZq0fyJjySZ7dyLi1dCIzOWFIppEyEwbDHvoCEOHAwVLQi-Yc-mMbaP1q6RJiau-rH3A8YsXn85iDlJtVztmeVSCFJPYkmI29hFqIqhlE_3_ivKk4H-cy2wq4HStI1ICMnep5Itj-YQfA.jpg" alt="" loading="lazy"> <p>Friday mood</p> <p><a href="https://t.me/chillhousetech/769?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For anyone interested, guys put together the rounds in the AI landscape. https://darkolorin.com/posts/768.html https://darkolorin.com/posts/768.html Fri, 24 May 2024 13:58:00 GMT <p>For anyone interested, guys <br> put together (<a href="https://airtable.com/app5Wwd1uz4jLGAK6/shrzdJxtNvtbDlvKM/tblTdw08HhECrQisJ?viewControls=on&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://airtable.com/app5Wwd1uz4jLGAK6/shrzdJxtNvtbDlvKM/tblTdw08HhECrQisJ?viewControls=on</a>) <br> the rounds in the AI landscape. Just for studying.</p> <p><a href="https://t.me/chillhousetech/768?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Google rolled out its generative answers in search (US-only, I think). https://darkolorin.com/posts/764.html https://darkolorin.com/posts/764.html Fri, 24 May 2024 13:14:49 GMT <img src="https://cdn4.telesco.pe/file/WyOKntxfCxilxCe28w2e1bS4fgs77_o5YXrXB7QdAcubu0k1HW55RqABZCt9nN35E3NflUBHpPHzONxnz-ygP5jcFVre2gkNa1m-H_343murbWR9KZnFIo799RBQlca1R_Ie_YJU6sD0sNweYaxiq_S5hy57brg1NHFQJzCGKXAN-lvVX5lBHMGhJrQhnzck1zbQOW-GhVx0rCJf9g7jcIOYRCSrnvMLQ75N-7RfTlog7TgJuhhbXXaxvwNbw3RN_joj48lPtHzBcfS35ZuBoViQXhD3PnuD9DdNjatU17O1snm50HBfDg9m_9MDYPyeNleT0svQ8vjPC7w41GEmgA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/CR1l9BAU__8ieXgBrEuD3VOQCaHzWf2vrN4-CeOHqKk6GBJvepeN4RCzjY-rszu_q4jdGIT1aVUHITRJ86A8td2NC1kWsrIC7w6_0dh1cUNvYY0HP-9u0LcXG8kqeXCcBt3BbDJ3wXJSiCBQFq6zKCGywofx8OrnX8yBJXqCWZAdIRuwugYc7tVndfkRB8fPmBbFQg1C0x3Ib4Ee_gf93lg3csnp3JIQw44LJTtsA9ZhkgPHscSO-6YXyyQwgHTW92PPnAXA6XMcfoLAVcvT0VqwqMHeNuNcSHanMuGC3Hg7iY82tjeXstRDvG-2-TpUjGdX2CUu5O-WlOqSQuMOtg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/EqsDIv2OppsaS0CPsW81oluZJJPZSLuYQmzQcH7gtIqabUbDahyRgIOE868z5VeOeIkx5PP1wyQXmez-kwqqEMVJUUfMUpDCdhf9cHgeJlmrWVdm8aaYOq7FgeBYIvV3iCCKTT7ulM3-tSOuuyYFF9i1QaeusnbhSLibAcDTXx44ty9fWgeW5H9bGDZY2892QLT1mDBc1FUfSL5CGUuIs-usb4BDngLrlx7JzRQjOcI4chXYdQq9lb30rSkPVV6IJ-sMZa4FHdNBDE0byGrJbquUJufExNDL45h4cG7uHpOCLUwqR786D3QyyYtUTUkTANEpsWQ0nK1BmrYDWXWO6Q.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/bvBjFBxFuL2vL2sdnkumeGn-vEqjLJkPeG5yZd84rM5MqR4Z5aztRoC7DIBVeIp6Wtyszz9G109QTTQneAx4tQKvtP0kxRE2FWVrvH8qjUvlW88aln2VbtD-Mc_XmHPu-L06pbuhXQ6Zl_69cHAzJSJuqKzaAj9wHQwGaOblsvrOlBE-B5LAGPHCe12ziZ45JqjdIY7woE1C92Tp3g3FkCDxbsvyhzCYF5PErcrpTfTQA_FL6Occc6attWSQEc50R3So-_EBJn3t2aiK2H37Ip3p_fxALod9d9zi8I9sINXeU4djcBAmijKnqpGaNZ9BLyI_mzXla4YIfT1C9mdZOw.jpg" alt="" loading="lazy"> <p>Google rolled out its generative answers in search (US-only, I think). Of course it slapped in some ads. But the answers are horror, reminds me of the release of their image generation model. Looks like they’ll roll it back too.</p> <p><a href="https://t.me/chillhousetech/764?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> AnimateDiff + inpaint https://darkolorin.com/posts/763.html https://darkolorin.com/posts/763.html <p>AnimateDiff + inpaint</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/763?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Went through some of the questions I’d been collecting. https://darkolorin.com/posts/762.html https://darkolorin.com/posts/762.html Wed, 22 May 2024 18:32:03 GMT <p>Went through some of the questions I’d been collecting. There are a couple where I really need a lot, a lot more time. <br> <a href="https://telegra.ph/AMA-05-22-20?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://telegra.ph/AMA-05-22-20</a></p> <p><a href="https://t.me/chillhousetech/762?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In general, all of Anthropic’s work is super interesting to read. https://darkolorin.com/posts/761.html https://darkolorin.com/posts/761.html Tue, 21 May 2024 15:32:41 GMT <p>In general, all of Anthropic’s work is super interesting to read. And today there’s something new. <br> On interpretability <br> a new series just came out. (<a href="https://www.anthropic.com/news/mapping-mind-language-model?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/news/mapping-mind-language-model</a>) <br> The previous series (<a href="https://transformer-circuits.pub/2021/framework/index.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site#mathematical-framework" rel="noopener">https://transformer-circuits.pub/2021/framework/index.html#mathematical-framework</a>) <br> I consider it one of the best works out there on understanding how a transformer works.</p> <p><a href="https://t.me/chillhousetech/761?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Not a bad framework on product-market fit. https://darkolorin.com/posts/758.html https://darkolorin.com/posts/758.html <img src="https://cdn4.telesco.pe/file/hPpx1NCIC0kt8vVWMUVIG8wO91ewndZHQJTHysXzWAL_76o-HmD4eyXEH0OYoN4z9XWMqSRHATL8yJ6zh25cxUiZxcS6g9sTeNHfTkwK5hZlwazV8fOgaHiOBTlSOs921AKNcUSN-WncN7gNQZYYBYQLMPJW9k6zFkXJar7Tnq9ydZt-6B13JoW0ZdCodBDZrsMl3F8YfsXuo5xPetuKVS8wYyrYPlytf_CCVHvrN4QnVTPnHcBBmPncGeOM6QPdSrvzIgu49NVVomTsn1bx9QqjclfF2psR9zOXVeXxWnbREa6mNLdqsH6v4NOPTUiRQWvAgm-ITvqxoinUEg26Hg.jpg" alt="" loading="lazy"> <p><a href="https://www.sequoiacap.com/article/pmf-framework/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.sequoiacap.com/article/pmf-framework/</a><br> Not a bad framework on product-market fit. Also, I was in Scotland for a few days. A couple photos for you, and I highly recommend you go.<br> Let’s get posting :)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/758?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If you need, right at the start, at the very beginning, help getting the… https://darkolorin.com/posts/757.html https://darkolorin.com/posts/757.html Wed, 15 May 2024 22:26:48 GMT <p>If you need, right at the start, at the very beginning, help getting the product to market from experts, mentors, trackers, consultants — don’t do that startup.</p> <p><a href="https://t.me/chillhousetech/757?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Still can’t stop being surprised. https://darkolorin.com/posts/756.html https://darkolorin.com/posts/756.html Wed, 15 May 2024 16:27:23 GMT <img src="https://cdn4.telesco.pe/file/vAJAYW1a2xYh6TEfDK12y4cm6PbNvIayEvo1dlxDE6iWh8INWmp7K1536WqP4oSGQspC9V6JOHMBHycBb5k8vf-0lquoVpvAFll8jQqRvxVcDDURtT7lAXcz9YvRuatvdUSwDlT_1pjB5hUpbjrhJSAolJKKW6TtDk3Q3Zm6V4MDBuV8sYSjTwH1FBRE05Nipz9RFyB1KH6h7c1NRl3chtTPPvpB5Cyy0BitoBXJ4m0j6Ta23YP9p_VR_oi2zI0nA-nWvJJqgEPw-RJMq_uS0N2PprRO_MWTW4fdBIhUk6fjx8wSAHlE3Xxdor6C8XQ8ZkGDXTgsA-IjN-EiRqodHA.jpg" alt="" loading="lazy"> <p>Still can’t stop being surprised. <br> Anthropic (<a href="https://www.anthropic.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/</a>) is hiring Instagram’s cofounder. You know, like a big product guy with “strategy.” But who, I’m guessing, hasn’t really done anything in product for the last 7–8 years. <br> Artifact (<a href="https://artifact.news/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://artifact.news/</a>), which was supposed to be for news, basically never went beyond the announcement.</p> <p>What’s funny is that formally he was CTO. (<a href="https://www.linkedin.com/in/mikekrieger?utm_source=share&amp;utm_campaign=share_via&amp;utm_content=profile&amp;utm_medium=ios_app" rel="noopener">https://www.linkedin.com/in/mikekrieger?utm_source=share&amp;utm_campaign=share_via&amp;utm_content=profile&amp;utm_medium=ios_app</a>) <br> And was doing team scaling. <br> Well then, waiting for new filters from Anthropic. <br> And who’s the ideal CPO for you? Drop a LinkedIn link.</p> <p><a href="https://t.me/chillhousetech/756?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Expected. https://darkolorin.com/posts/755.html https://darkolorin.com/posts/755.html Tue, 14 May 2024 23:09:08 GMT <p>Expected.</p> <p>Ilya is leaving OpenAI (<a href="https://twitter.com/ilyasut/status/1790517455628198322?s=61&amp;t=rh4M-tIq0pwheQB0rsBmgg&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/ilyasut/status/1790517455628198322?s=61&amp;t=rh4M-tIq0pwheQB0rsBmgg</a>)</p> <p><a href="https://t.me/chillhousetech/755?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Google, predictably, rolled out a bunch of new stuff, but I think commercially… https://darkolorin.com/posts/754.html https://darkolorin.com/posts/754.html Tue, 14 May 2024 19:34:58 GMT <img src="https://cdn4.telesco.pe/file/tWCXLDhyJcX5fWj3iLCIguT0TQ8HmgJd7w5HuKRqSPg-EUpy-WghzKw1ZScE15yYOZh5n5XC_JIoml-00fXMl1IQXwdGyLGKTK8KWFVJx7wlGbHVTSIIcLV-33ukce0ffTbOPLAck1aQYe9j98TPkHw4wQQevmPMRw-HVKJgJj0FRoBe3EJCyh5Q1MzEZ4ykBm8o2KedA3fM0KlZumGkOs9uG6iatS8JeSXwfA-I9d4jcVwmtx6ShE06erXzZWPrbrlU-YQlJc3KyvgOBN-GrjgSfJFPGa0URa-5VtiYixzz7fSUbCqQ1uZR1_JmmbYIaxL_cpf-qcYYCSzzwUIMvw.jpg" alt="" loading="lazy"> <p>Google, predictably, rolled out a bunch of new stuff, but I think commercially they’re ahead in terms of understanding how not to lose money in all this.</p> <p>AI Overviews. When they hand off part of the queries straight to the little model, right in search. Curious where ads end up in that case, and how much the guys lose. They’re shipping it in the US as early as this week — I’ll be watching the report. But I’m sure they’ve crunched everything and most likely they’ll be leading it into YouTube where ads are growing nicely.</p> <p>Other than that: a huge window size, a “competitor” to video generation — kinda meh, a bunch of doodads with visual search.</p> <p><a href="https://t.me/chillhousetech/754?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Thought about it a bit. https://darkolorin.com/posts/753.html https://darkolorin.com/posts/753.html Mon, 13 May 2024 19:25:16 GMT <p>Thought about it a bit. Still think this whole tower of tech is needed as the only way to shake up search. Because if you have the better model in terms of user experience of conversation and language understanding, you can just shove a web index underneath + your model and boom, you get a new Google. <br> The current phase of OpenAI and competitors is beautiful, directly it’s enterprise. But everyone understands where the money is. <br> Probably one of the few moments in 20 years when you can wobble the search market. The trick is even that, in theory, you can kill Yandex or any other local search engine. Only regulation will help, but most likely it’ll just push the whole industry’s progress back by 1–2 years (time to catch up).</p> <p>My timeline so I can check later (good exercise for any manager: write down what you think and then check):<br> - by the end of the year there’ll be releases of this whole thing in clients on all platforms (real-time interaction with ChatGPT)<br> - next year, natively in every phone<br> - by the end of next year, Apple headphones will have their own Jarvis<br> - in two years I’m betting an OpenAI search engine will have some noticeable share (it’ll be called something like some assistant)</p> <p>Most important problems to solve:<br> - building your own index<br> - very fast iterations on improving the model in local markets<br> - moving inference to CPU<br> - caching everywhere</p> <p>I even sort of believe today’s release is more important than an LLM like ChatGPT 3.5. Because in theory you don’t need to have this huge data archive that gets trained for a hundred years in GPU-hours. All hail deep learning. Oh man, how much more we’ll do through data and curve fitting.</p> <p><a href="https://t.me/chillhousetech/753?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So here’s the vid for you, I think it’s a great release and it’s clear why… https://darkolorin.com/posts/752.html https://darkolorin.com/posts/752.html <p>So here’s the vid for you, I think it’s a great release and it’s clear why Apple wants this kind of thing. Waiting for Google tomorrow and their little clapback.</p> <p>I still see a big enterprise business in this and don’t really get how they’re planning to do b2c, because there still aren’t enough GPUs to go around. Plus, waiting to try it ourselves, because practice shows a demo is one thing, and real-world use is another.</p> <p>And also <br> some thoughts from Sam (<a href="https://blog.samaltman.com/gpt-4o?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.samaltman.com/gpt-4o</a>) <br> Model page (<a href="https://openai.com/index/hello-gpt-4o/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/index/hello-gpt-4o/</a>)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/752?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Remember I wrote about the model here, well, that was the new model, and it’s… https://darkolorin.com/posts/751.html https://darkolorin.com/posts/751.html Mon, 13 May 2024 17:18:03 GMT <img src="https://cdn4.telesco.pe/file/NqU7PFnz5IrD-07CdL9-BVo07s2FVVAmDNhHbsCALLTTDwsfP7l_W62J9jOhTWVFkhzj5fYgm5d4uwRb4WRBfqBTfGaMPX2CcKlshEvhxvPo5N8Hi0ILQh0LG7LnZBkUI5PCrvYsHw7g_AUwP8w2JiDRCYEELx2NRTSFgqG9u9D3YXg5ohwJtUma5srrJeXW-O0DAObO1JUBn9PLHYr7c8UYfQGCBY5jN0pEwRdgtSkuq0C0tGqgt-Q5lsUzvEo-RjXz1PTCYtCGiaELJN3cuSzXOMDbWhvho803T1TWfbNxUA_wIjdxwDuEDEYYAUoX2mgLLVPL7q0YKhSoPvTIIQ.jpg" alt="" loading="lazy"> <p>Remember I wrote about the model here (<a href="https://t.me/chillhousetech/732?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/732</a>), well, that was the new model, and it’s very good judging by the benches on ChatArena. To put it mildly, it beats everyone and it’s really good at coding (which I noticed right away).</p> <p><a href="https://t.me/chillhousetech/751?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Just in case, the OpenAI announcement stream is here. https://darkolorin.com/posts/750.html https://darkolorin.com/posts/750.html Mon, 13 May 2024 16:51:24 GMT <p>Just in case, the OpenAI announcement stream is here. Starts in 10 minutes.<br> <a href="https://www.youtube.com/watch?v=DQacCB9tDaw&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=DQacCB9tDaw</a></p> <p><a href="https://t.me/chillhousetech/750?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I think it’d be a decent conversion, by the way https://darkolorin.com/posts/749.html https://darkolorin.com/posts/749.html Mon, 13 May 2024 12:55:45 GMT <img src="https://cdn4.telesco.pe/file/ui15mCqAqvGaNY2QGe7tBGB5v6qsWAH0A_Zwz_U_5ggGRMxOXgfzkuIjfWnjoNwJvTpWTPn2HQCLZnd3KUk23N5m_L-0U6QQnbui1WwmCPvNesBOmzmbF3HWRkXk1yQPQAw8BSWYHIGvhJXSxTayL2plTwcxOY1WO9l9qHKL3Ay_dKqNRnAZN-FDxPxcr8RI4XZezXsy6fT9rL8BRodKkSZKLDKLzEAMqiJF51eH7ohmeV0R5_RdiAt7xdF2INYr9Kr9WESF_zsSMjf5uaxxLsX8wUR49JDnuhyR-6OqBxiSAXSmk6uFhYi6Up9fzRYhAnoiY4Yag78Ej_y2w7uVaA.jpg" alt="" loading="lazy"> <p>I think it’d be a decent conversion, by the way</p> <p><a href="https://t.me/chillhousetech/749?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Overall it doesn’t surprise me and doesn’t leave me with any questions why… https://darkolorin.com/posts/748.html https://darkolorin.com/posts/748.html Mon, 13 May 2024 10:59:41 GMT <img src="https://cdn4.telesco.pe/file/DW8BZgaYj8zA-BL1m6Uk5Tkd49-mFq97rvP94D3_PkjWtUrmel5CBo47za0Jat5JKstt-eqg-UuJTDQ0KBs4rgCAGJCmrDKjRm2bPR4SCNkSH8gkGvylA0xfKUGKFbC9Q-p_IKj326SYDhDe3ap_Ch9czqI0uQ0kSum_y3m7jKpNhIMyIp0Edo8siyuKvV1PgVvijWmzcN8GAKiLDZx9iesa-tpAGoK6WElfe361Qb1h9GjJD3NZHn3KK8uNr0RO6fuM6VbrCuRWfHRcEZ4NAgjlhe0Kpj7_tTzWsrVFs-OZm1rIUxRLFZufr7XmPb4lRJI_eT5brITXtO9yILkFxw.jpg" alt="" loading="lazy"> <p>Overall it doesn’t surprise me and doesn’t leave me with any questions why Meta’s model performs so well on dialogues and chats, and worse on other tasks like exams and whatnot. <br> And all the authors and researchers keep reasoning like, well, they fed it a dataset neatly and blah blah. Well no shit: a huge social network has dialogue data with data about when it’s good and successful and labels and any damn context you could want. And then good luck proving to the model that they trained on some private labeled data. I’m waiting for Google to also start shoving everything it scraped into a dataset. It’s all a matter of time and managers’ KPIs. Here too they should’ve pushed the spot higher on the Chat Arena.</p> <p><a href="https://t.me/chillhousetech/748?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Rumors again about upcoming releases. https://darkolorin.com/posts/747.html https://darkolorin.com/posts/747.html Sat, 11 May 2024 14:16:26 GMT <p>Rumors again about upcoming releases. It’s gonna be a hot week since it’s both Microsoft Build and the OpenAI and Google confs all at once. We’ll see. <br> But the coolest innovation in the business model is paying for tokens upfront :) then it’ll be cheaper. That’s solid. <br> Otherwise everyone’s waiting for search, a Siri competitor (while they themselves already agreed to sell the API to Apple), and a bunch of other stuff. I’m waiting for a 10M-token context window so I can try a ton of new use cases. Your hypotheses? :)</p> <p><a href="https://t.me/chillhousetech/747?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> 30 papers, that, rumor has it, Ilya Sutskever (one of the top scientists at… https://darkolorin.com/posts/746.html https://darkolorin.com/posts/746.html Thu, 09 May 2024 23:22:13 GMT <p>30 papers, (<a href="https://arc.net/folder/D0472A20-9C20-4D3F-B145-D2865C0A9FEE?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://arc.net/folder/D0472A20-9C20-4D3F-B145-D2865C0A9FEE</a>)<br> that, rumor has it, Ilya Sutskever (one of the top scientists at OpenAI) recommended to John Carmack (legendary programmer and founder of idSoftware, the Quake guy) to figure out what’s what in AI. Says if you understand these works — everything will fall into place. <br> P.S. I read them and understood about half. Like, enough to explain it to someone else.</p> <p><a href="https://t.me/chillhousetech/746?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Music by elevenlabs, well, an expected move. https://darkolorin.com/posts/745.html https://darkolorin.com/posts/745.html <p>Music by elevenlabs, (<a href="https://x.com/elevenlabsio/status/1788628171044053386?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/elevenlabsio/status/1788628171044053386?s=46</a>)<br> well, an expected move. You think they’ll actually be able to make even one popular album fully via generation? And when?</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/745?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> looks like they’re making an editor with all sorts of AI goodies. https://darkolorin.com/posts/744.html https://darkolorin.com/posts/744.html <p><a href="http://KREA.AI/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://KREA.AI/</a> <br> looks like they’re making an editor with all sorts of AI goodies. Like, a Canva-ish approach, I guess, just modernized and also so you can mess around more. Overall, Canva started out with story templates too.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/744?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Hmm https://darkolorin.com/posts/743.html https://darkolorin.com/posts/743.html Wed, 08 May 2024 21:50:55 GMT <img src="https://cdn4.telesco.pe/file/OvILTctvN7v7supPZTOqS9omFX2y12hUOVYnBc3IHNkYuI4hIrIO_rZ3jEMenKkkqC-cTnmgOeVLNGx9uAZ9TKXLryd1S8kc9HDDM4tKV33QeeUUjAb2p9gVecuZFn8Au-xL27KKUDr8Zh48qrbGw8qbgLFcFA7KbDSUc-4KzCZ7cfu101xsX72TSeZxYifvSWkacFZ0xPnz2_q-RRHNMTkpCg_lB6rn4c9knDMGjBGnCpMeGDbGuG0iZAxnoWiGqFJ5LSCNaWfc-ydV6bd92IEhITJOq9q8yeogqLQ54rvgcrIu5yBZaYyHgOcxiRyWQWEhV4Xi0D0yY7aF8KCyTg.jpg" alt="" loading="lazy"> <p>Hmm</p> <p><a href="https://t.me/chillhousetech/743?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So what’s the deal with molecules and proteins and all that stuff. https://darkolorin.com/posts/742.html https://darkolorin.com/posts/742.html Wed, 08 May 2024 19:25:13 GMT <p>So what’s the deal with molecules and proteins and all that stuff. They say a nice little model will help us beat viruses faster, make cool convenient fertilizers, and so on. <br> <a href="https://blog.google/technology/ai/google-deepmind-isomorphic-alphafold-3-ai-model/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.google/technology/ai/google-deepmind-isomorphic-alphafold-3-ai-model/</a></p> <p><a href="https://t.me/chillhousetech/742?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> TLDR status of AI/ML: 1. https://darkolorin.com/posts/741.html https://darkolorin.com/posts/741.html Tue, 07 May 2024 23:55:42 GMT <p>TLDR status of AI/ML:<br> 1. Scaling and advancing deep learning will keep bringing more and more new capabilities and features and will make current tasks better and better.<br> 2. And at the same time, it won’t get us to AGI.<br> 3. We’re still far from AGI, and LLMs haven’t really brought us much closer to it.<br> 4. We’re still nowhere near full adoption and integration of deep learning methods. There’s still a huge number of tasks you can solve with it, and there’s still a lot left to build.</p> <p><a href="https://t.me/chillhousetech/741?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Really want to, and my hands are itching, to write a detailed teardown of this… https://darkolorin.com/posts/740.html https://darkolorin.com/posts/740.html Tue, 07 May 2024 20:41:52 GMT <p>Really want to, and my hands are itching, to write a detailed teardown of this article about “increasing retention” methods (<a href="https://a16z.com/7-ways-ai-native-companies-can-improve-retention/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://a16z.com/7-ways-ai-native-companies-can-improve-retention/</a>) by a16z partner Bryan Kim, with whom I have a very contentious history of interaction. Because his usual response to your pitch is like, so where’s your d30 ~30% and a million DAU, you say you don’t have it, well then bye. <br> But I think I’ll package my version of his advice as a separate big post. Especially since I’ve piled up plenty of material.</p> <p><a href="https://t.me/chillhousetech/740?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> MAI-1 - and in general LLM names are looking more and more like names of… https://darkolorin.com/posts/739.html https://darkolorin.com/posts/739.html Tue, 07 May 2024 12:24:12 GMT <p>MAI-1 (<a href="https://arstechnica.com/information-technology/2024/05/microsoft-developing-mai-1-language-model-that-may-compete-with-openai-report/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://arstechnica.com/information-technology/2024/05/microsoft-developing-mai-1-language-model-that-may-compete-with-openai-report/</a>)<br> - and in general LLM names are looking more and more like names of spacecraft. Voyager, Soyuz, TKS and so on.<br> In short, Microsoft hired a former DeepMind leader and now, obviously, is training its own foundation model with 500B parameters, most likely a release at their Build conf (<a href="https://www.theverge.com/2024/2/14/24073220/microsoft-build-event-schedule-may-21?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/2024/2/14/24073220/microsoft-build-event-schedule-may-21</a>)<br> on May 21. Well, we’ll see how they’re going to compete with their own startup. Reminds me of the DeepMind &amp; Google setup.<br> And in general<br> here’s a useful resource (<a href="https://huggingface.co/spaces/ArtificialAnalysis/LLM-Performance-Leaderboard?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://huggingface.co/spaces/ArtificialAnalysis/LLM-Performance-Leaderboard</a>)<br> for comparing price/speed of available model providers. Hope Microsoft won’t make it as expensive as OpenAI.</p> <p><a href="https://t.me/chillhousetech/739?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> From Andrew Chen’s new article I’d probably make a summary like this, which in… https://darkolorin.com/posts/738.html https://darkolorin.com/posts/738.html Mon, 06 May 2024 15:49:46 GMT <img src="https://cdn4.telesco.pe/file/RrawO3nYkTirNrtYejM-9t9FJjSscjm72A62GUIeBctSEGRSuGYU6Y71uiY7v2n6QNrZyD_t46s0pxIKZHv-5h7ExpMnmXWu1L0kMgE1X0xggC42SuM2YyBqF1A1YeYBcyAjZuKfzTXG9_INCPR4lIi3RuUyGHHv8vfrBzQqRZIPn7XXcBY0tXl2dZi3SHP3koSW3cMnRVShsYvilOijsCroY4nIwYdjILFX2phOoWdPeQPsjxDW_rCs68u8I9zUEzKhjsNK9k9A9Oi8y1ldhF9L1zZlpSbJe5QT8AEdQBOwjwp-ylkWBEU4d-JMIO0h9WkP7HFczMPiEQ9H3s8UyQ.jpg" alt="" loading="lazy"> <p>From Andrew Chen’s new article (<a href="https://open.substack.com/pub/andrewchen/p/how-novelty-effects-and-dopamine?r=3p9xs&amp;utm_campaign=post&amp;utm_medium=email" rel="noopener">https://open.substack.com/pub/andrewchen/p/how-novelty-effects-and-dopamine?r=3p9xs&amp;utm_campaign=post&amp;utm_medium=email</a>) I’d probably make a summary like this, which in many ways shows the trend of the so-called dopamine multiplier.</p> <p><a href="https://t.me/chillhousetech/738?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A minute of math on the channel, while I’m over there looking at what… https://darkolorin.com/posts/737.html https://darkolorin.com/posts/737.html Sat, 04 May 2024 22:43:37 GMT <p>A minute of math on the channel, while I’m over there looking at what questions who wrote, I’ll start answering sometime during the week. <br> Recently, like I said, I talked about ICL, I’m prepping a proper piece for here to post. And there, before me and my little talk, there was this nod to the idea that like, in LLMs everything is sooo non-deterministic. To which I objected, like no, not really, it’s the interpretation that sucks, and in general if you do token sampling by the highest probability — the degree of, so to speak, “random” drops really low. And I decided, overall, I’m thinking, let me see just how much math can squeeze certain types of tasks where there’s a lot of randomness down to clear determinism. And it turned out, actually, the question is solved. <br> For a 1994 paper, he and the computer scientist Noam Nisan illuminated that connection. They proved that if any natural hard problems exist, as most computer scientists suspect, then every efficient randomized algorithm can be replaced by an efficient deterministic one. “You can always eliminate randomness,” Wigderson said. <br> Source (<a href="https://www.quantamagazine.org/avi-wigderson-complexity-theory-pioneer-wins-turing-award-20240410/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.quantamagazine.org/avi-wigderson-complexity-theory-pioneer-wins-turing-award-20240410/</a>)</p> <p><a href="https://t.me/chillhousetech/737?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Quarterly AMA — ask me anything https://darkolorin.com/posts/736.html https://darkolorin.com/posts/736.html Fri, 03 May 2024 18:26:19 GMT <p>Quarterly AMA — ask me anything <br> <a href="https://forms.gle/uQiHqpd86g2NLgQR6?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://forms.gle/uQiHqpd86g2NLgQR6</a></p> <p><a href="https://t.me/chillhousetech/736?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> You remember how recently OpenAI opened an office in Japan. https://darkolorin.com/posts/735.html https://darkolorin.com/posts/735.html Thu, 02 May 2024 11:47:44 GMT <img src="https://cdn4.telesco.pe/file/fzE6wie15lG0MFCL4hndYG1AOmZAaS_7nlG19ZSBTdfkaafgv-6_gqwGr4V7dTR6huP5ztw1PvV0GaoOvitM_AtEd-GDXNp2IzfunPLSDftMc3q8QXMTBz-EhyKKiz-aG_-qHhaHiW9sBtpmLRWyTAA2Ps6ERiJLa-vqshn38NfTl-4WWQ1G_sgjIOxnyT4EFNGhuCFfWv3e_aisVs0x31QKuaFc96ANhe1dWIEgmN96rF9tzGNSQ8hbeBXq5N2gT4Z5eh_sB_heWqBouvDeNjbWm5sf-529oCWG7qn1n2mSv269rfAFvtoav-hTLgs0-tB_ZHiicSMdZoFOcl8fTg.jpg" alt="" loading="lazy"> <p>You remember how recently OpenAI opened an office in Japan. And now, out of nowhere, xAI of Elon Musk opens an office in Japan. Coincidence? Don’t think so.</p> <p><a href="https://t.me/chillhousetech/735?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A decent career — a decent rest https://darkolorin.com/posts/734.html https://darkolorin.com/posts/734.html Wed, 01 May 2024 22:29:10 GMT <img src="https://cdn4.telesco.pe/file/ipA2Ks15FbtAxHQ7hNwpibab5NuW88O4j_8F8RQwYcTVPH92Ea9MZnsNJtx_tAWgnC8W-cnWcr5bxF_d-OhSYSYkSqkHky5eNsfhxKS9eMrxIRnfUdxFRZz40lv0jFfi7hMOu7NHsSIlXQzrkWwl5In0uYpOWyxcGHCkAE99lDHatW3UH1f7O1arie4t-5Ib7GuNFqCq_PWYNOADo3ey4OtS8TFDtWyg4XhLX1N04mFgjPDFbY6plHT5KH_tchGmTf_IhLQobvrNHOM6NOHZszizCu1FG0jRnVkFGaDSzO7SangRXnGPGySEouee2UXiACigC9eetdYFHSCnF056Mw.jpg" alt="" loading="lazy"> <p>A decent career — a decent rest (<a href="https://www.linkedin.com/in/dryuan/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.linkedin.com/in/dryuan/</a>)</p> <p><a href="https://t.me/chillhousetech/734?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Anyone making games or services around them will probably find this list of… https://darkolorin.com/posts/733.html https://darkolorin.com/posts/733.html Wed, 01 May 2024 16:24:56 GMT <p>Anyone making games or services around them will probably find this list of gaming funds and the corresponding contacts useful. <br> <a href="https://nextgengamingclub.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site#investors" rel="noopener">https://nextgengamingclub.com/#investors</a></p> <p><a href="https://t.me/chillhousetech/733?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Something new came out and it’s very powerful. https://darkolorin.com/posts/732.html https://darkolorin.com/posts/732.html Mon, 29 Apr 2024 22:52:20 GMT <p>Something new came out (<a href="https://www.reddit.com/r/LocalLLaMA/s/c51pkyvN1T?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reddit.com/r/LocalLLaMA/s/c51pkyvN1T</a>)<br> and it’s very powerful.<br> This little model, in my tests, handles stuff that GPT-4 doesn’t really. Like, it writes ready-to-run code for mini-games that launch right away. Most likely we’re waiting on an official release in the next few days. You can test it on<br> chatbot arena (<a href="https://chat.lmsys.org/?leaderboard&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://chat.lmsys.org/?leaderboard</a>)<br> pick gpt2-chatbot</p> <p><a href="https://t.me/chillhousetech/732?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Twitter’s on fire again. https://darkolorin.com/posts/731.html https://darkolorin.com/posts/731.html Mon, 29 Apr 2024 22:32:23 GMT <p>Twitter’s on fire again. Or all the techno-optimists ran face-first into the California State Senate, which today pushed out this little topic (<a href="https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202320240SB1047&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202320240SB1047</a>) into the public (and soon probably to a vote).</p> <p>The funniest paragraph is obviously this one:<br> &quot;If not properly subject to human controls, future development in artificial intelligence may also have the potential to be used to create novel threats to public safety and security, including by enabling the creation and the proliferation of weapons of mass destruction, such as biological, chemical, and nuclear weapons&quot;</p> <p>This is basically a bundle of all the memes about Terminator and Skynet.</p> <p>In short: open source and startups are gonna suffer and get regulated hard as hell.</p> <p>Like, they’ve got examples in there about automating calls via AI and phone services, cameras at traffic lights, and other models you can use for that kind of thing (fraud, identity theft, etc.) — you’ll have to, so to speak, “check” them and get the company approved.</p> <p><a href="https://t.me/chillhousetech/731?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Memory in ChatGPT, I don’t think it’s gonna boost their metrics much, I can’t… https://darkolorin.com/posts/730.html https://darkolorin.com/posts/730.html <p>Memory in ChatGPT, I don’t think it’s gonna boost their metrics much, I can’t even picture there being users who really need it that bad. Available starting with v4.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/730?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ll repeat once again what seems right to me. https://darkolorin.com/posts/729.html https://darkolorin.com/posts/729.html Mon, 29 Apr 2024 01:27:00 GMT <p>I’ll repeat once again what seems right to me. What matters isn’t the model architecture (if we’re talking quality, not speed), but the dataset. Nothing else matters in the long run. Data will win, not fancy ways of shuffling layers and hyperparameters around, even if at times it feels like that’s super important.</p> <p>And if data wins, then we’re back in the wonderful world of “whoever has the users, wins.” I’m sure the folks at OpenAI understand this extremely well and figured out a long time ago that Microsoft is good.</p> <p>An ancient paper (<a href="https://arxiv.org/pdf/2212.10544" rel="noopener">https://arxiv.org/pdf/2212.10544</a>)<br> for you to think about and<br> an old post (<a href="https://nonint.com/2023/06/10/the-it-in-ai-models-is-the-dataset/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://nonint.com/2023/06/10/the-it-in-ai-models-is-the-dataset/</a>)<br> for you to read.</p> <p>P.S. By the way, soon I’ll be talking about in-context learning in a short presentation. This is the stuff people call magic, reasoning, and so on, even though in reality it looks like Bayesian inference. But the phenomenon is interesting. Are you interested? (Heart reaction then)</p> <p><a href="https://t.me/chillhousetech/729?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Haven’t had Saturday posts in a while. https://darkolorin.com/posts/728.html https://darkolorin.com/posts/728.html Sat, 27 Apr 2024 22:00:01 GMT <p>Haven’t had Saturday posts in a while. It’s been pouring here all day and I stumbled on an interesting document authored by the CIA from 1944 (for fairness’ sake: the CIA was formed in 1947 out of the Office of Strategic Services, OSS, so the doc is from them) — SIMPLE SABOTAGE FIELD MANUAL (practical recommendations for sabotage). Attached the document.</p> <p>So there’s this paragraph that hooked me: General Interference with Organizations and Production (aka how to properly undermine the work of organizations and production), which, well, really reminds me of how a lot of product teams and small businesses/startups operate. I’ll leave the best bits, tweaked the text a tiny bit for modern reality :)</p> <ul> <li>Insist that everything be done through “official channels.” Never allow shortcuts to speed up decision-making.</li> <li>Give “speeches.” Talk as often as possible and for a very long time. Illustrate your “points of view” with long stories and tales from personal experience. Don’t be shy about dropping in a few appropriate “patriotic” remarks.</li> <li>Whenever possible, refer all matters to committees for “further study and consideration.” Try to make the committees as large as possible — never fewer than five people.</li> <li>As often as possible, bring up irrelevant issues.</li> <li>Haggle over the exact wording of messages, minutes, resolutions.</li> <li>Refer to issues decided at the previous meeting, and try to again call into question whether that decision made sense.</li> <li>Call for “caution.” Be “unreasonable” and urge your colleagues to be “reasonable” and avoid haste that could lead to awkwardness or difficulties in the future.</li> <li>Worry about the legality of any decision — raise the question of whether the planned action falls within the group’s jurisdiction or might conflict with some higher-up level’s policy.</li> </ul> <p>Tips for managers in the comments :)</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/728?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> AGI is close! https://darkolorin.com/posts/727.html https://darkolorin.com/posts/727.html <p>AGI is close!<br> A video by my good friend Kirill<br> from here (<a href="https://x.com/kirillzzy/status/1784273266083041475?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/kirillzzy/status/1784273266083041475?s=46</a>)<br> The device (<a href="https://www.rabbit.tech/rabbit-r1?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.rabbit.tech/rabbit-r1</a>)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/727?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Some Chinese “mesh thing” competing with OpenAI Sora. https://darkolorin.com/posts/725.html https://darkolorin.com/posts/725.html <img src="https://cdn4.telesco.pe/file/Fnd4MU47f8kELpPo24yFywpOeFCuzYzKTEUvesTZAbRQlW6euzVw0rht6RWTETop4R0h5FXAarLUHx19iN-4dHV-yIpSd6rx8PiAGcKoi3qJTHPe4vHjEJOqJbkvzUQwcONlDwyK7Jf1q9TWBLSEDNPIVflUoP-fWVnddKjd-Vz3dHLnEeD8z-g80tKgSLYQftvD5VHJE7MP0pN_rXY0eIwjblm6QOX54xabd1JKwslvDVcbgBobEpx2Zh-GQyVlzpycJ5hMNWTfNTgsNVv2XD_QIeNFljCxe0f5b9kvA-4wiVpxIaJmdU4i6uPTm5q9Tz9vaYy0FrhpycQFUFAu2g.jpg" alt="" loading="lazy"> <p>Some Chinese “mesh thing” competing with OpenAI Sora. You can’t imagine how many Valley investors got all worked up and immediately go like, yeah the videos are cherry-picked, and that’s definitely the max it can do and all that. Not used to getting competition from some no-name Chinese folks, when you’re pouring in billions of venture money.</p> <p>Screenshot has a comment from an a16z partner.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/725?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Today I stumbled on a post in a familiar channel about searching for a COO for… https://darkolorin.com/posts/724.html https://darkolorin.com/posts/724.html Fri, 26 Apr 2024 18:59:19 GMT <p>Today I stumbled on a post in a familiar channel about searching for a COO for a startup. And I decided to share some thoughts on hiring / finding people for a team. <br> A long time ago, at the dawn of my career, while still working at Yandex, I went to a design conference (don’t ask why), and there were a few people there from early Facebook who, back then, were kind of significant to me, and they talked pretty extensively about their hiring principles. Which I tweaked a bit and I’ll lay out below.</p> <p>In short, people work for 4 reasons. Those reasons can be sorted by importance for each candidate. <br> 4P = people, product, process, profit.</p> <p>Let’s go through each.</p> <p>1/ People. This is that exact case of “come work with me.” Most often it actually boils down to two things. Either friends work with their friends (fun, ease). Or with really strong folks. Strength in this case has to be measured unambiguously (number of exits, releases, number of people on the team, ARR, references, results in hard skills like papers, GitHub stars) and not by any abstract qualities that everyone will rate their own way (empathy, a culture of respect, etc.).</p> <p>2/ Product. The main thing is what I’m doing and for whom. That’s exactly why I once wanted to work at Yandex. Because the product I was building was used by millions of people a day. And I didn’t give a shit with whom or for how much. What matters here is product-person fit. Easiest in consumer, hardest in enterprise.</p> <p>3/ Process. This is when an engineer goes where the problems are harder. Or their internal motivation sits in the area of solving the most interesting problems possible in terms of hard skills. That’s why these people have very high standards for the quality of what you’re doing. And they love crisp task setting and the reasons why those tasks are needed here.</p> <p>4/ Profit. Here I think everything’s clear. Money is money, even in Africa. It’s important to think about it over a longer distance, not just short-term.</p> <p>And then we overlay all of this onto some small company or a startup. And in a startup’s case the chances to hire anyone are: <br> - a startup is synonymous with growth and ambition; if your project isn’t raising rounds and isn’t growing like crazy every quarter and there aren’t rumors wandering around you about a new unicorn, it’ll be hard <br> - a startup is experience first and foremost, and experience that’s directional and very ambitious; the higher the goal, the easier it is to attract people <br> - and, well, founders’ track record — everything’s pretty clear here, if there are no refs on you online or no meaningful releases (exits) — the chances are lower</p> <p>The company’s goal often is to help a person get to their goal or to find them that goal.</p> <p>I wrote about hiring also <br> here (<a href="https://t.me/chillhousetech/606?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/606</a>) <br> and <br> here (<a href="https://t.me/chillhousetech/560?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/560</a>) <br> and <br> even like this (<a href="https://t.me/chillhousetech/559?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/559</a>)</p> <p><a href="https://t.me/chillhousetech/724?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One of the biggest buybacks I can remember from Google, first-ever dividends. https://darkolorin.com/posts/723.html https://darkolorin.com/posts/723.html Thu, 25 Apr 2024 22:36:37 GMT <p>One of the biggest buybacks I can remember from Google, (<a href="https://www.nbcnews.com/tech/tech-news/alphabet-issues-first-ever-dividend-70-billion-buyback-rcna149427?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.nbcnews.com/tech/tech-news/alphabet-issues-first-ever-dividend-70-billion-buyback-rcna149427</a>) first-ever dividends. Time to short Nvidia. And overall, this is a very strong market move and confidence in your own business, when you’re spending 70% of the company’s cash on buying back shares.</p> <p><a href="https://t.me/chillhousetech/723?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A very interesting article about the experience of working with Sora from… https://darkolorin.com/posts/722.html https://darkolorin.com/posts/722.html Thu, 25 Apr 2024 18:18:03 GMT <p>A very interesting article (<a href="https://www.fxguide.com/fxfeatured/actually-using-sora/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.fxguide.com/fxfeatured/actually-using-sora/</a>) about the experience of working with Sora from OpenAI, the thing that generates video based on your prompt. All the problems are right there, in plain sight. And that’s exactly why all of this is still in deep alpha and I’m afraid it won’t reach normal video production anytime soon. But as a tool for sketches or quick clips, I think it’ll be enough pretty soon. By the way, right now it generates at most 720p, but it’s still slow, so testers do 480p. As for consistency, I’ve been talking about that for a long time. Here it’ll be critically important.</p> <p><a href="https://t.me/chillhousetech/722?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A pretty interesting paper came out of the depths of OpenAI. https://darkolorin.com/posts/721.html https://darkolorin.com/posts/721.html Thu, 25 Apr 2024 17:18:37 GMT <p>A pretty interesting paper (<a href="https://arxiv.org/pdf/2404.13208" rel="noopener">https://arxiv.org/pdf/2404.13208</a>) came out of the depths of OpenAI. They talk about LLMs being vulnerable to attacks like prompt injection and jailbreaks, which let attackers rewrite the model’s original instructions. The authors claim that one of the key vulnerabilities underlying these attacks is that LLMs often treat system prompts as having the same priority as user requests. To deal with this, they propose an instruction hierarchy that explicitly defines how models should behave when instructions of different priority conflict. Then they propose a method for automated data generation to demonstrate this hierarchical instruction-following behavior, which trains LLMs to ignore lower-priority instructions. After testing, they say it gets better and there are fewer vulnerabilities.</p> <p><a href="https://t.me/chillhousetech/721?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Media on Telegram https://darkolorin.com/posts/720.html https://darkolorin.com/posts/720.html Thu, 25 Apr 2024 16:01:38 GMT <p>This media is unavailable in the public Telegram preview. View the original post on Telegram.</p> <p><a href="https://t.me/chillhousetech/720?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> NASA restored contact with Voyager 1, which is now 24 billion kilometers away… https://darkolorin.com/posts/719.html https://darkolorin.com/posts/719.html Wed, 24 Apr 2024 22:44:57 GMT <img src="https://cdn4.telesco.pe/file/DaD4Q5OpnJKjDuoWmwmoHenVRGA2PctV6zUC_YR5dghHH19Tk8KmxkoxTa74Zz9GCJ8ig7PQhTtZF1XABShTpL6SQUBwPRJxCWPH5JJNVQHjoacBQ9xUGDtjGE0OJpfT0KhTNWD5C_mcrzNikmP1LIH7iC3h0yJ1_xlC5AD2__DmDMOUKBKRutMoKxFNGj2KU-HWT8q0iZeMJVaSrixeIOqWxjZldqNnAxWa5Tdfp3Ge9cAlkO-XbTHF4wOlAHonWzNM9cKarTuTX69kbplKOEq7ECkkKGwTlqqRgHLypJ2n8LQ_lwrnnJF3L0LL89illO6bXVO-LE25aqlY_bS6ow.jpg" alt="" loading="lazy"> <p>NASA restored contact with Voyager 1, which is now 24 billion kilometers away somewhere out in deep space, it’s 46 years old.</p> <p>In November 2023 the signal from Voyager turned into an unreadable sequence. They started sending it different commands from Earth, trying to kick off at least some program (the signal there-and-back takes 45 hours now.) On March 3 they spotted a standout sequence in the unreadable data. Turned out the craft had sent a full dump of its memory contents.</p> <p>After decoding it, it became clear that 3% of the memory was corrupted; one of the ROM chips died, the one storing part of the program code. The engineers redistributed the control code across system memory, on April 18 they sent new firmware, and since April 20 meaningful data has been coming in from Voyager again.</p> <p>Voyager was designed for 5 years of operation, by the way.</p> <p>Startup bros are thrilled!</p> <p>P.S. Found the news somewhere in the vast wastelands of the internet.</p> <p><a href="https://t.me/chillhousetech/719?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Hmmmi https://darkolorin.com/posts/718.html https://darkolorin.com/posts/718.html Tue, 23 Apr 2024 19:59:45 GMT <img src="https://cdn4.telesco.pe/file/vE0j8jJUNrUdQgxkrRo26JorALpaMnZZCJi6ZqyOVagfIsFRmZdfi3gO6MM1C27BfrVlkGnxPmVzW4SW4HLAZmbwtwbxMY_EgdRU0vHMe_g7-ihlmLznpgqJ2tcuAVmLU9Y9oL1_ggJ8EZHP9iQihNb-dTWZwoMQyJbFznLc9FwoXsi_nQ6xq2bt4HucJ1dZyqf_LJ2c38236cVxoOM9NqdC-Z4hWaJ7wPtVPKCCkY2Zc3uWVqKHH9sLWHi7qzIKsB6iJE2H3_X2kE22Z-PGcXGTOG62u1tajjqzVfKKH2SEvWwOItKYAmkvpWcgHkbIaix511zRwFKn5nnyaKh_rg.jpg" alt="" loading="lazy"> <p>Hmmmi</p> <p><a href="https://t.me/chillhousetech/718?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Predictably, the folks at Perplexity have figured out how to beat Google. https://darkolorin.com/posts/717.html https://darkolorin.com/posts/717.html Tue, 23 Apr 2024 14:54:44 GMT <p>Predictably, the folks at Perplexity have figured out how to beat Google. Easy: start selling an enterprise solution. (<a href="https://www.perplexity.ai/enterprise/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.perplexity.ai/enterprise/</a>)</p> <p>Overall, given the amount of stuff the company is doing:<br> - API<br> - search? Where they literally say in interviews they’ll win via UX (also building their own search crawler)<br> - enterprise solution, like “sell your colleagues a search bot so query data doesn’t leak” (corporate search as a business is something new)<br> - subscriptions<br> - news<br> - distribution partnerships for their search</p> <p>You get this feeling that after raising a ton of money and becoming a unicorn (<a href="https://www.perplexity.ai/hub/blog/perplexity-launches-enterprise-pro?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.perplexity.ai/hub/blog/perplexity-launches-enterprise-pro</a>) they (probably) don’t actually understand how to build a growing business or product.<br> No focus. And the founder is constantly taking photos with a bunch of cool guys and doing interviews/podcasts. My bet: if they don’t figure out what to do this year as their core, they’re in for a slow death. Yet again I’m convinced that raising a ton of money (<a href="https://techcrunch.com/2024/04/23/perplexity-is-raising-250m-at-2-point-5-3b-valuation-ai-search-sources-say/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://techcrunch.com/2024/04/23/perplexity-is-raising-250m-at-2-point-5-3b-valuation-ai-search-sources-say/</a>) before PMF (product-market fit) isn’t always the best idea. What do you think?</p> <p><a href="https://t.me/chillhousetech/717?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Decided to collect a few links to my old posts, for those who, so to speak… https://darkolorin.com/posts/716.html https://darkolorin.com/posts/716.html Mon, 22 Apr 2024 23:49:52 GMT <p>Decided to collect a few links to my old posts, for those who, so to speak, didn’t scroll that far. <br> I think they’re more or less worth paying attention to:<br> - <br> about networking on Twitter (<a href="https://t.me/chillhousetech/10?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/10</a>)<br> - <br> about TikToks for dogs and cats (<a href="https://t.me/chillhousetech/17?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/17</a>)<br> - <br> about how I was the first official visitor of Apple Park (<a href="https://t.me/chillhousetech/76?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/76</a>)<br> - <br> about pushes in apps (<a href="https://t.me/chillhousetech/172?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/172</a>)<br> - <br> about mute (<a href="https://t.me/chillhousetech/174?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/174</a>)<br> - <br> my draft deck for an early-stage consumer product (<a href="https://t.me/chillhousetech/214?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/214</a>)<br> - <br> about early-stage startup analytics (<a href="https://t.me/chillhousetech/395?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/395</a>)<br> - <br> what an investor update looks like — email template (<a href="https://t.me/chillhousetech/449?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/449</a>)<br> - <br> about patterns of fast-growing apps (<a href="https://t.me/chillhousetech/482?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/482</a>)<br> - <br> a bit about community (<a href="https://t.me/chillhousetech/503?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/503</a>)</p> <p>- <br> why I gave up on being public (<a href="https://t.me/chillhousetech/543?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/543</a>)<br> - <br> about what I would do differently if I were starting Prisma now (<a href="https://t.me/chillhousetech/544?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/544</a>)<br> - <br> who I’m following with the bell on on Twitter (<a href="https://t.me/chillhousetech/556?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech/556</a>)<br> The rest seems more or less fresh</p> <p><a href="https://t.me/chillhousetech/716?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Media on Telegram https://darkolorin.com/posts/715.html https://darkolorin.com/posts/715.html Mon, 22 Apr 2024 16:00:05 GMT <p>This media is unavailable in the public Telegram preview. View the original post on Telegram.</p> <p><a href="https://t.me/chillhousetech/715?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Media on Telegram https://darkolorin.com/posts/714.html https://darkolorin.com/posts/714.html Mon, 22 Apr 2024 16:00:05 GMT <p>This media is unavailable in the public Telegram preview. View the original post on Telegram.</p> <p><a href="https://t.me/chillhousetech/714?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Useful UX patterns. https://darkolorin.com/posts/713.html https://darkolorin.com/posts/713.html <p>Useful UX patterns. Part “no idea which number.” How to show which switcher or setting to flip on—turn on a video and leave it in a window. Правда there’s a risk the window gets covered, and you can’t control the position of that kind of player (the video will cover the on/off switcher, for example, that’s a problem).</p> <p>App by the Zenly folks — amoamoamo</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/713?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, onto actually useful inventions. https://darkolorin.com/posts/709.html https://darkolorin.com/posts/709.html Mon, 22 Apr 2024 13:52:25 GMT <p>So, onto actually useful inventions. Special urinals for urine analysis in China. <br> Yeah yeah, you show up, pay some cash, and a full urine test lands on your phone. Now *that’s* what I call health-improvement measures. I’m sure hardly anyone tracks their markers in general, and these kind of preventive measures could totally boost public awareness. Payment via WeChat QR code. <br> Source (<a href="https://x.com/chris__pc/status/1782245409857102262?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/chris__pc/status/1782245409857102262?s=46</a>)</p> <p><a href="https://t.me/chillhousetech/709?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sometimes I think what kind of tool should even exist for designing modern… https://darkolorin.com/posts/708.html https://darkolorin.com/posts/708.html <p>Sometimes I think what kind of tool should even exist for designing modern apps (UI/UX). Anyway, I found a reference point and I believe some hypothetical Figma will come to this kind of look in like 3–5 years.</p> <p>P. S. The game is<br> Tiny Glade on Steam. (<a href="https://store.steampowered.com/app/2198150/Tiny_Glade/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://store.steampowered.com/app/2198150/Tiny_Glade/</a>)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/708?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well looks like TikTok is a goner in the US. https://darkolorin.com/posts/707.html https://darkolorin.com/posts/707.html Sat, 20 Apr 2024 19:48:54 GMT <p>Well looks like TikTok is a goner in the US. Such is life. <br> The US House of Representatives voted for a potential TikTok ban; now China’s ByteDance will have to sell the social network or stop operating in the country.</p> <p><a href="https://t.me/chillhousetech/707?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So what about how the App Store steals from devs? https://darkolorin.com/posts/706.html https://darkolorin.com/posts/706.html Fri, 19 Apr 2024 12:21:35 GMT <img src="https://cdn4.telesco.pe/file/vq1aBHMq1JbabBkBytj3Wov0kXKwSvQ9dq16ZIWJ4uDQBdNv3dBT4e9o9_L56hHHUfLEZNmA7o8x8ozw0uyYtXNUsVLJj_I441gh6RcFxYDAaWbCGT66tmPqaurb8dKQgRRSpfcBWxLx1pY1ZzoXqUQvWMhsIqdgCWoyunZRi06-O-UstsNPgC9c6Kyo3aIjMT6in9rZSJJK0GtOV1FjMyb1M_XoK0WMAAY62I_MdG31IS4sVnWWsS7tqfrbCkEkqULvkOuW8iVP8i_uzz01FxjyscL4m0xzydqvMy7IQEJpNDroNhy4cWFXbn-g8eOX5dzbMB7GtmJv3xy7sEIGMg.jpg" alt="" loading="lazy"> <p>So what about how the App Store steals from devs? And those horse-sized commissions?<br> But yeah, overall it’s just the only viable (something you can actually survive on) option for a product like this. And then USDT on TON finally made it over too.</p> <p><a href="https://t.me/chillhousetech/706?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Dropped a new little model from Meta AI — Llama3, by the benchmarks it comes… https://darkolorin.com/posts/705.html https://darkolorin.com/posts/705.html Thu, 18 Apr 2024 17:28:58 GMT <p>Dropped a new little model from Meta AI —<br> Llama3 (<a href="https://llama.meta.com/llama3/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://llama.meta.com/llama3/</a>)<br> , by the benchmarks it comes out the best in Open Source (though looks like post-Mistral isn’t in that comparison).<br> Once I try it — I’ll share impressions.</p> <p><a href="https://t.me/chillhousetech/705?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Not an ad. https://darkolorin.com/posts/704.html https://darkolorin.com/posts/704.html Thu, 18 Apr 2024 14:11:59 GMT <p>Not an ad. My good friend Yura, who recently went through YC (<a href="https://www.ycombinator.com/companies/fluently?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.ycombinator.com/companies/fluently</a>) and raised $2M USD, wants to do a live thing and all that and share various wisdom about what it’s like in fundraising right now. He’s got a long, rich story, worked with a bunch of people. I’ll drop the info in the comments — welcome.</p> <p><a href="https://t.me/chillhousetech/704?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Product post. https://darkolorin.com/posts/703.html https://darkolorin.com/posts/703.html Wed, 17 Apr 2024 22:56:47 GMT <img src="https://cdn4.telesco.pe/file/MUm19LDssTtAGTT6WuqwheW1jy2XDTc7n3iGI_rfND3WBk2qOtITbiXKsXlF533gWCIkWEsnqURstOwrEB4TXJh0GeR-7JCU40uidEtVlbUODQNEvB-wkiASjyKOuTWX4IdbAa3pTGqNuqO9LxgXkL78yYv-s4NW5tHofC_ygW3DdFu1-Gx6Q1wmxbJQ_0LUFs4O1E6RetcdMPSyzKSLHdYXoIPfbmOd1hoGksG9fX-TllMysagZCxwnTI74MvnOJKbVLKZDnzYsYyA3DFpKEDf4qn80ozwUv_CaQAiCk9V7rFdSLmZVPy2YqHWOBT4MFckBXmkp5xVpFXbrG4CzQQ.jpg" alt="" loading="lazy"> <p>Product post. For the last 10 years I’ve been doing product one way or another in the role of some kind of manager. In all my companies I’ve always been the CEO and the product manager (the one and only). Not always successful, but I always tried to understand what people on the other side of the screen actually need. <br> On top of all that good stuff I’ve always been into a product like search. And now there are attempts to reinvent it or at least change the сценарии. Because, finally, an answer can be generated based on the information that already exists. But that’s not what I’m talking about. <br> Across the CIS (yeah, and I think outside too) everyone knows Yandex. I worked there, started my career there, so to speak. And I always considered their search a great product. But recently the guys introduced Neuro. To keep it simple it’s some kind of аналог of Perplexity. (<a href="https://apps.apple.com/gb/app/perplexity-ask-anything/id1668000334?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/gb/app/perplexity-ask-anything/id1668000334</a>) <br> And everything would be fine, if not for the product “bold” decision to put a switcher for toggling right under the search bar. Why this is “not it”:<br> - For like 10 years already Yandex has been hammering “anything can be found”, “ask Yandex”, basically bringing themselves closer to the verb “to Google” (just google it). Ads, videos, marketing — everything. And that was very right. There’s a cool Western example. <br> - Yandex will find it. Except nope. They themselves launched a competitor to themselves and right in the same place. <br> - And here I am, a regular user, I go to Yandex and see two options — so where do I type my query? Because both answer my questions. <br> - Cognitive dissonance when choosing between hard-to-distinguish options is a disaster. And I’m extremely surprised the product folks behind this decision don’t know that. <br> The reason most likely exists and it’s hidden in chasing the hype from Western коллеги. Because we can do it too, and even cooler. And I also suspect they weren’t afraid to do it because mostly queries without expensive ad listings will flow there (facts and the like).</p> <p><a href="https://t.me/chillhousetech/703?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Alright then. https://darkolorin.com/posts/701.html https://darkolorin.com/posts/701.html Wed, 17 Apr 2024 14:33:12 GMT <img src="https://cdn4.telesco.pe/file/Xaetzf4AhnbjWsJrtjaw3FTDp0DLBul4fc-MqqRItQvS8XlQp0xNU1OErnHQTMG9L-A9tPdqmiZwlr9kFxk37d-07--jaKpJK85MLm6DSEDj9bUJXwIZbhzmAqvW0v0BHh2ftprJyyWlWru6HOp-nJqi0iNqKKf5SqArnLTL-1Q9iTLh-4Mk9QJ9Cl-LntEiUC49hFapqzAlIODr2REKEIKzXzS_RbOcN2mPTl-bhkfCaeORjwYGmn9L7abMMdGnud7YpYXdsQiL79KqE49vzGUbMUVlVSpnIV3rbRLL7YHivXT9TZoKXVpgnfjIkk0-ymnyXETXeKhMBOsL7gdfng.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/dwNFGAg10aMEklZE6vFPUFNwEJCKmB1qNWvquRyeEbohze9EX0ReUtt50i0UcNKOJ59vGze8Y0HUG6qWcTGRGXYIQVwot8zaiCTPCTgn5o468AE8QtZFzcYJR7jYISrfqbM3_kSiIzfQBFez4qzlKS0ylIVBFGTU-eUmrGWnrKVHnyvTBpvQNJ-d77b6ROcR2xDiuVs6gMkmwumRp3ojzomIymZ5N_B2RCs0hjN9heiIhB081t4WYV0iI_4CpBRv35aesMvnganjFQeJMk0GCHbryhbcn6056Dos6bzOPQniVq9lLwwBHNxgGvV8C7F6h2pd0ibuo79cANvOSdSrXA.jpg" alt="" loading="lazy"> <p>Alright then. I stumbled on a contest (<a href="https://www.waicas.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.waicas.com/</a>) for AI models (not the LLM kind, but the ones in Instagram or other social networks) — Miss AI. Interesting times we live in. And the creators of other such models are the judges, too. <br> Reminds me of a prompt contest.</p> <p><a href="https://t.me/chillhousetech/701?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The price of training foundation models. https://darkolorin.com/posts/700.html https://darkolorin.com/posts/700.html Wed, 17 Apr 2024 10:31:41 GMT <img src="https://cdn4.telesco.pe/file/T41yDEGKn3_8gHRX4LENIsQN9VLyR1K-y-UdJ_r0LOGaJ0yKHFSHoVUOkq23EXouwXFNZKntdTv8bul1MexHou2qHz0gWXvHLJeLtpdLrztXq9zFGfjyYkFWmBa1sfjqnPRPEjUYWc-P8M5qum75cAp1mG6hKl_rA2oNxOWHzoe2D9SG9r_HqGajJ7fwYNNpflLJ4LkjtZDGYsaXy8obM6xwpgkdRjei5NwiPboZW6SPlse0Xem5oDrY1TWYNiUWgRw5FF4FwPET5NB-Sfc2WmNCCjzLj6ochLLJQ0khM1sfVUb8dfHoV3uK3fULRXK7YM0wgupdfhTLQuH0dFIF-g.jpg" alt="" loading="lazy"> <p>The price of training foundation models. Well then, ~200M USD for Google Gemini Ultra. And about 75M for gpt4. <br> And yeah, compute is officially over :) wait for this whole thing to roll onto CPUs. <br> Wonder how much mistral costs</p> <p><a href="https://t.me/chillhousetech/700?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Nice move with the interview. https://darkolorin.com/posts/699.html https://darkolorin.com/posts/699.html <p>Nice move with the interview.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/699?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Everyone gets so hyped about mentoring, helping, searching for meaning, being… https://darkolorin.com/posts/698.html https://darkolorin.com/posts/698.html Tue, 16 Apr 2024 21:48:10 GMT <p>Everyone gets so hyped about mentoring, helping, searching for meaning, being a teacher. Well yeah, in general. <br> How does that saying go — <br> * <br> it’s not like hauling sacks. <br> I get hyped too, that’s why I teach students when there’s a chance. And I treat it definitely not like a way to make money. Wonder how many of those guys are ready to help startups for free and without equity. Rhetorical question, obviously.</p> <p><a href="https://t.me/chillhousetech/698?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well if it’s 5 sources then we’ll believe it https://darkolorin.com/posts/697.html https://darkolorin.com/posts/697.html Tue, 16 Apr 2024 15:50:43 GMT <img src="https://cdn4.telesco.pe/file/o-WBXvgbTGSmDAQMfIhgLHZbaXNt7fisMjSFM1drc68Lt_m_XU19RPwRKeomfM6vikPeIgwaCG_-m4VdrZ_97q7q2DbYrxHzJwsmU70_TKEvgXgPdMZyAiyzqjByBVlRbwV14DEnuktwHo4RvfHR3S4ZELAZK_YbSYt8HUTCJl7WVmed3r4WDV9BJuoaPFrpokiNkM36ydKYx6Z7K7CVQyAZT2_blL-xhW53I6Upat1HSXMW9iKgOpUVQuHcT0LUZTWWBVUgXmuOLYIwOWltCZLJOU0ahoGd-0QVPHECYtJ9FOqlU_inuXeiK7E9iKxjcI8Vtb8FGEySvLDNb5Ng-w.jpg" alt="" loading="lazy"> <p>Well if it’s 5 sources then we’ll believe it</p> <p><a href="https://t.me/chillhousetech/697?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Let’s talk about the hyped-up release in certain circles, AI Pin by Humane. https://darkolorin.com/posts/696.html https://darkolorin.com/posts/696.html Tue, 16 Apr 2024 14:59:19 GMT <img src="https://cdn4.telesco.pe/file/N7OPoEAU_HsiA6nrUV5BU5KfwR5oj2_TGI3J-CEvsH1-mb9Lo6KcqIpv6YdDZ_E-MPNOhp048dOw_vtNYi6_SlBFbO53BYh1TkD0MoWBv-WX_JPGULueV4R7p8W6PPufJtmfrIXbr07B7OQUxzQhQp4sCOYnYFWb-G7XLgvDVdtgqIJly8P-UG_ZKhguile5IjVc0hbKKlzpKgz0tCxkwlaZdid6xu2iUDpsGm4Z0XjoJ926ZOVfqJiBDlRdsA65aU_JiMpMMUkkq72bx7PeYoaJ8QVHqxYKGxWjyBTDsOYSZSZPLAbNQWmu7N6a3QTrbPNB0fto_x4laSuyNgzi2g.jpg" alt="" loading="lazy"> <p>Let’s talk about the hyped-up (<a href="https://www.youtube.com/watch?v=TitZV6k8zfA&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=TitZV6k8zfA</a>) release in certain circles, AI Pin by Humane (<a href="https://humane.com/aipin?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://humane.com/aipin</a>). Besides the fact that the company name is kinda straight-up disgusting to me, I’m surprised how people who worked in corporations have absolutely zero understanding of balance between product and features. Like, the guys announce just a ton of functions, create ridiculously huge expectations, crank marketing to the max (Apple-style, “inspiration” and all that). But it turned out all of this is basically a byproduct of having 100M USD onboard from investors.</p> <p>You can assemble a cool device, make the casing nice and all that. But actually mastering the product—turned out they couldn’t. And can we stop spinning this idea that the first iPhone was shit too, because it wasn’t. Sometimes you need to admit you’re making some garbage and take all the criticism.</p> <p>If I were them I’d make just one feature and make it really damn good—especially since it was there from the start—and that’s Live Translation. Because that’s actually convenient, anyone who travels knows, especially in Asia. Everything else is just some incomprehensible crap to me, where any $250 phone does it faster and better (and this pin costs $700, btw). And the price bites so hard it means they’re targeting an adult customer, not Gen Z with pocket money.</p> <p>And yeah, the typical launch mistake is in the screenshot.</p> <p>P.S. You have to be ten times better than second best. (c) Peter Thiel</p> <p><a href="https://t.me/chillhousetech/696?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In short. https://darkolorin.com/posts/695.html https://darkolorin.com/posts/695.html Mon, 15 Apr 2024 22:53:44 GMT <p>In short. Got reminded of something.</p> <p>Or let’s talk about the topic of “money for advice.” Like paid expertise. Where, like, I’m so badass that my advice costs a shitload.</p> <p>It’s not that I don’t believe in a thing like paying for advice. But not in the moment, probably. Advice by itself isn’t work. I see it more like handing out social credits. Like, I’ll help you with advice, and then you’ll help me later too. Basically you’re building up not just reputation, but experience (since people show you the situation, give you the inputs). Plus your credit score in the community goes up.</p> <p>I do believe there are, like, omega-cool guys who don’t need any of that. But in my experience guys like partners at a16z, various unicorn founders will easily agree to give you advice, take a look at your idea. Even if they write back “bullshit” (more often they just won’t reply after the first message), it still matters for you.</p> <p>So selling advice is scary crap to me. Self-limiting and partly smugness. Consulting is work over a stretch of time, but definitely not a one-off answer to a couple questions. No time — just say no.</p> <p>But everyone decides for themselves.</p> <p>To make this world a tiny bit better, even with just a piece of advice — that’s already cool.</p> <p><a href="https://t.me/chillhousetech/695?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Venture industry crisis. https://darkolorin.com/posts/694.html https://darkolorin.com/posts/694.html Mon, 15 Apr 2024 21:46:31 GMT <p>Venture industry crisis. Winter! Nobody’s giving money. Oh, right. <br> <a href="https://www.axios.com/2024/04/15/andreessen-horowitz-7-billion-new-funds?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.axios.com/2024/04/15/andreessen-horowitz-7-billion-new-funds</a></p> <p><a href="https://t.me/chillhousetech/694?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Western Twitter is hyping up a new Clubhouse from Naval Ravikant — AirChat. https://darkolorin.com/posts/693.html https://darkolorin.com/posts/693.html <p>Western Twitter is hyping up a new Clubhouse from Naval Ravikant — AirChat. <br> It’s an app, a feed of discussions, where you type text like it’s a voice message and then there’s also transcription happening. Basically you get both a text message and your audio (kinda like an audiobook or podcast experience). <br> The whole Valley crowd has flown in there again and is sitting around chatting (apparently they’ve built up things to say since the Clubhouse launch). As always, everything’s invite-only. Which regions it’s available in — no idea, the invite seems to go through a phone number. If you need one — DM me. <br> Still think that in the long run audio content loses to video. Though here they decided to get clever: combined it with text and kept the author’s voice. We’ll see. <br> App link (<a href="https://apps.apple.com/gb/app/airchat-just-talk/id1469462351?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/gb/app/airchat-just-talk/id1469462351</a>)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/693?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> This is basically me when I use some “new revolutionary AI device or software” https://darkolorin.com/posts/692.html https://darkolorin.com/posts/692.html <p>This is basically me when I use some “new revolutionary AI device or software”</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/692?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ve got a bunch of vacancies, people are always asking me to help fill them… https://darkolorin.com/posts/691.html https://darkolorin.com/posts/691.html Fri, 12 Apr 2024 22:51:40 GMT <p>&quot;&quot;&quot;<br> I’ve got a bunch of vacancies, people are always asking me to help fill them, on top of the eternal help with fundraising. CPO, Product Managers, AI/ML. Interested? Should I post them?</p> <p>- Yes<br> - No<br> - Better help me raise a round<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/691?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> News on context length. https://darkolorin.com/posts/690.html https://darkolorin.com/posts/690.html Fri, 12 Apr 2024 18:23:09 GMT <p>News on context length. Well, I expect that soon you’ll be able to toss books, videos, movies into the net, better just by link so you don’t have to bother. Then they’ll bring all that to the device too (at that moment dump your NVIDIA shares).</p> <p>Don’t confuse it with the model getting smarter or learning to understand the world and all that. It’s just that now the same technique the transformer is built on will work, like, on a very long input prompt. Looks like that’s how the latest Google release is done.</p> <p>Read the paper (<a href="https://arxiv.org/pdf/2404.07143.pdf" rel="noopener">https://arxiv.org/pdf/2404.07143.pdf</a>)</p> <p><a href="https://t.me/chillhousetech/690?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> You: pay for a subscription to ChatGPT, Claude3 and all that. https://darkolorin.com/posts/689.html https://darkolorin.com/posts/689.html Wed, 10 Apr 2024 22:51:40 GMT <img src="https://cdn4.telesco.pe/file/m3z1YKcPXlraDErOa_IJPXyTu0bPW4a2fKbhbjqIwiju3EZh6Jy1BJhrGdtnaQM1nJs1sBR18BSHupi0xyNvJ_CBtLETife5Dgs5TmEWY57vkHTJxeWKKdOisdPADe8QaifCUJg27kQ2Uqp5GttxIx83gsYSRs-6aKpntpgXAgF3_Ach1UJcuSJREBY2XRZyionb407f5QaLN5kIdLeksU0KOjNb60Bdf2f-dp3gSrqDp1c-HEctJgV8YFT_9hU0wCkWbUcEdQjX1tEcGdYM4YLL8v4Enps-On6_2LK8hCsrDLLal3mqqTEMXLaOXHNuiwuZ7XkwPc_UAfulROCYeQ.jpg" alt="" loading="lazy"> <p>You: pay for a subscription to ChatGPT, Claude3 and all that. <br> Me: use the assistant in the Shopify app for free. <br> P.S. It even understands with typos.</p> <p><a href="https://t.me/chillhousetech/689?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile Pavel Durov is telling us that Telegram’s DAU/MAU is ~1/2, which is… https://darkolorin.com/posts/688.html https://darkolorin.com/posts/688.html Wed, 10 Apr 2024 19:40:34 GMT <p>Meanwhile Pavel Durov is telling us that Telegram’s DAU/MAU is ~1/2, which is an insanely high metric. And combined with the fact that it’s a messenger, retention should be through the roof too. In short, a grandiose result. Waiting for it to turn into a business. <br> Similar numbers were seen across all successful consumer (social) companies, and their IPOs were very strong as well. Almost everyone went with an ad model (the exception is probably WeChat, which started building a superapp, but the explanation is more in its locality). <br> So: DSP, auction, discover, etc. Or we’ll be watching something new. Right now the movement is closer to a Snap/Discord strategy. And their future as a business looks unstable to me. <br> Would be nice, of course, to see daily time spent per user. Then we’d immediately know the ad potential (and possibly the need to build a feed like in that same WeChat).</p> <p><a href="https://t.me/chillhousetech/688?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In one line. https://darkolorin.com/posts/687.html https://darkolorin.com/posts/687.html Wed, 10 Apr 2024 10:52:01 GMT <p>In one line. A bunch of interesting releases dropped over the last couple days.</p> <p>Google launched Gemini 1.5 (<a href="https://developers.googleblog.com/2024/04/gemini-15-pro-in-public-preview-with-new-features.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://developers.googleblog.com/2024/04/gemini-15-pro-in-public-preview-with-new-features.html</a>) with a huge context window (up to a freaking million tokens, in the example they load a Jeff Dean lecture at 117k tokens) and the ability to upload video/audio.</p> <p>Today the French guys at Mistral dropped (<a href="https://twitter.com/mistralai/status/1777869263778291896?s=61&amp;t=BDFpg9yVlGUdk189o0twQA&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/mistralai/status/1777869263778291896?s=61&amp;t=BDFpg9yVlGUdk189o0twQA</a>) a new model version. The foundation model market is getting hot and it’s already hard to say who’s gonna be the leader in the next few months, while OpenAI isn’t really shipping anything new. They dropped it in their usual style, just leaving a torrent link on Twitter.</p> <p>MoE architecture (<a href="https://huggingface.co/blog/moe?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://huggingface.co/blog/moe</a>)</p> <p>World-famous AI expert Andrey Karpathy is starting to write his own library (<a href="https://github.com/karpathy/llm.c?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/karpathy/llm.c</a>) for training nets in C. I suspect he’s prepping us for a world where nets get trained inside smart devices at home — in fridges and stuff.</p> <p>A new version of GPT4-Turbo with Vision rolled out (you can shove images in) with a 128k token context. No detailed description or changes yet. Waiting for tests. Meanwhile, take a look over here (<a href="https://chat.lmsys.org/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://chat.lmsys.org/</a>)</p> <p><a href="https://t.me/chillhousetech/687?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Remember the release of DeepMind’s AlphaGo, the program that played Go better… https://darkolorin.com/posts/686.html https://darkolorin.com/posts/686.html Tue, 09 Apr 2024 08:20:16 GMT <p>Remember the release of DeepMind’s AlphaGo, the program that played Go better than humans. Up to that point this game was considered the most high-variance and difficult board game, and by beating a human the AI showed, like, yeah, you can even learn by playing against yourself and still be better (i.e. without a dataset of games, just the rules—basically the games were generated by the program itself. That version was called AlphaZero) than all players.</p> <p>So yeah, that was in 2015–16. And there’s an effect. You won’t believe it, but Go players became more inventive and better. After a few years, the weakest pro players started playing better than the strongest players before the AI release.</p> <p>This is a great example of the thesis “AI will help pros get better, you just need the desire to grow and learn new stuff.”</p> <p>Source (<a href="https://www.henrikkarlsson.xyz/p/go?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.henrikkarlsson.xyz/p/go</a>)</p> <p><a href="https://t.me/chillhousetech/686?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Although why the hell am I even talking about salaries. https://darkolorin.com/posts/684.html https://darkolorin.com/posts/684.html Sat, 06 Apr 2024 08:46:02 GMT <img src="https://cdn4.telesco.pe/file/uSqGb4Snj59Ygyrhnzi9BDt4g9RZlhIMI3ZR9RFie4Mmr67KmH1JGLQgWg8jHBS2YX809L4el2o5t1S_bTrU9aWPIdciDW1rIN4HUJwxGBaRVcnsWueFEVrUKvZV6Tdg4pQC4xjdKHr3Y038HaXHpsgVh-ctwQcOl_d1tyoID63FTbiIfN1hduwPCxtFd_sQlrfOAPK86jDGz-TFrJhYW9AeFMOq-tiY2Mvr7gFBmxgFN30GwNypPnhBEG_uzXLxVpglFVRXNGreVBXzAYerwULxE3S8ExfkJvLN3bclQdlaXiaZZF2U9xfCKu56lYGSg1upjwJ31JXliEsKqec1kQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/IcqVW7eBdcfkBhW4U9Vei9KswrqX56eRotIFoUxPKv2X-v0liQbCDzWQ1mpHOeqBtTELqkPR5lOXVYPbcpo5Su1n90XcpiAhrqdnK0BSNziwuvujJmA0VI6-ytE6SjhElwM88WNGYxBk4wG-2zAoCpA5m1n4tPgVBcJ0Gpd15AR5jW269L6f5ilz5yKFbxgYfxldOBWEIoIjBUft81P2IIXeh-rqvo8RyC5YH5kFWxE_cUUlYZDopxTaUpxNkMW2DqngWAQo76sRSLouDmPfcfjyaUZWAwxKQuBq3jxCrJU73Pl7iovcphIGY-e0foLNWlfCNuQv2hFuTOPEczlTng.jpg" alt="" loading="lazy"> <p>Although why the hell am I even talking about salaries. Pff. Librarian. <br> P.S. And not just like that, but remote. Wtf.</p> <p><a href="https://t.me/chillhousetech/684?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Friday poll. https://darkolorin.com/posts/683.html https://darkolorin.com/posts/683.html Fri, 05 Apr 2024 15:42:37 GMT <p>Friday poll. There are more of us now, so I’ll ask: how many posts and what kind would you like to see? (You can pick multiple options)</p> <ul> <li>Post like it is now</li> <li>More personal takes based on experience</li> <li>Fewer news and comments</li> <li>More about startups</li> <li>More about AI</li> <li>More about building products</li> <li>More about career stuff</li> <li>More memes</li> </ul> <p><a href="https://t.me/chillhousetech/683?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Once upon a time, there were a lot of meme design posts in my channel from… https://darkolorin.com/posts/682.html https://darkolorin.com/posts/682.html Fri, 05 Apr 2024 15:29:22 GMT <img src="https://cdn4.telesco.pe/file/i0k3uDBcjQfhoLcAkDjwy4eIz1c2Q2hsUHzBNGx-P_SSocUoHonuhJfmab-TqgNyN0riwPz4saxAYFiY2oZ3QvtHOBjRsUUcgq2MUhJQFfYn1v5a8TGZ2ijKpEkuBeRfCHDJoAFoZDN5fEl4ntuKKPHr5XrNw6yHffjox37C_S20ri4MzsN6AKupx6DJy0sUzt-5uOmtIgLGzUj7RbGLvfNhWy2Pu8XLX294U991obUEBaJeFr8L-C07f8Vla6OMFl8sqXNoskxPfce_MF4b5KvMTAgGfzJKGVEq0voUbZlXnmARwWqxSfihnb53ARdEEr6YSj3jZTOyaMVFnrn0GQ.jpg" alt="" loading="lazy"> <p>Once upon a time, there were a lot of meme design posts in my channel from this (<a href="https://x.com/soren_iverson?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/soren_iverson?s=21</a>) gentleman. And honestly, he’s actually offering some not-bad stuff.</p> <p><a href="https://t.me/chillhousetech/682?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The Chinese pretty often put out interesting models. https://darkolorin.com/posts/681.html https://darkolorin.com/posts/681.html Fri, 05 Apr 2024 12:31:31 GMT <p>The Chinese pretty often put out interesting models. This time it’s a pretty worthy competitor in the area of generating portraits and photos of people. <br> <a href="https://cosmicman-cvpr2024.github.io/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://cosmicman-cvpr2024.github.io/</a> <br> Do you use any image-generation nets at all? I still haven’t managed to find any use for them for myself.</p> <p><a href="https://t.me/chillhousetech/681?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Yann LeCun’s presentation at Harvard. https://darkolorin.com/posts/680.html https://darkolorin.com/posts/680.html Fri, 05 Apr 2024 08:21:29 GMT <p>Yann LeCun’s presentation at Harvard. Even though he’s a hardcore critic of the current approach: more data, more servers, and our everything is deep learning only. Picking a kinda weird, IMO, way to criticize — like “a person can learn to drive a car at 17 in a few hours, but a computer can’t.” But it’s worth noting all his examples ignore that before that you’ve spent 17 years seeing cars on the road, racing movies, playing games, your parents driving you places, and that’s all the same dataset, and you and your brain remember all of it. <br> Overall, valid points, and anyway it’s always interesting to listen to a coherent talk from the head of one of the strongest AI labs at Meta. <br> PAPER: (<a href="https://openreview.net/forum?id=BZ5a1r-kVsf&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openreview.net/forum?id=BZ5a1r-kVsf</a>) <br> Slides: (<a href="https://drive.google.com/file/d/1Ymx_LCVzy7vZXalrVHPXjX9qbpd9k_bo/view?usp=drivesdk&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://drive.google.com/file/d/1Ymx_LCVzy7vZXalrVHPXjX9qbpd9k_bo/view?usp=drivesdk</a>) <br> Video: (<a href="https://youtu.be/MiqLoAZFRSE?si=wk-ESI_i-_Q7539w&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/MiqLoAZFRSE?si=wk-ESI_i-_Q7539w</a>)</p> <p><a href="https://t.me/chillhousetech/680?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Couldn’t let this piece of news slide by. https://darkolorin.com/posts/679.html https://darkolorin.com/posts/679.html <p>Couldn’t let this piece of news slide by. (<a href="https://cloudblogs.microsoft.com/quantum/2024/04/03/how-microsoft-and-quantinuum-achieved-reliable-quantum-computing/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://cloudblogs.microsoft.com/quantum/2024/04/03/how-microsoft-and-quantinuum-achieved-reliable-quantum-computing/</a>)<br> Even if it looks kind of funny, it’s a small step forward toward a very interesting future.<br> I believe the real breakthrough in information processing tech and working with it will come with the era of actual quantum computers.<br> This particular breakthrough probably won’t have an immediate impact outside of R&amp;D, unless it scales well. The two biggest obstacles to practical use of quantum computers right now are the not-enough qubits for anything beyond experiments, and the lack of effective error correction—which could change thanks to this breakthrough.<br> I think that in the next 4–5 years we’ll see the first real problems being solved on quantum computers more efficiently than on classical ones.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/679?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> It so happened that one way or another over the last half a year I’ve been… https://darkolorin.com/posts/678.html https://darkolorin.com/posts/678.html Thu, 04 Apr 2024 18:48:37 GMT <p>It so happened that one way or another over the last half a year I’ve been running into a lot of different companies and cases, and especially Russian-speaking startups (in different locations).<br> Noticed an interesting thing. I’ll call it the “low tax base effect.” Specialists (more often managers) who decided to move (doesn’t really matter when), really want to keep their standard of living, or even improve it. And here a moment comes that’s kinda weird to me.</p> <p>Let’s say a good manager with experience at, say, Yandex gets $5k a month net. And when he moves to work somewhere in Europe or I don’t know where else, he wants to both keep his standard of living and raise his salary. And after a quick napkin calc he ends up with, like, $10k+ net (I’m counting in bucks for convenience). And you know, that’s, well, a lot.</p> <p>Salaries at the $10k+ net per month level (I generally forgot people are used to thinking per month; everywhere it’s per year, because taxes are per year too) — that’s the US or FAANG (Facebook, Google etc, where everything’s in options). But for a manager to get 10k in a startup — that’s, uuuuh, a very крутой dude or the founder. Or better yet, don’t pay anyone that much in a startup at all. Too expensive.</p> <p>Meaning:<br> - I want to raise my standard of living<br> - and live the good life where taxes are pushing 40%<br> - and preferably the company is Russian-speaking<br> - and housing just as cool (count that as x2 in price)</p> <p>No offense. But it’s easier to hire a Stanford intern for $50k a year than an “successful” manager for $200k.</p> <p><a href="https://t.me/chillhousetech/678?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Seems like it’s been a while since I did a roundup of books that I consider… https://darkolorin.com/posts/677.html https://darkolorin.com/posts/677.html Thu, 04 Apr 2024 14:40:42 GMT <p>Seems like it’s been a while since I did a roundup of books that I consider the most, well, awesome, must read, so to speak. I’ll make a list:<br> 1/<br> The Road to Reality, or the laws governing the universe. Roger Penrose.<br> Yeah, I like hardcore, and this is it in its pure form. A heavy monograph with a ton of formulas and exercises. Properly makes your brain work. Took me about 2.5 years to get through.<br> 2/<br> Thinking, Fast and Slow. Kahneman.<br> A classic and I don’t think it needs an intro, Nobel stuff in behavioral economics, lets you really understand how decisions work in people’s heads and why advertising is the way it is.<br> 3/<br> A History of Western Philosophy. Bertrand Russell.<br> This really lets you take a solid look at the core ideas in culture, history, and the world. Super useful for your horizons.<br> 4/<br> Surely You’re Joking, Mr. Feynman.<br> I don’t even know, just read it — the best thing out there about science and physics in a light, accessible form.<br> 5/<br> The Hard Thing About Hard Things. Ben Horowitz.<br> I think if you want to understand what startups are like and how life is there and what gets done and how success comes — then this is probably the only book you need. It even seems to me you don’t need to read anything else about startups, better to keep your head clean and just this book.<br> 6/<br> Commentaries on the Gallic War. Caesar.<br> A modern guide to conflicts in the world.<br> 7/<br> The World of Warcraft Diary: A Journal of Computer Game Development.<br> I’ve spent in World of Warcraft definitely more than ten thousand hours and, probably, it’s one of the best games of all time, and the author managed to very sensitively and interestingly convey the process of making it. For everyone who loves games — must read.<br> So, kind of like that, a short list of nonfiction. From fiction — Dostoevsky and Tolkien. The rest didn’t click<br> 🙂</p> <p><a href="https://t.me/chillhousetech/677?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Who here’s a deck fan? https://darkolorin.com/posts/676.html https://darkolorin.com/posts/676.html Thu, 04 Apr 2024 13:57:20 GMT <p>Who here’s a deck fan? Catch a collection of AI startup decks —<br> <a href="https://decks.chiefaioffice.xyz/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://decks.chiefaioffice.xyz/</a><br> Overall, I’ll make a remark right away — decks don’t raise rounds, founders do, and often without a deck at all<br> 🙂</p> <p><a href="https://t.me/chillhousetech/676?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> My acquaintances launched their startup Higgsfield (clearly took inspiration… https://darkolorin.com/posts/675.html https://darkolorin.com/posts/675.html <p>My acquaintances launched their startup Higgsfield (clearly took inspiration from particle physics) <br> <a href="https://higgsfield.ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://higgsfield.ai/</a> <br> Besides my respect for Mashrabov &amp; co (the guys Snap bought back in the day), after Sora it’s gonna be kinda rough. Still, almost all current projects look like variations on <br> <a href="https://animatediff.github.io?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://animatediff.github.io</a> <br> (that thing’s about a year old, if someone hasn’t seen it), except, obviously, OpenAI. <br> Still, congrats — the guys raised $8M right off the bat, which makes sense: you’ve gotta blow a lot on GPUs and training. Hope it’ll all be awesome! Waiting for the paper :)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/675?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Lenny sometimes puts out good summaries after interviews. https://darkolorin.com/posts/674.html https://darkolorin.com/posts/674.html Wed, 03 Apr 2024 20:56:39 GMT <img src="https://cdn4.telesco.pe/file/iy06ILH5FztI2RgIiN9538XTSo0PAYHZQE8LTRm4KkKCTKLZ1x4gj1UV12F_dkjMFlzUxFZYwxP0I9_rgyKjZvYIIMN4n68yJbS_ceHqX6G4fRbKZu1kr3SG_nDZkG0GeVowY8RDQJoXSRfvRIuMN0JB-F4gvNAGCOLXhmUCnI65v5PldC5wdovosHMu8W1RZJATMww3zdu7ou_rkwDKmWb2Jhk_PlF-35lrSMkhNipZ1ZUDsls6Vh0830_jXoZ0zLh96zpaejW3OCreBWauEdr7at9SoQD-HTRWrL2ll4JDA6aDzyMQaubWpFFqAKSmA1Codmn5e4IW-wjpIVp7FQ.jpg" alt="" loading="lazy"> <p>Lenny sometimes puts out good summaries after interviews.</p> <p><a href="https://t.me/chillhousetech/674?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One of the early-stage funds has a great interactive map across different… https://darkolorin.com/posts/673.html https://darkolorin.com/posts/673.html Wed, 03 Apr 2024 19:30:18 GMT <p>One of the early-stage funds has a great interactive map across different landscapes (market directions). Use it. <br> <a href="https://mad.firstmark.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://mad.firstmark.com/</a></p> <p><a href="https://t.me/chillhousetech/673?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Seen those deepfake videos with different actors on TikTok. https://darkolorin.com/posts/672.html https://darkolorin.com/posts/672.html Mon, 01 Apr 2024 14:28:14 GMT <img src="https://cdn4.telesco.pe/file/a9ZR9-qs16H54fMWbHmHb-XeQrmxtE8G9tXXHAKuiFeZpGEQ3wi55Lq3EtjhLqf0VvBBGKKqulHnUxydWEeRjxAsduj0eI02P9qltoTJ5gpLCBr83Iq1HuuGvrQawqMlOThEMxoOqQJGuV_Mc8GWbVbkb8RMqgVNTOk8zgIXnji60WS7osgpe9-gWRKspw8otMnOFt-XsYj8vhyF493SPfGMQpHe8IY09XPE7FRV41XWys13liqq3ulc1Vac1TY88rfMt1doLS5rH4Li0pgfURFwz7AmMqrhIzOboG_Bi2d-DyGDYg1CbPAqMYSAJ85iiqUINcIaXzKbyrU8EZ72yg.jpg" alt="" loading="lazy"> <p>Seen those deepfake videos with different actors on TikTok. Like Tom Holland or Keanu Reeves.<br> So, Intel has a technology (<a href="https://www.intel.com/content/www/us/en/newsroom/news/intel-introduces-real-time-deepfake-detector.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site#gs.6q7j6k" rel="noopener">https://www.intel.com/content/www/us/en/newsroom/news/intel-introduces-real-time-deepfake-detector.html#gs.6q7j6k</a>) for detecting that a video is faked (not exactly the latest and greatest, just so you know). But the most interesting part is how they do it.<br> FakeCatcher from Intel uses a digital version of photoplethysmography (<a href="https://ru.wikipedia.org/wiki/%D0%A4%D0%BE%D1%82%D0%BE%D0%BF%D0%BB%D0%B5%D1%82%D0%B8%D0%B7%D0%BC%D0%BE%D0%B3%D1%80%D0%B0%D0%BC%D0%BC%D0%B0?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ru.wikipedia.org/wiki/%D0%A4%D0%BE%D1%82%D0%BE%D0%BF%D0%BB%D0%B5%D1%82%D0%B8%D0%B7%D0%BC%D0%BE%D0%B3%D1%80%D0%B0%D0%BC%D0%BC%D0%B0</a>) (rPPG) to detect blood flow.<br> This method works by detecting changes in blood vessel volume, analyzing color variations in the video pixels that correspond to blood flow across the face.<br> Under the hood: good old convolutional nets (they’re a great fit for this task). Meaning, formally we train a net to understand changes in facial blood flow between a deepfake and a real video.<br> Thanks to my former colleague Artur for the tip. And here’s the source. (<a href="https://x.com/thatguybg/status/1774084382527226096?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/thatguybg/status/1774084382527226096?s=46</a>)<br> P.S. In the latest YC batch, not just one (<a href="https://www.nuanced.dev/team?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.nuanced.dev/team</a>) company is doing fake detection.</p> <p><a href="https://t.me/chillhousetech/672?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> How to properly keep an eye on your competitors. https://darkolorin.com/posts/670.html https://darkolorin.com/posts/670.html Mon, 01 Apr 2024 04:49:01 GMT <p>How to properly keep an eye on your competitors. <br> As usual, the most interesting stuff surfaces in lawsuits of one company against another. This time it’s a suit against Meta (formerly Facebook). Pretty educational:<br> - the tone of the emails <br> - the structure of the decisions (obvious autocracy — well, hopefully that’s not news, because in big tech it only works like that) <br> - they write that Zuck has a lack of emotional intelligence (empathy, in a word), which isn’t surprising — what “the hell empathy” in business decisions</p> <p>Overall I’d like to emphasize that empathy and all that fellow-feeling rarely have anything to do with big money and decisions. <br> <a href="https://storage.courtlistener.com/recap/gov.uscourts.cand.369872/gov.uscourts.cand.369872.736.0.pdf" rel="noopener">https://storage.courtlistener.com/recap/gov.uscourts.cand.369872/gov.uscourts.cand.369872.736.0.pdf</a></p> <p>And the text of Mark’s email to the top managers: <br> Whenever someone asks a question about Snapchat, the answer is usually that because their traffic is encrypted we have no analytics about them. <br> Given how quickly they&#x27;re growing, it seems important to figure out a new way to get reliable analytics about them. Perhaps we need to do panels or write custom software. You should figure out how to do this.</p> <p>And yeah, they figured it out — they bought a company to track competitors through traffic analytics.</p> <p><a href="https://t.me/chillhousetech/670?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If you even need a management book, this one will be enough. https://darkolorin.com/posts/669.html https://darkolorin.com/posts/669.html Sun, 31 Mar 2024 13:31:13 GMT <p>If you even need a management book, this one will be enough. No others. Seriously, the best there is. <br> <a href="https://g.co/kgs/joN9taF?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://g.co/kgs/joN9taF</a></p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/669?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m waiting for Telegram to start pumping its ad platform hard too, because… https://darkolorin.com/posts/668.html https://darkolorin.com/posts/668.html Sun, 31 Mar 2024 13:29:23 GMT <p>I’m waiting for Telegram to start pumping its ad platform hard too, because you can’t live off premium alone. Because right now these two platforms are, like, very similar in their release structure. Except Telegram is multiple times bigger. Discord is gonna launch its own DSP, as I understand it. <br> <a href="https://www.wsj.com/tech/discord-to-start-showing-ads-for-gamers-to-boost-revenue-bf5848b9?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.wsj.com/tech/discord-to-start-showing-ads-for-gamers-to-boost-revenue-bf5848b9</a></p> <p><a href="https://t.me/chillhousetech/668?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sometimes I DM with various “famous” people on Twitter. https://darkolorin.com/posts/667.html https://darkolorin.com/posts/667.html Sun, 31 Mar 2024 12:07:25 GMT <p>Sometimes I DM with various “famous” people on Twitter. Either in DMs or in replies. Usually we argue about some stuff.</p> <p>One of the recent topics was: you don’t need any money / VC money to reach PMF. Usually talking about consumer (b2c). And the examples they throw out are Snap/FB/WhatsApp. And yeah, overall I agree with the basic idea that under certain conditions you really don’t need money to get organic growth and, so to speak, hit the market, but there’s a nuance and for some reason everyone is very professionally silent about it.</p> <p>Most of the successful examples above are stories of so-called growing along with the market, or like when the iPhone launched and boom—apps, or when cameras suddenly became ok—boom—in Insta, or when the web started spreading through campuses—FB showed up (don’t judge too hard, the point is clear, and it’s obvious that’s not the only reason for the growth). In short, as the market grows up and matures, shipping a new WhatsApp becomes very hard. User attention is expensive now because TikTok is on your path. Even if you look at it, the last explosive growth that more or less stuck is OpenAI, and there without venture money at the start—well, kinda no way (yes I get it, it’s a technology).</p> <p>And tldr: in a mature market, just taking it and getting to pmf like that is harder without money than with it.</p> <p><a href="https://t.me/chillhousetech/667?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Goodbye voice-based authorization. https://darkolorin.com/posts/666.html https://darkolorin.com/posts/666.html Sat, 30 Mar 2024 09:55:38 GMT <p>Goodbye voice-based authorization. But the most interesting thing is that more and more models are basically just scared to ship because of scammers. That’s all of today’s safety, yeah? Or am I missing something?</p> <p>They’ve learned to simulate a voice very fast and with high quality.<br> <a href="https://openai.com/blog/navigating-the-challenges-and-opportunities-of-synthetic-voices?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/blog/navigating-the-challenges-and-opportunities-of-synthetic-voices</a></p> <p><a href="https://t.me/chillhousetech/666?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A pretty cool thing from the founder of 37signals, remember those guys? https://darkolorin.com/posts/665.html https://darkolorin.com/posts/665.html Fri, 29 Mar 2024 03:59:59 GMT <p>A pretty cool thing (<a href="https://once.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://once.com/</a>) <br> from the founder of 37signals, remember those guys? <br> They say SaaS is getting old, you wanna pay once and it’s like it’s all yours. They made their own Slack <br> <a href="https://once.com/campfire?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://once.com/campfire</a> <br> But so far I kinda don’t get how to migrate. Maybe I missed it. And that’s the hardest part. Would you switch your team to something like this or just keep paying for SaaS?</p> <p><a href="https://t.me/chillhousetech/665?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> They’ve already rolled out the next version, and I’m sure this open source was… https://darkolorin.com/posts/664.html https://darkolorin.com/posts/664.html Fri, 29 Mar 2024 01:08:43 GMT <p>They’ve already rolled out the next version, and I’m sure this open source was fucking unnecessary. They tossed out a version that’s already totally not relevant. <br> <a href="https://x.ai/blog/grok-1.5?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.ai/blog/grok-1.5</a></p> <p><a href="https://t.me/chillhousetech/664?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, they’ve learned to integrate emotions into voice assistants pretty well now https://darkolorin.com/posts/663.html https://darkolorin.com/posts/663.html <p>So, they’ve learned to integrate emotions into voice assistants pretty well now <br> <a href="https://demo.hume.ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://demo.hume.ai/</a></p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/663?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> What are they thinking about AI over at sequoia Here. https://darkolorin.com/posts/659.html https://darkolorin.com/posts/659.html Thu, 28 Mar 2024 00:59:34 GMT <img src="https://cdn4.telesco.pe/file/WlUstm-XZ7CntYh3Dnm3ZVCngZqK4KShvWiLHXdJzoHXdw-JdkMaTFDf2yqCJKRHWl7GOkpXQ46gg3rFp-VQkJwWotzAmh7ACMpk2Mk3jJssdDpBsQYyhFhyUbkhQJ8tpXauDw0CUUzqbfcaNdP50RZ8zGCzSTWhuGCqkLwDNTWE3bcIMz7TVlNfaZlcyy4yjlF6BUe388HBGjKDUZc-RrlKDhiCMC_aaQwXdwYZhW9zXE4MB_7OAEGQS3im3vwhWH8A_rHN1j7tVm4EMLQ1DrdKCakqom-_j6vIZHAGb2nSdDNnJ5GtUNEUN05wVZ30GOwXOZLldTlBgD48nZxsuw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/OyBh6PdL5_Gz00NFIYc-Kelu8s4pQU7dNrE7Ro5jRneWv6b_9G3XDGQhYbscbjd77ijnwWxHI7Pt40eXL_pNfQlvVqWA5ILf6Dumq4upJUNWMmtMrGAAEJvhLQZ29ButNrldl2AQXsDI7WMH6U40ey9FIynaDDOgQTIiilG4vJ84xGZ9DgEwsK6Jzk0dNQYQiatV9AhPcT7v_giD06ugMr4gSbBiweM-0d8leH3CJjuRgo0vKNYYi40u346tlGsbfDNqsj0xaBbJfa6yLmLPmFBshXPMCmVwilo_9m2JRAe3hvUzRGgXiLN25OgJYHVOZMO94kxLbkp0yX_saGPkwA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/JLh4ULE7xHXO74ERWBgfnZq2z1U-Fy2iX3diXp9-Dd0W9HML2_J7JGAFXLu3g8J4210HIlXiA6pmGSmJcJ3jjf0Xp5ejSBie7h7xdoakXaUdjPii8FPDO8k6gFvh7RAxi0Ulxi2ASk266iFlFJdHNMQF8okCRblLzyN1peRJDcpqXlz5LrgRT2xHDnFEn8YmtyKZ79jRTouAv0-nDbTKCjta5bVNvvGtBuD-M8wNzowF9H2VjTB7SQqr_oLkcV7qDzff9rmHouzOPXqvL0mOCkioRXRKX6yuR3tQmo6lF5GCxZN3KYa-IKqmUJswL93KGJnu6ZzhoM8YARIpT5KVcQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/mExC0bLUpEnkGco_290U7w210NkBEWEJ4zrj7_K5bmtBA9W9ynw0UA4ctsX2sR3hiOdmpXTeX6ippX0PzjRmEdESfBp5ghfEXnNkbRKtikNm0hhSh1PT6z803hTBz6FL3fdy3mnI3neu0zKFV5p5Vb-FAiqANAt5nLG7ycDzGp3Q-TwfKHP-wqXxK_3rW6zZxKq59ZgDd3r4k9hJobhfx5J74fMIZyfsbG0FKQAQYuh2I1KvpbsvhvM2owzdiJoLPYlP6eXK2pL3i-9nTCl5sGk51m6rRQJ-btzYr9pCMMqco1JZnJwVXR_ES-aIvTgV82LP0aiYqp7dv0cC9e0ACA.jpg" alt="" loading="lazy"> <p>What are they thinking about AI over at sequoia<br> Here.<br> A playlist (<a href="https://youtube.com/playlist?list=PLOhHNjZItNnOoPxOF3dmq30UxYqFuxXKn&amp;si=X51k22qGuIcLc9uf&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtube.com/playlist?list=PLOhHNjZItNnOoPxOF3dmq30UxYqFuxXKn&amp;si=X51k22qGuIcLc9uf</a>) of talks. Overall they believe this is a new turn for the birth of super-huge companies.</p> <p><a href="https://t.me/chillhousetech/659?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Found this kind of study on Twitter. https://darkolorin.com/posts/658.html https://darkolorin.com/posts/658.html Wed, 27 Mar 2024 14:35:09 GMT <img src="https://cdn4.telesco.pe/file/LJj0NA9EXHWMtH3jHsivrIDOHnxhyXDFTFTLew67gnkoTInHPlYO1U5bFL_TEMPHQcC0zxXj2h6_9-_1w0h5aup6WT8xSbBWA8JlSB-1q33-fUzb0ZE1AIVKRKbIMuIm0t8MYrIK70vRSA5LE5iL2I2rWRPi1Gktl3dM3V637pFFTWZOjgJHEAjAprEAl0LFz8qnZiTpORAAukCzzKx2LPaAtBIdizIrbNl2VQ4kSd2O1UFEAzF4PlEBBjFZXRyP6Vgio6AAhj46B0GIRXjq7GSjzthrRHKl39ScrZvVSmdh7uvgqaahxZLUJ6R2FWf_Njp6Qffx4TdYsjh6wpfU5Q.jpg" alt="" loading="lazy"> <p>Found this kind of study on Twitter. But it doesn’t say whether they count project managers as PMs, I think overall yeah.<br> Talking about headcount, obviously.</p> <p><a href="https://t.me/chillhousetech/658?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Alright. https://darkolorin.com/posts/657.html https://darkolorin.com/posts/657.html <p>Alright. One post about lifestyle. <br> I’m in Seoul right now. Second time, but the first was so business that I didn’t have time for a damn thing. A few impressions:<br> - super clean, just everywhere, in any neighborhood, in any back alley.<br> - neat, tasteful, everything is kinda, damn, beautifully done. Any place, not even talking about coffee shops.<br> - and yeah coffee shops and coffee, hands down the best in the world, quality, service, vibe and taste, I haven’t seen better<br> - prices are like minus 25–30% versus London or, like, Europe</p> <p>Feels like it’s comfortable to live here. Don’t know. Just a nice feeling. Everything’s made for people. Wi‑Fi is basically everywhere :) very little weird stuff, and a lot is generally understandable, people are all neatly dressed and look nice. Which I think matters for the overall impression. Video from a regular coffee shop (there are hundreds like this here).</p> <p>Subway fare is 1500 won, about a buck with one transfer, there are zones but the difference isn’t that big.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/657?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Marissa Mayer just won’t calm down and keeps launching various startups. https://darkolorin.com/posts/656.html https://darkolorin.com/posts/656.html Wed, 27 Mar 2024 01:05:17 GMT <img src="https://cdn4.telesco.pe/file/KmL7PA70pmLDfsIcLbdsuQZYE1VSu4_oTIt54MxvBJkeNwJkPka5xDuQFDXgdZiShjUrc0f7EzkG4W3Lfz0kKKexRSTxkULuqMvffP0aJ-9BC_qdJorQR5LNKHtxGH0KP1k4z28MLvtvVzVEZARvGwPt2_5X9TWm4LKi0dn0bvtvHUq1FGOxjjAJntcQBEVL9Ih09Fzwb6_n98sfo5Fg54-jGANGArdYsn4s5yl4kG8SOnlHTJm09RPrJktTKBXFiplcZEI4SCl6AsrqxZUG9RwCXOFOm6WQOtdkK66xil4hMybU7e0PdynWKX4zY6incPy5_qFVwh29OhWAQf0uoA.jpg" alt="" loading="lazy"> <p>Marissa Mayer (<a href="https://en.wikipedia.org/wiki/Marissa_Mayer?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://en.wikipedia.org/wiki/Marissa_Mayer</a>) just won’t calm down and keeps launching various startups (<a href="https://x.com/marissamayer/status/1772668635217010853?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/marissamayer/status/1772668635217010853?s=46</a>). This time, in the age of AI, we’re watching photo albums you can share with friends (wow). And a very cool interface (gen z will get all nostalgic, of course). <br> But the very fact that people like that keep doing startups is, of course, interesting. Given that behind them there’s a pretty successful corporate career and a lot of money.</p> <p><a href="https://t.me/chillhousetech/656?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Put this together for you специально from DocSend. https://darkolorin.com/posts/655.html https://darkolorin.com/posts/655.html Mon, 25 Mar 2024 02:36:25 GMT <p>Put this together for you специально from DocSend. Sequoia’s memo on the decision to invest in YouTube. One of their best investments, both by speed of payback and by how big that payback was.</p> <p>Sure, the memo got disclosed during the Viacom vs YouTube proceedings. But which is obviously a plus for us — it’s really interesting to read what factors top partners at funds were weighing back in the 2000s already.</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/655?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And a bit of enterprise AI from a16z P.S. https://darkolorin.com/posts/654.html https://darkolorin.com/posts/654.html Sun, 24 Mar 2024 22:22:22 GMT <p>And a bit of enterprise AI from a16z <br> <a href="https://a16z.com/generative-ai-enterprise-2024/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://a16z.com/generative-ai-enterprise-2024/</a> <br> P.S. Took off to South Korea, I’ll be posting a bit less this week. Not gonna be cranking out lifestyle content, right.</p> <p><a href="https://t.me/chillhousetech/654?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Ahem ahem Microsoft CEO Satya Nadella to board members: “If OpenAI disappeared… https://darkolorin.com/posts/653.html https://darkolorin.com/posts/653.html Thu, 21 Mar 2024 19:26:19 GMT <img src="https://cdn4.telesco.pe/file/OfCO5ZfquJivnEbrCDRqdXKVztp1-XRzBDgsR0oE-2RSS_18MDmZ_bqsdsQZ-rC5LB8IUarQHrulzX5d_mitY2nQ0KlxTI-hINWYc6xknlis6dVYy_eKztLcoT9ci2mks_33V3LY54xqOuj4Ws-iQVfZSOEI_3HVMrsbMxe0XMoJHsN0Jl5KfVi9_g6YrqEVLcg8-y92gc_aJ9KW1BxhMqizHdtNoKhQPZZqQFbmA7vir5wjV7kpKE36NZkHuvCPcFACiz_gNkQWxiW-5i9m8c3rpNLPCMMugN0hZFUfOb-V9MES8fb7wmBVXmB7yxrYd2eY0A3tDbcyHDeW-fgI5Q.jpg" alt="" loading="lazy"> <p>Ahem ahem <br> Microsoft CEO Satya Nadella to board members: <br> “If OpenAI disappeared tomorrow, we have all the IP rights and all the capability. We have the people, we have the compute, we have the data, we have everything. We are below them, above them, around them.”</p> <p><a href="https://t.me/chillhousetech/653?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> They dropped the monthly retention benches from a16z — consumer apps Based on… https://darkolorin.com/posts/652.html https://darkolorin.com/posts/652.html Thu, 21 Mar 2024 00:46:36 GMT <img src="https://cdn4.telesco.pe/file/VmLZz6QJ3mm_64M9ZEBS3Nxt3nUF9pdV4vPyfwDvMDg6WgdPz6ycoM_fYt7--8nhuie_FjzLbczLoPya7L0J2A7VQryWdoyxw-uytkdLQR58aj-INOw_xrbAAsmmJHZZYKTLIHqyAITI9ktM_Vb0tdXQ3Zq8-NQ8MpU1zIDs0ZijsL87U4TCvCI50gK0kvd9saXbCAIkdmx6Vlpg6dG4uuxWCZlGUzfzx4UY7JH8Ui4DbSVzaggWSf6umFaMfn2T3CsiFPv5sAkOkj3oyDb7P5BsH_t71BlQ_scGtxWieXmJRYuw6Igp3GY_2qbIhO1Nvhc2rWgEiz7R9I8mYmorZA.jpg" alt="" loading="lazy"> <p>They dropped the monthly retention benches from a16z — consumer apps <br> Based on how it shook out in the comments — these are activated cohorts (aka paying users etc, the ones who got a cat and so on)</p> <p><a href="https://t.me/chillhousetech/652?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Not bad https://darkolorin.com/posts/651.html https://darkolorin.com/posts/651.html Thu, 21 Mar 2024 00:43:21 GMT <p>Not bad <br> <a href="https://x.com/elonmusk/status/1770565942168420750?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/elonmusk/status/1770565942168420750?s=46</a></p> <p><a href="https://t.me/chillhousetech/651?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A good list of ai repos on GitHub And also an analysis of all this stuff https://darkolorin.com/posts/650.html https://darkolorin.com/posts/650.html Wed, 20 Mar 2024 23:18:09 GMT <img src="https://cdn4.telesco.pe/file/ly5E9PYSGjmRMGwoCProB7WkRYpoX8jvRIwsd1qEjqIUiCoOxVfRADSJQtXt5DrhFacAhkJE9IHATCUmrezIMuk1DTD8V7k9OO4bjA2PybqJmd8xqAa9UDhfuITbuABsPLecYZXr5S6k9oVMziVBmOViKSFfX1YedaqEdoAEF2OR5anTunIaWz3NLZLS3owlFW4rB8IIaN4pc8GED6oTP8_Ymva1fkWbbWiY8lxVIx6oHVjds0Y4FUHqN6Ni_f1ubyXe9ZRF2rcG010Heuf3cAys-hKvjT4s4Jke5mlV9xF4SS5zyYNoOAwnzzKpGK6HF_cUIduTpCi27k9uX4GXqw.jpg" alt="" loading="lazy"> <p>A good<br> list (<a href="https://huyenchip.com/llama-police.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://huyenchip.com/llama-police.html</a>)<br> of ai repos on GitHub<br> And also<br> an analysis (<a href="https://huyenchip.com/2024/03/14/ai-oss.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://huyenchip.com/2024/03/14/ai-oss.html</a>)<br> of all this stuff</p> <p><a href="https://t.me/chillhousetech/650?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I love Feynman’s essays. https://darkolorin.com/posts/649.html https://darkolorin.com/posts/649.html Wed, 20 Mar 2024 20:21:17 GMT <p>I love Feynman’s essays. Still, science gives you irreplaceable knowledge and experience. Let’s not forget that. <br> The first principle is that you must not fool yourself. <br> And don’t forget to repeat the experiments :) <br> <a href="https://calteches.library.caltech.edu/51/2/CargoCult.htm?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://calteches.library.caltech.edu/51/2/CargoCult.htm</a></p> <p><a href="https://t.me/chillhousetech/649?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I remember Facebook once bragging about their philosophy and culture, that… https://darkolorin.com/posts/648.html https://darkolorin.com/posts/648.html Tue, 19 Mar 2024 22:34:08 GMT <img src="https://cdn4.telesco.pe/file/pN_nGWuKo92sKjsxO2e7Ktqb1wrLofjM0W1q-cXjWp2vDDNJM_DXZBhGARZeNNggSr_lwejB_sCQIlma5S34pnp4CIPmDperLcChCiq9lAePAxWcM3Z2KtMCXf5WHRpL8-FtDZaWuiIHozJat74yL5sVbQjoJvohsiZx2TxKMP6MJP0otLuiHnzcpOaQnpkV8DxjsjxM2Dxf_yARNdEyRvNRQZtrpwZ4EOykd6pdlwTiVyNT4BME1K7SG6XA3uw2bQIgm1k_Gim9TlsP6YUGqP-0TiSjdXowtU5S-vrkrFezTLP-vxySti2W3B0BGcs4LNO-5WedZ_9pPJACRX1lVw.jpg" alt="" loading="lazy"> <p>I remember Facebook once bragging about their philosophy and culture, that they don’t have users, they have people. Yeah right. Yeah right. <br> <a href="https://about.meta.com/uk/media-gallery/executives/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://about.meta.com/uk/media-gallery/executives/</a> <br> Probably after the rebrand to meta.</p> <p><a href="https://t.me/chillhousetech/648?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Great LinkedIn and a career. https://darkolorin.com/posts/647.html https://darkolorin.com/posts/647.html Tue, 19 Mar 2024 20:10:36 GMT <img src="https://cdn4.telesco.pe/file/dLYI8KMHrTHv1wPShvGFmEpxSX2yeIWyhSH_ZXpl5GZ_WNzreROVCvW8fQWdQIbMsyMLy_ZO9DqVkF1qykMTRvT_0mvHK4Xv13H7uyB1d4yvKL8O1vW8U0PCUgl7jzaxe6ykV6HPuM8vcXvY4YHMtJe0LJSN8IUl_PirdEA9igPL8IYGZiO-5FzT9OrqL-XbwNLpWlCMqD2ERS8Yaz64NIQGNaIWPfnRMR86RPJrFfZTDAfTQMPKsafrapI92WQkHUnG5mlpSjoLBkuRqz5nxB887LvVFrJWjCvgul3E2sRNLTKzixXQ88V6J47QlpS834_kPjiu00aTvqhX6TXQyQ.jpg" alt="" loading="lazy"> <p>Great LinkedIn and a career. At Google, a new chapter of search.</p> <p><a href="https://t.me/chillhousetech/647?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Man, Sam Altman’s got such a ridiculously good position. https://darkolorin.com/posts/646.html https://darkolorin.com/posts/646.html Tue, 19 Mar 2024 00:56:15 GMT <p>Man, Sam Altman’s got such a ridiculously good position. Just think about it:</p> <ul> <li>you ship super-heavy closed models that need unreal resources (compute)</li> <li>you bend everyone into accepting that training this stuff takes a shitload of servers</li> <li>you go raise money for your magic</li> <li>and then you go “boys, the future money of the world is compute”</li> </ul> <p>Genius. As a seller of startups and technologies, I’d say he’s absolutely top of his era.</p> <p>Click (<a href="https://x.com/aisafetymemes/status/1769600345171481073?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/aisafetymemes/status/1769600345171481073?s=46</a>)</p> <p><a href="https://t.me/chillhousetech/646?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Index Ventures put out a really big piece on how to scale a company in the… https://darkolorin.com/posts/645.html https://darkolorin.com/posts/645.html Mon, 18 Mar 2024 14:47:49 GMT <p>Index Ventures put out a really big piece on how to scale a company in the growth stage. Highly recommend digging into it. <br> <a href="https://www.indexventures.com/index-press/scaling-through-chaos/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.indexventures.com/index-press/scaling-through-chaos/</a></p> <p><a href="https://t.me/chillhousetech/645?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And here’s the release https://darkolorin.com/posts/644.html https://darkolorin.com/posts/644.html Sun, 17 Mar 2024 19:38:00 GMT <p>And here’s the release <br> <a href="https://x.ai/blog/grok-os?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.ai/blog/grok-os</a></p> <p><a href="https://t.me/chillhousetech/644?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> That very pitch deck I actually like. https://darkolorin.com/posts/643.html https://darkolorin.com/posts/643.html Fri, 15 Mar 2024 13:42:33 GMT <p>That very pitch deck I actually like. Everything is crystal clear, all the scenarios are laid out, the tech enablement factor, the market, and it’s insanely easy to read and even without details to picture the service. Progress and plans are spelled out. <br> The original Uber deck. UberCab.</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/643?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Very lucid and clear from Hugo Barra, and I guess I’m of exactly the same… https://darkolorin.com/posts/642.html https://darkolorin.com/posts/642.html Fri, 15 Mar 2024 13:28:29 GMT <p>Very lucid and clear from Hugo Barra, and I guess I’m of exactly the same opinion about the Apple Vision Pro <br> <a href="https://hugo.blog/2024/03/11/vision-pro/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://hugo.blog/2024/03/11/vision-pro/</a></p> <p><a href="https://t.me/chillhousetech/642?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The US House of Representatives backed a bill to force TikTok to be sold or… https://darkolorin.com/posts/641.html https://darkolorin.com/posts/641.html Wed, 13 Mar 2024 17:10:20 GMT <p>The US House of Representatives backed a bill to force TikTok to be sold or banned, the document was sent to a Senate vote — <br> Reuters (<a href="https://www.reuters.com/technology/us-house-vote-force-bytedance-divest-tiktok-or-face-ban-2024-03-13/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reuters.com/technology/us-house-vote-force-bytedance-divest-tiktok-or-face-ban-2024-03-13/</a>) <br> So, what’ve we got on regulations? <br> 🙂</p> <p><a href="https://t.me/chillhousetech/641?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A bit more of the gen AI Landscape from a16z https://darkolorin.com/posts/639.html https://darkolorin.com/posts/639.html Wed, 13 Mar 2024 16:45:27 GMT <img src="https://cdn4.telesco.pe/file/MzELNjE8FggcsGlvM4TjxMzAOVPII1oiEDAAjcv2FnIbPnIorYLeqhdFW48Dzj888hm8Gvw1OreZVKyw0_E8Ypk2bxkv0XpX30gE0PTHqhenu8-iqAT-QcTikL3JDr7RvMFOi2Vb7IVUQ1KHJ5zwkoBpW-LbwqQCY6zkAUec7ncObN0UlVBZx6ad3Y69shWFKJy0rHqZO8lvot3f0Y_7fC0PdmfqSM3p3QfB-eMDQdVB6H_-trH4JbUmNvDkRzFWiVv3MSUszNqQKmecz34ISW_qDgZOTUBpVNYriPs4CZWluov8K79B1zNK5pLeBomOVpOKRNwrXgW_mUJlTnAaxg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/G_5iByinEAq_Hm8sES_lggiIclSBi7PzIXCOpTEMLv9i_ufGa3UBZP62xnfLpghzdTfEj0K1r3JXH50F90ldWjtA4uE8ImFzfnI7KePjrHVgp2OrgI6l1ym-9uoyYfMqtgNPn-jOijh7chyuas_gouvlYAMs-qV-ZyBpZIYsGhcxIvIH-L1V08Tszd5R3b3IEn-N0GPPcW7dEQ_PU86VzTxqEBVtgagg8KDVwq98tu3GyZlvuWoTwcB6H0T8zeopSAzNFnBA4ZSqncP60XTOGpOgWx4PZEM_OberNTnDKd5sV4H4VukFCNunjEb7KofLCzGG5x_C9JiL7YkXyD73_w.jpg" alt="" loading="lazy"> <p>A bit more of the gen AI Landscape from a16z</p> <p><a href="https://t.me/chillhousetech/639?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Automation. https://darkolorin.com/posts/638.html https://darkolorin.com/posts/638.html Wed, 13 Mar 2024 14:28:30 GMT <p>Automation. <br> VC-backed AI employee startups are a trend. <br> Here are some that raised in 2024 + total funding: <br> Software Engineer - Cognition ($21M+) <br> Software Engineer - Magic ($145M+) <br> Product Manager - Version Lens ($1.6M) <br> Data Scientist - TextQL ($4.1M) <br> Data Analyst - Fluent ($7.5M) <br> Chief of Staff - Mindy ($6M) <br> Universal Employee - Ema ($25M) <br> Financial Analyst - Finpilot ($4.5M) <br> Financial Analyst - Rogo ($7.5M) <br> Compliance - Norm Ai ($11.1M) <br> Receptionist - Arini ($0.5M) <br> Loan Officer - Casca ($3.9M) <br> Hotel Concierge - Runnr ($1.2M) <br> Designer - SevnAI ($0.5M) <br> Customer Support - Sierra ($110M) <br> Customer Support - Rasa ($70M)</p> <p><a href="https://t.me/chillhousetech/638?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Recipe for a “modest” seed round: - AI - more researchers from fancier… https://darkolorin.com/posts/637.html https://darkolorin.com/posts/637.html Tue, 12 Mar 2024 21:32:19 GMT <p>Recipe for a “modest” seed round:<br> - AI<br> - more researchers from fancier universities<br> - the main thing is to do a foundation model so you can justify the investment size: “we need the biggest GPU clusters”<br> - make a demo on the training data so every investor loses their mind<br> <a href="https://www.bloomberg.com/news/articles/2024-03-12/physical-intelligence-is-building-ai-for-robots-backed-by-openai?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bloomberg.com/news/articles/2024-03-12/physical-intelligence-is-building-ai-for-robots-backed-by-openai</a></p> <p><a href="https://t.me/chillhousetech/637?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Constantly spams this ad. https://darkolorin.com/posts/636.html https://darkolorin.com/posts/636.html Tue, 12 Mar 2024 20:04:25 GMT <img src="https://cdn4.telesco.pe/file/aMcb0oLybdQIAB1WntTabfrigtgHWftrixxQgstdWdzJP8NsWc5JCuoC9ysq0q9_m6NcOeVFzkI4BprMfk12TSGcLrGxEL_e1rtnqaH-rRzPT97_YX6CaHrxQVVMknZU394EPn024Cjyf8iyGY47Jr9QQrihAyZHAgAN8-GNJVlF5kSnQMJBnwEfOXYUwHOB019LAYLzY4ZPF8HM7S6aid07hWeyb2nxEGN2lB_BSz2T8PTT1uQDgmn2zsqBfiH6KUpANkabwipLp4eIVjY6wizBOQyxJQWEgkJxvnzUuB8hJCxvvVtq4l6uhVis9_xyX2XJMDFgD7zeYyMrTs3uUw.jpg" alt="" loading="lazy"> <p>Constantly spams this ad. I used to suffer from the same bullshit myself for a long time. But seems like in a workshop like that you need people who actually worked at companies, to really understand whether this or that feature came out by accident or there was an actual strategy behind it. <br> And yeah, still didn’t get it — is this about business, product, or what? <br> P.S. For me to write a strategy usually takes at least a month.</p> <p><a href="https://t.me/chillhousetech/636?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So what’s up with subscriptions? https://darkolorin.com/posts/635.html https://darkolorin.com/posts/635.html Tue, 12 Mar 2024 18:27:31 GMT <p>So what’s up with subscriptions?<br> <a href="https://www.revenuecat.com/state-of-subscription-apps-2024/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.revenuecat.com/state-of-subscription-apps-2024/</a><br> P.S. this little fact made me happy: Billing errors account for over 23% of churn on the App Store</p> <p><a href="https://t.me/chillhousetech/635?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, what’s the deal with programmers/engineers there Karpathy described the… https://darkolorin.com/posts/634.html https://darkolorin.com/posts/634.html <p>So, what’s the deal with programmers/engineers there <br> <a href="https://www.cognition-labs.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.cognition-labs.com/</a> <br> Karpathy described the progress of these systems really well and in a very natural way <br> <a href="https://x.com/karpathy/status/1767598414945292695?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/karpathy/status/1767598414945292695?s=46</a></p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/634?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Alright, everything’s on point. https://darkolorin.com/posts/633.html https://darkolorin.com/posts/633.html Mon, 11 Mar 2024 18:49:49 GMT <img src="https://cdn4.telesco.pe/file/FIRBSJtrLVItm8ttayGjjzrnKbLR-JnSbAD-kKeCRTILunmqUJkeGkaXFk6HXv_YSOenyTuOT39B4u2_eZ-eoGrb5nRWgljAgkrK628VN1eRIOKOF2rPvmMritpFzwIIOr1HxJ-zZzrgopcIoAg7zL8t8-rl7MvCgzNyaimU6o6EJGr8yvPPCfyrSEJB_-IBFeQ1-liQ7ILAfXPMu5zPMQlpo4QKcECGADqPLKYS4qKg49-YRpMKEtJT0Kqv-5zt3I17zRl_Ff4vNVx8WpYtKWZTo4qOXlmJLPaa7iIDlEmUVxcV7K3cH9MZQxjXfNvXRn2CZ1whU0-jPPXssXRdNQ.jpg" alt="" loading="lazy"> <p>Alright, everything’s on point.</p> <p><a href="https://t.me/chillhousetech/633?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A little polling. https://darkolorin.com/posts/632.html https://darkolorin.com/posts/632.html Sun, 10 Mar 2024 20:27:08 GMT <p>A little polling. How many LLMs do you use?</p> <ul> <li>1</li> <li>2</li> <li>3–5</li> <li>A fuckton</li> <li>None</li> </ul> <p><a href="https://t.me/chillhousetech/632?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> You’ve gotta make a movie out of this plot, straight like those rumors. https://darkolorin.com/posts/631.html https://darkolorin.com/posts/631.html Thu, 07 Mar 2024 13:27:47 GMT <p>You’ve gotta make a movie out of this plot, straight like those rumors. <br> <a href="https://www.businessinsider.com/google-engineer-ran-secret-startup-china-stealing-ai-tech-doj-2024-3?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.businessinsider.com/google-engineer-ran-secret-startup-china-stealing-ai-tech-doj-2024-3</a></p> <p>The Chinese guys (2) were doing actual industrial espionage. Worked at Google, leaked docs, flew to China with them, and built a startup there off that stuff. And meanwhile a buddy was tapping his badge at the office so it’d look like you were on site. Basically a classic.</p> <p>And what the guys were stealing was schematics for how to make TPUs to run AI. And against the backdrop of Vinod Khosla literally just recently saying AI should be nationalized altogether, the drama started moving fast.</p> <p>Reading materials:<br> Andreessen on X (<a href="https://x.com/pmarca/status/1765637872965853298?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/pmarca/status/1765637872965853298?s=46</a>) <br> The lawsuit (<a href="https://www.justice.gov/opa/media/1341356/dl?inline&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.justice.gov/opa/media/1341356/dl?inline</a>) <br> Vinod Khosla thread (<a href="https://x.com/vkhosla/status/1765424996548333997?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/vkhosla/status/1765424996548333997?s=46</a>)</p> <p><a href="https://t.me/chillhousetech/631?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Everything you need to know about datasets for training LLMs https://darkolorin.com/posts/630.html https://darkolorin.com/posts/630.html Wed, 06 Mar 2024 10:03:22 GMT <img src="https://cdn4.telesco.pe/file/Ithi-SnaAnMkYnp9OApwZC_oIRLEdSiuIs-YWHVKCGZGMGi3FJbTNjDW1nL23zQsWqoJWN8VNx_LY2sY4KF7RwCF_9hAGCbDl15F7-QL7b_psxa1FnOveWoTrS_GaPR6JVR4G0wc74y_s8-IBuijlTVhoQdC_p-kHPQCC7NT0VNga61hDNNUjpZ4JM7ddxkqKcXR_S_mhT9hObju6oxNtiL4qSTGmLrtjkqtggECrMwdc9N0VDSQOcd7syri6lo7xL_CbWp2GADYcu2CZaegcsJ-mdzGRssYSx8vqd5GrxBauEFC4WpQV0NAabXaaIvFZzlh5egbPugqI9n4-ETqjA.jpg" alt="" loading="lazy"> <p>Everything you need to know about datasets for training LLMs</p> <p><a href="https://t.me/chillhousetech/630?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Feels like it’s not that interesting now with LLMs vs the way Jobs used to talk. https://darkolorin.com/posts/629.html https://darkolorin.com/posts/629.html <p>Feels like it’s not that interesting now with LLMs vs the way Jobs used to talk. A virtual character just has very little appeal vs a live one. But the very fact that even back then people were having thoughts like that is fascinating.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/629?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A guide to fundraising. https://darkolorin.com/posts/628.html https://darkolorin.com/posts/628.html Mon, 04 Mar 2024 16:31:22 GMT <p>A guide to fundraising. Free. Okay fine, your time is money anyway. <br> Let’s go. First, some disclaimers: <br> - I don’t know how to raise from angels, funds in MENA, family offices. I can only guess that personal connections there matter even more than everything I’m going to list. <br> - in 2 out of my 3 companies I raised money from top-tier funds (okay fine, in the prism of term sheets it was more like I collected from everyone I could). We’ll be talking about them. The mighty top-tier. <br> The approach I’m describing is not a recipe for success, but it will definitely increase your odds.</p> <p>1. The ability to find a point of overlap with a top fund is super important, even if you have zero chance of raising from them. There’s a whole team there basically doing market making. Meaning their projects will most likely define the landscape of your product. Wouldn’t hurt to learn their concerns and so on, right?</p> <p>2. Prep stage. Probably the thing I’ll pay a million attention to. <br> So what does it consist of: <br> - Idea. I’m not judging the idea “as such,” but the level of how worked-through it is is your first output. “The idea doesn’t matter” is bullshit imo. It takes me about half a year to form the first version of the vision in my head. What matters: <br> - understanding the tech or the enablement factor. Because you rarely build the actual tech—well if you do, I envy you. It’s expensive. So it’s important that the market believes in this tech (both customers and investors). You can’t just whip something up fast; it’s painstaking and often technical work to figure out what exactly and which problems you’ll solve. You don’t know how attention works and you’re doing AI — it’s gonna be rough. <br> - testing it through your network (I usually struggle here), you go to all the meetups/chats/etc and try to understand what bothers people and why. You look for feedback. You change your beliefs almost every evening. Like gradient descent. <br> - team. They invest in people, and it’s important to find who you can work with for the next 10 years and who can bring the idea to life and share the vision. And yeah, any top investor instantly sees whether you actually believe in your idea and product — you won’t bullshit your way through. And basically they’re making a bet: will you pull the next three rounds, will you be able to fire everyone when needed, will you grind 24/7 and drag everyone along. Because it’s really hard. <br> - vision is about what the world will look like after you do what you plan to do. The job of the pitch is to make them believe in the world you’re creating. If they don’t believe in that world — nothing else matters. Honestly I think this is one of the most common reasons for rejection. It’s extremely easy to formulate in someone’s head (like damn, what a load of crap), so a clear explanation of your concept is half the success. Hence all these elevator pitches and other nonsense. It’s a check that you knoooow what you’re talking about.</p> <p>Let’s say you actually listed out all the leaders in your domain and all the investors from top funds who invest in it: <br> - you followed them on Twitter, hit the bell, and you sit there tracking what they’re into, how they see the market and the world in 5 years <br> - any top fund has technical advisors or partners, you studied what they did, what they’re good at, you know how they can help you (knowledge, connections, trends in corporates etc), because the classic American question “how can I help you” is exactly about this, money needs to be multiplied by help from partners. Then the probability they’ll work with you is higher. Even 1% is already a solid uplift.</p> <p>Next: trigger-based communication. You see a post/opinion/release that’s about your landscape. You write a substantial comment/email. Context works. Half of my intros are literally just a cold email at the right moment. Should I write about how to do those emails? FYI ChatGPT suggests decent ones if you give it the topic and the idea. And yeah, this cycle goes on for a month or two. Don’t think the trigger will show up immediately. Patience and work.</p> <p>That’s probably enough for today. And notice — not a word about your current results. Those later. <br> Happy to take your questions. And don’t turn off your camera on calls plz :)</p> <p>P.S. <br> I’ll write separately about this great tactic of sending 1000 emails <br> .</p> <p><a href="https://t.me/chillhousetech/628?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Thanks to everyone who voted. https://darkolorin.com/posts/627.html https://darkolorin.com/posts/627.html Mon, 04 Mar 2024 12:21:22 GMT <p>Thanks to everyone who voted. I’ll throw in my two cents.</p> <p>There are a few aspects I’ve run into, and I’ll just list my take on them:<br> - founders &amp; founding team are different things to me: the first are the exact people the question is about, the second are those who were just around at the beginning<br> - founders, besides carrying beneficial responsibility like shared risks, are the engines of the company being created, and my brain just can’t fit how the main people can own 1% at the moment the company is created<br> - if we’re talking about venture companies, what matters is how investors look at it, and there’s a consolidated opinion here — if you’ve got more than 10% it means you have to be taken into account, seriously (usually in my practice it’s already 5–10%); sitting on the sidelines and not deciding anything with 10%+ won’t fly anymore<br> - a position is possible where that status is agreed internally between shareholders and basically the person uses it (like, you’ve got 1% but you asked if you can be called that — they told you ok); investors’ attitude to this doesn’t change: if you’ve got less than 10% you’re unlikely to be able to later, when raising money, confidently say that (it’s often checked as part of the background check in PitchBook and they take references on you)</p> <p>In short, I consider it responsibility, not status. If someone assigns themselves that status — ask what the person was responsible for and what they did for the company, and then preferably verify it.</p> <p>And as YC writes:<br> If you don’t value your co-founders, neither will anyone else.</p> <p><a href="https://t.me/chillhousetech/627?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The trend is picking up steam. https://darkolorin.com/posts/626.html https://darkolorin.com/posts/626.html Mon, 04 Mar 2024 02:42:20 GMT <img src="https://cdn4.telesco.pe/file/YoP2MuNAmi30SBAChs-V2H4yrSyzIpcIqeLykxduIeTzj9Kpf5fGf8nEMhYSz1UXhJJd0I1acvH_rQ4jVihQPxNnPEPoYQFezLkffGFDayN8Mcb_rW34PnqtT0gYuI2vCQiRfsOSqpvKDa3mRIq1RPSkdgfKBrKcSRdjgCe_Iu5oAB9viDExDBKxmVr3vUKy04sznyrgxR0SW_-wxKgPZrnm1T2Gfb2XSaVoTaEL49c1TCoCH-iK9xT3KQLZfTnLKm-eHYF5yViV39gp_SBrndphtuVeJRlaNHrYzbwM0TwjETrOrXqDwL25MlDyCU32fe3odekryFUdPFX4udTXtQ.jpg" alt="" loading="lazy"> <p>The trend is picking up steam. Wanna deploy a model and generate some pictures — go to the ministry. India’s generally hardcore on regulations. But yeah, we’re gonna see the same kind of thing everywhere.</p> <p><a href="https://t.me/chillhousetech/626?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Who, in your opinion, can call themselves a startup co-founder? https://darkolorin.com/posts/625.html https://darkolorin.com/posts/625.html Sun, 03 Mar 2024 03:02:26 GMT <p>Who, in your opinion, can call themselves a startup co-founder?</p> <ul> <li>&gt;25% shares at the moment of founding</li> <li>&gt;10% shares at the moment of founding</li> <li>&gt;5% shares at the moment of founding</li> <li>Anyone who was there at the moment the company was founded</li> <li>Anyone who considers themselves one</li> <li>Statuses aren’t for me</li> </ul> <p><a href="https://t.me/chillhousetech/625?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And my favorite one One of the most underrated controversial ideas Peter Thiel… https://darkolorin.com/posts/624.html https://darkolorin.com/posts/624.html Sat, 02 Mar 2024 00:03:38 GMT <p>And my favorite one <br> One of the most underrated controversial ideas Peter Thiel has ever voiced was that failures don’t teach you anything.</p> <p>Just think about how much mythology there is around justifying failure as “well, I learned a lesson from it.” It’s basically a PMC (professional-managerial class) mantra: “learn from your failures and mistakes,” with all the boilerplate interview questions like “tell me about a time when something didn’t work out for you.”</p> <p>And yet, Thiel argues you get nothing useful out of failures — because a failure is always overdetermined.</p> <p>In other words, if something doesn’t work, it could also have *not worked* for 99 other reasons.</p> <p>Tested on myself and my startups.</p> <p><a href="https://t.me/chillhousetech/624?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I recommend this to everyone who believes in AGI and LLM magic. https://darkolorin.com/posts/623.html https://darkolorin.com/posts/623.html Fri, 01 Mar 2024 23:50:42 GMT <p>I recommend this to everyone who believes in AGI and LLM magic. Must read <br> <a href="https://docs.google.com/presentation/d/1DQQQJM0LwgdyPMB92LdwPqw-DrXkWeBAKbZ_bPPB0iM/edit?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://docs.google.com/presentation/d/1DQQQJM0LwgdyPMB92LdwPqw-DrXkWeBAKbZ_bPPB0iM/edit</a></p> <p><a href="https://t.me/chillhousetech/623?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Of course, everyone’s probably already read that Elon Musk filed a lawsuit… https://darkolorin.com/posts/622.html https://darkolorin.com/posts/622.html Fri, 01 Mar 2024 12:48:34 GMT <p>Of course, everyone’s probably already read that Elon Musk filed a lawsuit against OpenAI and against Sam in particular, like you’re such a traitor and started cashing in on everything. But I suspect that besides the ephemeral chance of winning (or just dragging it out, for example), one of the ideas might be this: often companies that have a lawsuit filed against them can’t sign serious B2B deals, and it takes a lot more time to come to an agreement. Feels like a tactical move to me. правда, what for? Not sure yet.</p> <p>P.S. Also people say like that’s how they’ll find out info about GPT-4 or like 5, because the law firm is small. But I still think it’s about slowing down the number of deals. We’ll see.</p> <p><a href="https://t.me/chillhousetech/622?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Falling a bit short on earnings, gotta bump my consulting rate for a second up… https://darkolorin.com/posts/621.html https://darkolorin.com/posts/621.html Thu, 29 Feb 2024 14:56:32 GMT <img src="https://cdn4.telesco.pe/file/Vo6ILxPmIzT9RGywDpd6mzfVDuSVM_U2nEX2PMz9zxYGrphC-5PrPmMMIaMmXUqddRuNJX7D0q1tRkyupAL3295IWOvPNH97IcQJ_ah8EZkipGa2Ihxx34jgZWVCL2eivQ2hrl_2UPoNMgJYNSuNvJQiaTanN6ah9CK91QrxblUgT09v94Wu-GJ5QxQypv4hrnmfCskkGQJ9NYo_s5jut0ucT-uMjFTWhzKGqqq6ObYpLqASWL4KHwFc_o65b63wDqD7Ov8-0odkLV0b9N5NkR18aIbFlDEIxGS6QlPalCsRlFPBTfVBmZOJ8XtdlC1tObaabrn3R2RbgQM98b4t8A.jpg" alt="" loading="lazy"> <p>Falling a bit short on earnings, gotta bump my consulting rate for a second up real quick</p> <p><a href="https://t.me/chillhousetech/621?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> News on model binarization (well actually -1,0,1). https://darkolorin.com/posts/620.html https://darkolorin.com/posts/620.html Wed, 28 Feb 2024 14:07:11 GMT <p>News on model binarization (well actually -1,0,1). If briefly:<br> - the paper <br> <a href="https://huggingface.co/papers/2402.17764?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://huggingface.co/papers/2402.17764</a><br> - they say the transformer itself gives the same quality (wow)<br> - obviously a whole playground for hardware optimization for these needs (Nvidia is gonna go down by the end of the year)</p> <p>Why am I writing about all this. Understanding LLMs and other generative models and keeping an eye on where everything’s moving will let you better understand use cases and which devices and hardware will win from all this движ. It’s obvious this year will be the year of optimizations and by the end of the year pretty much everything will run on device in one quality or another, which of course opens up all kinds of opportunities. But you can’t forget:<br> - all local models will most likely work only on the latest device models<br> - it still won’t be GPT-4 level quality, but quick evaluation and summaries right inside the keyboard will be there<br> - it’s still bad with facts and personal data (vector search loses way too much information along the way — though there are some tricks there)</p> <p>I’m sure the window of opportunity to jump into this world will close by the end of the year. It’s the perfect time to do something :)</p> <p><a href="https://t.me/chillhousetech/620?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Everyone’s obviously heard about making money off channels. https://darkolorin.com/posts/619.html https://darkolorin.com/posts/619.html Wed, 28 Feb 2024 12:37:49 GMT <p>Everyone’s obviously heard about making money off channels. Well now Telegram is going to share 50% of the revenue it gets. And it’ll pay out through the TON ecosystem, too. Great move — recapitalizing its own crypto. Bravo.</p> <p>But overall it reminds me of Musk’s move back in the former Twitter with payouts based on views. Which were fun the first month, and then suddenly tanked. Telegram’s audience in general isn’t the priciest (not that much usage in the most expensive countries), so it’ll be interesting to watch the payout dynamics. But in the moment it’s a great move and a step forward.</p> <p><a href="https://t.me/chillhousetech/619?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A paper just dropped in passing on generating already controllable environments. https://darkolorin.com/posts/617.html https://darkolorin.com/posts/617.html <img src="https://cdn4.telesco.pe/file/u5UYjB2-FueSJKAPkWv2HkoiXcx5NKCtBbo09pIx2WVOGbvxJfiJa1oqBVO9poiBXokCw67V_Ajn5QVbgVY6cM-2LCzRAKdeh3b7Xao2m3vLQ0QRPA5ihcIezDzCWGLX5Y_veJqFSnrrqkGhywA0epinX7F9Dk0q6ao1YUDrFq8RmRsAIJSBND2CmdteM96gknojd3iviLDB6hPQTL71GrGuwN8MCkvKrB1rVEBrlFVqJ77ilZkkCWJGXf8iQ2RUD8x7lBdxCd3d_l8L1nP8VIT9qP4gVZE7wIK0SLf-eFeXJipgtLhtlUhYHhwppspWbdM3riqR2CqrEAr03Gkstg.jpg" alt="" loading="lazy"> <p>A paper just dropped in passing on generating already controllable environments. You know, those minigames like Mario. <br> Still gotta remind you: this doesn’t mean it’s game over for all game devs. You need a huge dataset that’s basically all the same under the hood, which is exactly what works for platformers.</p> <p>Inside:<br> - LAM Latent Action Model that lets you do dynamics with actions (your joystick)<br> - the now-standard thing: a video tokenizer with a transformer on top<br> End result so far: 1 FPS and it guzzles millions of memory.</p> <p><a href="https://sites.google.com/view/genie-2024/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://sites.google.com/view/genie-2024/</a></p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/617?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Weekend post. https://darkolorin.com/posts/616.html https://darkolorin.com/posts/616.html Sun, 25 Feb 2024 20:22:05 GMT <p>Weekend post. Or my genuine surprise. A huge number of people read about all kinds of productivity lifehacks, checklists, how to manage to do everything and forget nothing, read books and do yoga while running a meeting. Basically all that GTD stuff.</p> <p>In my life:<br> - just plain Google Calendar, even without Calendly and other crap, most meetings we just agree on upfront and they toss an invite (3–5 a day and 2 1:1s on average)<br> - I’ve got 3–4 active inboxes, and not once, not once (even when it was 100 emails a day with content actually worth reading) was there a “pile-up”<br> - same thing, that pile-up has never happened in any messenger either, there are basically two modes — don’t reply, read and don’t reply, set an auto-reply for the next day like “sorry means I’m not up to you guys”<br> - I’ve never had either a personal or a work task list, yeah I keep everything in my head/email/messenger (unreplied stuff is often exactly for that)<br> - I manage to: grab high game ranks (diamond TFT, 3k RIO WoW in the first season though), read 10–50 pages a day (I read diagonally and very fast — useful skill), fuck around and work sometimes 3, and sometimes 10–12 hours a day</p> <p>And yeah I really believe that no methodology will help you become more effective. It’s habits.</p> <p>P.S. This week we’ll bash out some interesting insights.<br> Selling pricey courses on how to reach this kind of zen.</p> <p><a href="https://t.me/chillhousetech/616?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Pretty interesting data. https://darkolorin.com/posts/615.html https://darkolorin.com/posts/615.html Sat, 24 Feb 2024 20:13:40 GMT <img src="https://cdn4.telesco.pe/file/Vjmd7yGOr3ieGXyOOHy0PqHe4Pe0xxITM4t_sos_wGf8Tj1qvk1phrdbyQKy4K4DcLaQvZLJym9TElonW_QTMphdP9Nlz1oYc07oiNnivGxsEqWp2ibQ1wTp5jnM2RP8vxA2hBYIIiezzRbWQ8jpsIx90nEWrQ0BGXU3tVIuAyxctGDwlIZ-oCMLugHXoMZo6Wwv_oL47FsQ3hU-td5usl6zT7m3evB0HgQzf5T9Gxk9YA3DollBPFdFAfvvXCxUeagkBoZbqHglziczGChlapBs6WiEVrF3pyJnOWz8qDYKdMb9seC1OD_6NVM9E-lB_eVrcm4GrMQJGKkzdhR_NQ.jpg" alt="" loading="lazy"> <p>Pretty interesting data. Specifically, the audience overlap between Reddit and other services. Twitch is the leader, and yeah, that tracks — Reddit has a seriously strong gaming crowd.</p> <p><a href="https://t.me/chillhousetech/615?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Most product channels focus on product management processes or people… https://darkolorin.com/posts/614.html https://darkolorin.com/posts/614.html Fri, 23 Feb 2024 21:05:27 GMT <p>Most product channels focus on product management processes or people management. Whatever I read, it’s checklists everywhere, method descriptions, typical communication problems and career paths. This knowledge is already available in pretty large quantities everywhere. Same Reforge. Even easier: go to ChatGPT and ask, describing the situation. Trust me, the answers will be dope.</p> <p>I think it’s not that important. More precisely: there are, as it seems to me, more important things.</p> <p>Endless learning of frameworks helps settle the picture in your head, understand the structure better. But often the problem is in the question: “so what the hell do I do right now, and why?”</p> <p>Strategy. Priorities and, as a consequence, features. The ability to do things on time while keeping focus and to throw out optimizations in favor of 10x growth. For these tasks I consider important:<br> - constant practice in the sector/industry, you won’t win this with theory<br> - network, as a tool to validate your logic<br> - understanding releases from both the side of corporations and startups that are trying to outrun them<br> - macro analysis, because why make yet another wrapper around the ChatGPT API if you don’t understand the fundamental constraints and problems<br> - deep understanding of the underlying technologies at the base of products<br> - analysis of the VC market and preferably top-tier, because the whole team there works on the points above<br> - predict, write down what you think is relevant and watch the market dynamics—did you guess, what reasons did you see to make the forecast<br> - an extreme degree of self-criticism</p> <p>In this small channel I’m trying to develop this in myself and my readers. And yeah, I love open, honest questions and comments.</p> <p><a href="https://t.me/chillhousetech/614?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> YC’s Managing Director gives a good hint on the question of “what if Google… https://darkolorin.com/posts/613.html https://darkolorin.com/posts/613.html Fri, 23 Feb 2024 19:01:04 GMT <p>YC’s Managing Director gives a good hint on the question of “what if Google does the same thing as your product”:</p> <p>When a startup is competing against a large competitor, they aren&#x27;t competing with the *entire* company, they are likely competing with some PM focused on internal politics/career progression. <br> With this framing, it shouldn&#x27;t be surprising to see startups win as often as they do.</p> <p>But he forgets to mention the power of distribution. If you’re competing with a big company, you have to reach enough scale before they ship a release. And that already makes the task a bit more complicated than just competing with some product guy.</p> <p><a href="https://t.me/chillhousetech/613?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Everyone’s probably seen what kind of pictures Gemini from Google generates —… https://darkolorin.com/posts/612.html https://darkolorin.com/posts/612.html Fri, 23 Feb 2024 17:20:33 GMT <p>Everyone’s probably seen what kind of pictures Gemini from Google generates — or can’t generate. I think this is the best take.<br> <a href="https://telegra.ph/The-shining-star-of-corporate-surrealism-02-23?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://telegra.ph/The-shining-star-of-corporate-surrealism-02-23</a></p> <p><a href="https://t.me/chillhousetech/612?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Off-topic: right now I’m reading a curious little book by Hermann Weyl, “On… https://darkolorin.com/posts/611.html https://darkolorin.com/posts/611.html Fri, 23 Feb 2024 03:12:59 GMT <img src="https://cdn4.telesco.pe/file/rPIDqH4jNC2HwDC6d4V2v7sGAotWX9nxYltmOJAh8wcl5qfqwGMUDRzxz1Z0Vl0BbtxBovytbWdQtEIWayXPs1kgWPxDij0bl7_ByADB__Vd_h8B5Kr7oHrE-hQv_JN5zbk233vSknnm1iTqZ9ovTjHBe8INSesL2JM_Wmh7etwljXi4aS9bXymdOpuUPitZVKRqbwjoKYcMFpaJZq4DB7UE7vYdh4AtmvAKuU_QaxwErerxTRHF3Xt1s-nz5q3rUfiFU4RPBjuNohFi3kNtxxZmllFe7PLzLzv2l2cS5u97US0J0HcgxuOoqYVFJDzx-2anB1klA77QjQco1Cf1rQ.jpg" alt="" loading="lazy"> <p>Off-topic: right now I’m reading a curious little book by Hermann Weyl, “On the Philosophy of Mathematics.” Aside from the fact that I deeply respect the work of this scientist, who arguably had a hand in just about everything in math and mathematical physics, I’m generally interested in tracing the thought patterns from the revolution period in physics and mathematics—the early 20th century. <br> I noticed a word in the book that hurts my eyes—итти. Turns out, up until 1956 that’s how they wrote the verb “to go”—идти. Well then.</p> <p><a href="https://t.me/chillhousetech/611?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The power of the community. https://darkolorin.com/posts/610.html https://darkolorin.com/posts/610.html Thu, 22 Feb 2024 21:31:58 GMT <p>The power of the community. Probably the biggest one right now. <br> <a href="https://www.cnbc.com/2024/02/22/reddit-files-to-list-ipo-on-nyse-under-the-ticker-rddt.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.cnbc.com/2024/02/22/reddit-files-to-list-ipo-on-nyse-under-the-ticker-rddt.html</a> <br> Reddit is going IPO. What’s interesting: <br> - yeah yeah users with high karma apparently can take part <br> - when calculating the ad market, Russia and China are straight-up removed in plain text <br> - they’re planning to launch what looks like their own Avito (great move) <br> - they want, as I understand it, to keep ~10% of revenue in deals for licensing their content for whatever, including LLMs &amp; Search. Google, as I understand it, is the first client for $60M a year.</p> <p><a href="https://t.me/chillhousetech/610?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Stable diffusion 3 - like a decent text-in-images generator - they brought a… https://darkolorin.com/posts/609.html https://darkolorin.com/posts/609.html Thu, 22 Feb 2024 17:28:46 GMT <p>Stable diffusion 3 <br> <a href="https://stability.ai/news/stable-diffusion-3?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://stability.ai/news/stable-diffusion-3</a> <br> - like a decent text-in-images generator <br> - they brought a transformer in here too <br> - can’t read the paper yet <br> Waiting to test it and look at the results</p> <p><a href="https://t.me/chillhousetech/609?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A little bit of classics about ideas and their “unimportance.” The good old… https://darkolorin.com/posts/608.html https://darkolorin.com/posts/608.html <p>A little bit of classics about ideas and their “unimportance.” The good old myth that it’s all about execution. <br> <a href="https://cdixon.org/2013/08/04/the-idea-maze?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://cdixon.org/2013/08/04/the-idea-maze</a> <br> And also Jonny Ive’s speech</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/608?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And also those channels of all kinds of super-experts are so damn annoying… https://darkolorin.com/posts/607.html https://darkolorin.com/posts/607.html Tue, 20 Feb 2024 23:01:02 GMT <p>And also those channels of all kinds of super-experts are so damn annoying, where you can’t drop into the comments and criticize the bullshit in the post. True, right? <br> 😏</p> <p><a href="https://t.me/chillhousetech/607?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m watching this whole constellation of posts about hiring. https://darkolorin.com/posts/606.html https://darkolorin.com/posts/606.html Tue, 20 Feb 2024 22:57:26 GMT <img src="https://cdn4.telesco.pe/file/NDT27Mv5uZk9Lb0JE7dBlB-rgL9FyeVtHMPGCSzCUVmYucaoUw-hLw4qhaA-nBxUbZ_aLpeQy-mEXz6lEYlZ6IcEonyOopYhyQATIErXxcKFTMQQrA_Y6B5eCTu_FDiLpphHkyV6Nvhx7p9CeptxpdagQsWHYOe375DWcM1jIBnKNT--Jb505dq7sMGShfhQci8rW0o4hpd1TmY3JpSxYakfKewmhVqZKc_SlGG1aVvvS0u4InhLxTTlPENeBudaeUiYA2SoYTYcbmJxY4fmZEkBn_ks3u4T6m7cUbYTEMpTv0hNUf8f2cwCj5l-EZiRGRdgZFb8ea-uYChmP9ujtA.jpg" alt="" loading="lazy"> <p>I’m watching this whole constellation of posts about hiring. Like, some people have literally “thousands of resumes,” and some can’t find two programmers (the word programmer hurts my ears, for me it’s engineer already).</p> <p>And yeah, it’d be fine if these respected gentlemen understood the reason, but they write and ask how can this be. Well it’s actually pretty simple. We’re talking startups.</p> <p>I’ve never once had problems with hiring. I act according to a<br> secret<br> methodology:<br> 1) I only hire from my network, or via a referral<br> 2) usually people go to work for other people (especially the cooler ones who’ve already done everything, or наоборот the younger ones who haven’t done anything big with anyone yet) and globally they want to learn from the CEO / or like a VP and so on. People came in crowds to work with the emperor (Mikhail Parakhin, ex-CTO of Yandex and current head of Ads &amp; Portal at Microsoft), and of course crowds left too. But the effect is clear.<br> 3) everyone I’ve seen prefers b2c over b2b etc. It’s just boring to build admin panels for the hundredth time and solve “interesting tasks for our clients” and “make this<br> data sync pipeline between enterprises<br> make the world a better place.”<br> 4) the leader’s charisma means a lot<br> 5) HR at a startup level == a curse</p> <p>P.S. The picture is an example of a post about looking for sales folks.</p> <p><a href="https://t.me/chillhousetech/606?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Did everyone subscribe to OpenAI’s TikTok? https://darkolorin.com/posts/605.html https://darkolorin.com/posts/605.html Tue, 20 Feb 2024 21:00:57 GMT <p>Did everyone subscribe to OpenAI’s TikTok? We’ll be watching the “TikTokers are no longer needed” show <br> <a href="https://www.tiktok.com/@openai?_t=8k3A4dvLGmP&amp;_r=1&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.tiktok.com/@openai?_t=8k3A4dvLGmP&amp;_r=1</a></p> <p><a href="https://t.me/chillhousetech/605?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For all AGI and Generative AI fans, a must-read A bit of info. https://darkolorin.com/posts/604.html https://darkolorin.com/posts/604.html Tue, 20 Feb 2024 00:13:24 GMT <p>For all AGI and Generative AI fans, a must-read <br> <a href="https://x.com/fchollet/status/1758896780576739485?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/fchollet/status/1758896780576739485?s=46</a> <br> A bit of info. This is a pretty respected person in the world of Google AI.</p> <p><a href="https://t.me/chillhousetech/604?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For you, my dear B2B growth hackers: a gazillion MRR in just a month. https://darkolorin.com/posts/603.html https://darkolorin.com/posts/603.html Mon, 19 Feb 2024 03:21:57 GMT <p>For you, my dear B2B growth hackers: a gazillion MRR in just a month.</p> <p>If you are a startup founder and have a productized business:<br> 1. Go to ycombinator. com/jobs<br> 2. Find recently funded startups that are hiring for the role related to the productized service you offer<br> 3. Install the Chrome extension “Simplify”<br> 4. Use this extension to fill out the applications<br> 5. Instead of the cover letter, plug in your offer<br> ————<br> 6. Profit?</p> <p><a href="https://t.me/chillhousetech/603?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A minute of research. https://darkolorin.com/posts/602.html https://darkolorin.com/posts/602.html Sat, 17 Feb 2024 12:34:28 GMT <p>A minute of research. Judging by the info on Twitters, here’s the basis of OpenAI Sora’s new model <br> <a href="https://openaccess.thecvf.com/content/ICCV2023/papers/Peebles_Scalable_Diffusion_Models_with_Transformers_ICCV_2023_paper.pdf" rel="noopener">https://openaccess.thecvf.com/content/ICCV2023/papers/Peebles_Scalable_Diffusion_Models_with_Transformers_ICCV_2023_paper.pdf</a></p> <p><a href="https://t.me/chillhousetech/602?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The burning problem of a modern product or founder: “so what the hell do you… https://darkolorin.com/posts/601.html https://darkolorin.com/posts/601.html Sat, 17 Feb 2024 11:50:07 GMT <p>The burning problem of a modern product or founder: “so what the hell do you do if you don’t buy traffic?”</p> <p><a href="https://t.me/chillhousetech/601?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A super important topic in the release write-up (lolkek, not even a “work”) on… https://darkolorin.com/posts/600.html https://darkolorin.com/posts/600.html <p>A super important topic in the release write-up (lolkek, not even a “work”) on Sora. If you keep it short, of course they didn’t say jack shit about how and what was done, as usual. BUT there’s this:</p> <p>Simulating digital worlds. Sora is also able to simulate artificial processes–one example is video games. Sora can simultaneously control the player in Minecraft with a basic policy while also rendering the world and its dynamics in high fidelity. These capabilities can be elicited zero-shot by prompting Sora with captions mentioning “Minecraft.”</p> <p>This finally brings us closer to real VR and full personalization of games.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/600?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> First legit approach to video generation https://darkolorin.com/posts/598.html https://darkolorin.com/posts/598.html <p>First legit approach to video generation <br> <a href="https://openai.com/sora?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/sora</a></p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/598?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And then some more news right after that. https://darkolorin.com/posts/597.html https://darkolorin.com/posts/597.html Wed, 14 Feb 2024 21:39:14 GMT <p>And then some more news right after that. The only sensible move, and the expected one. <br> <a href="https://www.theinformation.com/articles/openai-develops-web-search-product-in-challenge-to-google?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theinformation.com/articles/openai-develops-web-search-product-in-challenge-to-google</a></p> <p><a href="https://t.me/chillhousetech/597?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Transfer window news in the AI world https://darkolorin.com/posts/596.html https://darkolorin.com/posts/596.html Wed, 14 Feb 2024 21:25:02 GMT <img src="https://cdn4.telesco.pe/file/P1-Jpz1MTKUC6vCEajRtwFrjETKMiuW4TuGdw9fI52dT56jjVvKM8yL2r4Pc91HfClu2z_YMfUrEYoAmstO5HyZavlkWeTK0_imsUl-zID2nLOJfsbpdyISbV3xcsB4PcJOn8Sza3KL-JcmX6T3vbdmo2x1Fj-eKQjvJGknWUEjjIQzXqVI7M_zeHM0sC660oiPPUQySswM0ovhkBQLLwSNaiW8NIhNZ6B6GXxya73_1ovO5FmoZ7sfgK15WJkjbwPtY17XfKFFADftZI0xIf79ll2M0v7P67Y1y_CJLibedhJboEvZVrGRBBlVZCfolpejCb8F7Jxb5Z1aCoc95vA.jpg" alt="" loading="lazy"> <p>Transfer window news in the AI world</p> <p><a href="https://t.me/chillhousetech/596?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> YC’s new list of startups they’re interested in reflects the trends extremely… https://darkolorin.com/posts/595.html https://darkolorin.com/posts/595.html Wed, 14 Feb 2024 15:04:35 GMT <p>YC’s new list of startups they’re interested in reflects the trends extremely well <br> <a href="https://www.ycombinator.com/blog/ycs-latest-request-for-startups?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.ycombinator.com/blog/ycs-latest-request-for-startups</a></p> <p><a href="https://t.me/chillhousetech/595?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Amazon’s new text-to-voice generator is really damn cool, check out the… https://darkolorin.com/posts/594.html https://darkolorin.com/posts/594.html Wed, 14 Feb 2024 12:51:48 GMT <p>Amazon’s new text-to-voice generator is really damn cool, check out the samples, especially the emotions! Awesome!<br> <a href="https://amazon-ltts-paper.com?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://amazon-ltts-paper.com</a></p> <p><a href="https://t.me/chillhousetech/594?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well the analytics are kinda meh of course, but it’s interesting there isn’t a… https://darkolorin.com/posts/592.html https://darkolorin.com/posts/592.html <img src="https://cdn4.telesco.pe/file/AUjZM91pyI6O73gwJDkCLxbiW6Fk2o5iY4TXUmCGcqFTONBRfk4zGa_tGYtvApHH118YxnrLXBixcxSfrRWr4sdHaDW-P9rzV3OV9cYSLV6o3LKVs7NnunI0Y_9kGnrDSdJVkLnxec8n25WFbsngj8Pw6P6SPR5A3pIYqfLaVjAYtVKHmyY6K3-soa53HeOi_esvCmAU3KI_t9fQcJ_w0uC0szLvG6Vzp3NCJ8yzZ1bNt8-CtR5yq-BMojB6n9urWTZuDv_-guBA59B-jhFPhXTd1bi57JYmSuGh4BDkHU7bIKClSLuAvSD31gIRXebMbE9fAyaMXySKQZY1RNpkIw.jpg" alt="" loading="lazy"> <p>Well the analytics are kinda meh of course, but it’s interesting there isn’t a word about full-on VR really, and where the hell is the Metaverse. But the real joke will be if Apple made it “worse, more expensive,” and still win the market. Then we’ll be laughing together.</p> <p>But honestly Vision Pro and other headsets, in my opinion, are still very niche devices and for the masses you need glasses. I still think the most important sign of “mass” is the ability to go outside.</p> <p>From my experience using Vision Pro:<br> - it’s the best personal movie theater ever<br> - the eye + finger-gesture control is really cool, seems like an extremely successful solution to me<br> - on the other hand, to use some UI element I have to “walk my gaze” over there, which to some extent contradicts real-life experience — to grab a mug I don’t have to look at it, this desync is noticeable here<br> - I get physically tired, it’s a heavy thing<br> - battery is 2 hours<br> - kinda pricey imo</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/592?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Finally someone came up with a solid analogy for me. https://darkolorin.com/posts/591.html https://darkolorin.com/posts/591.html Tue, 13 Feb 2024 17:25:40 GMT <p>Finally someone came up with a solid analogy for me. Nikita dropped a post on why I don’t do / people don’t do intros. Good concept. Think of intros like a credit limit: with some people you’ve got an infinite one, but with others it’s 2–3 a year. Reminder: infinite, or at least a big enough limit, is only for friends. <br> <a href="https://x.com/nikitabier/status/1757278531199926714?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/nikitabier/status/1757278531199926714?s=46</a></p> <p><a href="https://t.me/chillhousetech/591?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> My name’s Alexey and I made Prisma, Capture (not as successfully, but I did… https://darkolorin.com/posts/590.html https://darkolorin.com/posts/590.html Mon, 12 Feb 2024 19:36:10 GMT <p>My name’s Alexey and I made Prisma, Capture (not as successfully, but I did raise from top-tier funds in the Valley), lf.group (acquired by Xsolla). 2 exits, 3 startups. I’ll be writing all sorts of stuff. Just thoughts, because sometimes it hits. My opinion might change over time…</p> <p><a href="https://t.me/chillhousetech/590?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Overall, this whole “small cozy internet” topic gets a ton of people riled up. https://darkolorin.com/posts/589.html https://darkolorin.com/posts/589.html Mon, 12 Feb 2024 17:29:58 GMT <p>Overall, this whole “small cozy internet” topic gets a ton of people riled up. A lot of folks in VC talk about it nonstop, and I even got handed a million bucks under the slogan make internet smaller. I had that in my Capture project on one of the slides in huge letters.</p> <p>The idea is so not new that it gets reborn every 2–3 years: either into Mastodon, or into various protocols, or into old-new forums, or into some other kind of thing like that.</p> <p>People even put together a collection of posts about how good it used to be <br> <a href="https://projects.kwon.nyc/internet-is-fun/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://projects.kwon.nyc/internet-is-fun/</a></p> <p>In general, like I said in earlier posts on this channel, nostalgia is a very powerful thing and I’m sure something has to pop up in the consumer segment off the back of that emotion. Because over the long run it’s that very product for grown-ups.</p> <p>What do you think: is there room for a full-on product/company focused on the internet experience of the past?</p> <p><a href="https://t.me/chillhousetech/589?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, about what company acquisitions can look like (we’re not even… https://darkolorin.com/posts/588.html https://darkolorin.com/posts/588.html Mon, 12 Feb 2024 11:38:20 GMT <img src="https://cdn4.telesco.pe/file/JtsAeAgb05Dvq53pDwo--6sGfk7EROjUz4wh-gN38eRtiRD15xw1_eA9ycuxKPSmFN_MnD28df60q0y70tU5BXcIUJPDmWQOzULwHQXWtZIi_oLltXo5FvqFIj9GevccUBFo73PO1uvutIBTFwM2kS63BhBf7gY1kFjZ8qpy6G721eMS1iXhtTCJBkvfQZ1dJTu_jOWhoIPspnLT1R0HQbgnKZ_-TboYkoRLvf0MlWni9tMWOxpPWU2RZMMJgmkBDyQPpxfB7GLmz2ZePIkVf6TP0-iY42cxpfOsPBekE4RIK8CxcWYBEU245NmcIQqlCUcNQQ8m7CX_hh8pQtLjvA.jpg" alt="" loading="lazy"> <p>By the way, about what company acquisitions can look like (we’re not even talking about what the point of these companies is in the first place). This is just a CEO and director whim. Yeah yeah, that very style Mark Zuckerberg enshrined, where you decide everything solo. <br> A small case about Tidal &amp; Block. Looks like they decided to carve up the money with a buddy. Tired of working.</p> <p><a href="https://t.me/chillhousetech/588?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> What do you know about ambition. https://darkolorin.com/posts/587.html https://darkolorin.com/posts/587.html Fri, 09 Feb 2024 18:04:59 GMT <p>What do you know about ambition. Big deal, just gotta raise a couple-three trillion bucks. <br> <a href="https://www.wsj.com/tech/ai/sam-altman-seeks-trillions-of-dollars-to-reshape-business-of-chips-and-ai-89ab3db0?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.wsj.com/tech/ai/sam-altman-seeks-trillions-of-dollars-to-reshape-business-of-chips-and-ai-89ab3db0</a></p> <p><a href="https://t.me/chillhousetech/587?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> We love polls, right. https://darkolorin.com/posts/586.html https://darkolorin.com/posts/586.html Fri, 09 Feb 2024 04:13:46 GMT <p>We love polls, right. Me — yeah. How much would you ask for consulting in your professional area per hour in US dollars. Doesn’t matter whether you have experience doing that (then it’s easier for you, you can set, so to speak, the final hourly price) or not.</p> <ul> <li>&lt;100 </li> <li>100-200 </li> <li>300-500 </li> <li>500+ </li> <li>700+ </li> <li>1000+ </li> <li>I’ll pontificate in the comments (or not)</li> </ul> <p><a href="https://t.me/chillhousetech/586?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Startups. https://darkolorin.com/posts/585.html https://darkolorin.com/posts/585.html Thu, 08 Feb 2024 01:34:08 GMT <p>Startups. Or how venture makes bets. If you’re into launching products yourself, you’ve probably noticed that in the US there are big rounds, huge funds, and all that. And on the other side you see other markets where an investor immediately asks for a business model or positive unit economics. Both of these options work. But let’s answer: how does the first one work.</p> <p>Startups from the “knockoff” category grew into something bigger. Nobody (in the US) is interested in investing in yet another SaaS. You’ll ask why? Most top funds don’t want a startup in the classic sense. They want a market maker. Yeah yeah. So that you are the market, or most of it. This is an important point when you read these op-eds from Andrew Chen. And it’s important to understand when you start your project.</p> <p>Marketmaking is a very hard game and a path. It’s not about another ten mil. It’s about IPO and billions. Keep that in mind when you’re thinking why you got rejected by a16z, because they’re not about multiplying their capital, they’re about controlling the market and creating the future. And that’s fine.<br> <a href="https://gamma.app/public/a16z-Consumer-Abundance-Agenda-ieotbnzbxj81biu?mode=doc&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://gamma.app/public/a16z-Consumer-Abundance-Agenda-ieotbnzbxj81biu?mode=doc</a></p> <p><a href="https://t.me/chillhousetech/585?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The future of livestreaming. https://darkolorin.com/posts/584.html https://darkolorin.com/posts/584.html <p>The future of livestreaming. Or how it is in China. A female streamer is promoting products, each one for 3 seconds. <br> P.S. Come on, you gotta admit that’s pretty skillful.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/584?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> It didn’t fit in the channel, so I moved it here, since in twi it’s not always… https://darkolorin.com/posts/583.html https://darkolorin.com/posts/583.html Wed, 07 Feb 2024 21:20:54 GMT <p>It didn’t fit in the channel, so I moved it here, since in twi it’s not always convenient to read these long-ass threads either. But this is, like, a genuinely good guide/explanation about pitching in the US/Bay Area/anywhere. Why it’s like that there and why it’s not like that in other regions. Highly recommend.</p> <p><a href="https://telegra.ph/Andrew-Chen-about-pitching-02-07?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://telegra.ph/Andrew-Chen-about-pitching-02-07</a></p> <p><a href="https://t.me/chillhousetech/583?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And as for me, no interview happened today 🤙 Boring channel https://darkolorin.com/posts/582.html https://darkolorin.com/posts/582.html Tue, 06 Feb 2024 22:59:56 GMT <p>And as for me, no interview happened today <br> 🤙 <br> Boring channel</p> <p><a href="https://t.me/chillhousetech/582?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Up: OpenAI rolled out the same thing too, see comments https://darkolorin.com/posts/581.html https://darkolorin.com/posts/581.html Tue, 06 Feb 2024 15:46:55 GMT <p>Up: OpenAI rolled out the same thing too, see comments</p> <p>Back to the news, I’ve been waiting forever for them to finally start implementing this <br> <a href="https://about.fb.com/news/2024/02/labeling-ai-generated-images-on-facebook-instagram-and-threads/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://about.fb.com/news/2024/02/labeling-ai-generated-images-on-facebook-instagram-and-threads/</a> <br> I think it’ll be across the board, I was talking about this during a pitching in 2016, there’ll be like invisible watermarks and other stuff that helps instantly label content, sure I think I missed the mark on the “watermarks” part, but I got the direction of the thought across right</p> <p><a href="https://t.me/chillhousetech/581?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> From a recent discussion, my comment: https://darkolorin.com/posts/580.html https://darkolorin.com/posts/580.html Tue, 06 Feb 2024 14:03:33 GMT <p>From a recent discussion, my comment:</p> <p>There’s this kind of crap in the industry, probably dating back to the time of the first IPOs. It’s called “genius syndrome.” Or I also call it “iron man perception.”<br> VCs like Thiel and others pulled off a few genuinely very successful deals, and often those were companies run by geniuses—movie-type geniuses, you know, borderline insane and at the same time they actually made it. Mark, for example. And so the word spread so hard that even movies like Fincher’s The Social Network are just crawling with those images. And then it took off. In the Russian-speaking (and not only) segment, this “I know the recipe” mindset got cemented—what you need to do and how to be a badass CEO. You’ve gotta smash everything, talk bullshit, do whatever, because I’m raising money and I’m on top. Only this whole story doesn’t match the fact that all the Valley fairy tales were based on a very short time window, and then people went through drastic changes really fast and became very strong leaders (same Mark). And people like (I’ll omit the name here, but it fits a lot of folks) still live in a dreamworld of fairy tales about Alexander the Great and pretty quotes from TechCrunch.</p> <p><a href="https://t.me/chillhousetech/580?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A real experience https://darkolorin.com/posts/578.html https://darkolorin.com/posts/578.html <p>A real experience</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/578?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, gotta redo the deck, turns out there’s AI NATIVE! https://darkolorin.com/posts/577.html https://darkolorin.com/posts/577.html Fri, 02 Feb 2024 03:56:51 GMT <img src="https://cdn4.telesco.pe/file/pyrxIAfX_nkZN0OClQj8n7b910YOX6U3kWCk20oKQvn1n_jAQzy4ocCg0Q8oPXgYjxbgnERk3TUFIle0IBEy-z4cZLaFclr-UkczkdimjD-T7YM1S9V_0iHSzSh4DflPAB3BI5H_iIgKxpuL57VZtSYpHZrvXdcCJaXsGYJbk1Oix2PSR81p8fHG4XZ-V3XbXwTCflyOEIkLMxxj1EocWUbphVPpzVhIxTSZQI2Bh7-bE979XVCfdJtce4YKx66wKx9vqBwi0Eh7MvKtS0FATPz3-ceyheZKTDvdN89Wp_7JKvwoNpQvl2v8-_UgauDyHyv41Mww16Rv53CaRKOIHw.jpg" alt="" loading="lazy"> <p>So, gotta redo the deck, turns out there’s AI NATIVE! And most importantly, don’t mix it up with AI-ENABLED.<br> 👎</p> <p><a href="https://t.me/chillhousetech/577?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well this is purely for fanatics. https://darkolorin.com/posts/576.html https://darkolorin.com/posts/576.html Fri, 02 Feb 2024 03:51:48 GMT <p>Well this is purely for fanatics. For people who love hunting for success recipes. An analysis of who started a company how and why. And basically when you realize you’re cooler than the people you work with or for, you become a founder.</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/576?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Everything will be on-device, slowly but surely (3–4 years till it’s… https://darkolorin.com/posts/574.html https://darkolorin.com/posts/574.html Fri, 02 Feb 2024 01:29:00 GMT <p>Everything will be on-device, slowly but surely (3–4 years till it’s mainstream), and everything will be stock (libs/models and so on) <br> <a href="https://arxiv.org/pdf/2311.16567.pdf" rel="noopener">https://arxiv.org/pdf/2311.16567.pdf</a></p> <p>Talking about LLM/Diffusion and the rest of that stuff</p> <p><a href="https://t.me/chillhousetech/574?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Found the best app for Vision Pro https://darkolorin.com/posts/573.html https://darkolorin.com/posts/573.html <p>Found the best app for Vision Pro</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/573?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In the minds of the Russian-speaking segment, this notion of a “global market”… https://darkolorin.com/posts/572.html https://darkolorin.com/posts/572.html Wed, 31 Jan 2024 17:38:41 GMT <img src="https://cdn4.telesco.pe/file/EJlW3GM5iWdizgiPo0zCXncjNxemyyQ-zuXXF6LiwKpZyWp7F0DzNXkp9KbjKFZWU0RoH84fUAl5CN6SnZ9UUnRO4CDgq9gJoaWc4L2xMWXFZAo0f30Yebrg4S06739jZ2tTxZ0UU1bUmjP_Ynm5IWxDAaAYvfDQ9OWnKaJlnLGqcSONtAiMiE98RYGTk81aZKNfLHL6hdptkpgWOy_BVN0WQ3Odnl8_jrf5tIOGx0CXPpU8EBZpDK4QL5_EfmI5IxMLQA2c5GEttPOQPOwoJjo65kDtJ4HqUsfF1D7WzK81qlc99Kyou32Cy2_rsYHTVufrifb_VpN3m_Qao2kwsw.jpg" alt="" loading="lazy"> <p>In the minds of the Russian-speaking segment, this notion of a “global market” has really taken root—requests like: I need a global market, not the CIS. (Got curious how many of the askers are successful even in the CIS.)</p> <p>And there are tons and tons of webinars, lectures, podcasts, all that stuff. Everyone talks about cases and whatnot. And often the presentation is in Russian. And you know, it all starts with a tiny little fact: “at least make the content in English,” because culture is like that. How you think is how you do.</p> <p><a href="https://t.me/chillhousetech/572?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And I was wondering why the trains are always running late https://darkolorin.com/posts/571.html https://darkolorin.com/posts/571.html Wed, 31 Jan 2024 03:24:19 GMT <p>And I was wondering why the trains are always running late <br> <a href="https://www.tomshardware.com/software/windows/ms-dos-and-windows-311-still-run-train-dashboards-at-german-railway-company-listed-admin-job-for-30-year-old-operating-system?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.tomshardware.com/software/windows/ms-dos-and-windows-311-still-run-train-dashboards-at-german-railway-company-listed-admin-job-for-30-year-old-operating-system</a></p> <p><a href="https://t.me/chillhousetech/571?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Got curious. https://darkolorin.com/posts/570.html https://darkolorin.com/posts/570.html Tue, 30 Jan 2024 19:21:04 GMT <p>Got curious. LinkedIn keeps sending me how many times my profile got viewed per week. Usually it’s 300–500 times. What about you?</p> <ul> <li>&lt;100</li> <li>100–200</li> <li>300–500</li> <li>500–1000</li> <li>&gt;1000</li> <li>No clue</li> <li>Don’t use LinkedIn</li> </ul> <p><a href="https://t.me/chillhousetech/570?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For anyone who’s interested, a bunch of Vision Pro review vids dropped today… https://darkolorin.com/posts/569.html https://darkolorin.com/posts/569.html Tue, 30 Jan 2024 18:03:03 GMT <p>For anyone who’s interested, a bunch of Vision Pro review vids dropped today, and it feels to me like everyone’s writing super restrained, extremely reluctant to criticize — if this was a device from not-Apple they’d be like: what’s this? no way we can use it now <br> But here we’re seeing something totally different <br> - Apple Vision Pro reviews roundup: stunning potential with big trade-offs <br> - Apple Vision Pro Review: A Mind-Blowing Look at an Unfinished Future <br> IMO, judging by what I’m seeing, it’s like 5–6 years away from being a normal device <br> <a href="https://www.macrumors.com/review/apple-vision-pro/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.macrumors.com/review/apple-vision-pro/</a></p> <p><a href="https://t.me/chillhousetech/569?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So I’m going through the onboarding for the Arc mobile browser and I’m like… https://darkolorin.com/posts/568.html https://darkolorin.com/posts/568.html Mon, 29 Jan 2024 17:32:26 GMT <img src="https://cdn4.telesco.pe/file/MiWI3yqV4Z93JsgxVPCk_PaPQc2MqNXGd5qq_T-mummJQXOFpgGCt-aebSBXpknu4EXM5-Uqyu_9ksOqXSpBSjt49xmOh3OHa6NZ8hs-KNibOKEixPdOee8mqsmPc30Htmv4YWmyRHB6P_9LTSXWCgE6UXR0AY6kt0jmVSo1qITkS8p16J-weXfvOuCW70kKNLrOadYewu3-DZZlP3CiZ__Lx7fMjsumxLlRR7PD845X7pIOTE-DOyaVbl9VsKcNxD5DvVTli0-5A9SRU5bGzGJfkrxHj2tslRvmej-ANZAliCCYHwgYQE7RvGe1OEwmzB8KTyayCNyu-pAdeCBsQw.jpg" alt="" loading="lazy"> <p>So I’m going through the onboarding for the Arc mobile browser <br> <a href="https://apps.apple.com/gb/app/arc-search/id6472513080?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/gb/app/arc-search/id6472513080</a> <br> and I’m like, no need to set yourself as the default browser. And it throws this screen at me one more time. <br> And like, how are you even supposed to set some browser as default right away if you haven’t used it? Or is it assuming I came over from desktop. But there was no login. Some kind of nonsense.</p> <p><a href="https://t.me/chillhousetech/568?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, I’m still continuing to convince myself that what awaits us is the… https://darkolorin.com/posts/567.html https://darkolorin.com/posts/567.html Mon, 29 Jan 2024 16:39:25 GMT <p>By the way, I’m still continuing to convince myself that what awaits us is the de-globalization of the internet <br> <a href="https://fortune.com/2024/01/29/biden-administration-cloud-providers-amazon-microsoft-investigate-report-foreign-clients-developing-ai-apps-platforms/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://fortune.com/2024/01/29/biden-administration-cloud-providers-amazon-microsoft-investigate-report-foreign-clients-developing-ai-apps-platforms/</a></p> <p><a href="https://t.me/chillhousetech/567?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sunday-ish https://darkolorin.com/posts/566.html https://darkolorin.com/posts/566.html Sun, 28 Jan 2024 14:54:41 GMT <img src="https://cdn4.telesco.pe/file/OlAkdLVE3ZhUvl_EOqo3b4E8ZYcN-B1exS22gvPKKbmkpI8TkUv51jNJjsqMkjkIWVtGGlmrKWR7CsE9Q2JTdIrVjg-Rn5xYcudXwFHy6D_m7Lpvr068vUhXCTFlN6wrUfpt9DwLxaI-L_ZVrADMU3z81A0lCJpRFP4DYnvtMT942vRrHUoyzmcSIjGevgL6u-4Wzz0JPfy-x-z-0l62KSEvJq-5BjDHxbtFEd0magIWGuvASBGdmzasDlIw2CzzNeZXkmKSWRP5eYBoo2oUfy0VlrsLQe5ye1ndh-opgxMzWieBU5Xxvrp8TMwJyvv-i01MjgH6aD832_cU2fU9ZQ.jpg" alt="" loading="lazy"> <p>Sunday-ish</p> <p><a href="https://t.me/chillhousetech/566?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Just maybe someone will find it useful — a nicely structured list of LLMs https://darkolorin.com/posts/565.html https://darkolorin.com/posts/565.html Sat, 27 Jan 2024 23:50:04 GMT <p>Just maybe someone will find it useful — a nicely structured list of LLMs <br> <a href="https://llm.extractum.io/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://llm.extractum.io/</a></p> <p><a href="https://t.me/chillhousetech/565?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Nailed it. https://darkolorin.com/posts/564.html https://darkolorin.com/posts/564.html Sat, 27 Jan 2024 17:30:32 GMT <p>Nailed it. <br> <a href="https://x.com/nikitabier/status/1751292923931685076?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/nikitabier/status/1751292923931685076?s=46</a> <br> In case it gets deleted, the text is in the comments.</p> <p><a href="https://t.me/chillhousetech/564?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “Delusion” column. https://darkolorin.com/posts/563.html https://darkolorin.com/posts/563.html Sat, 27 Jan 2024 14:15:50 GMT <img src="https://cdn4.telesco.pe/file/j5nYhcTCXoS60JYtRsOtK_8cQM9SpcpmA-MVefGY719Skk2EytdS2VCutB15fWQClqfUSsYRqpPRn4f8B373aEvWsjJRm4KsYSu9xHuF7bZtjpAP4Rfw2oHZSddK8TL-WPN8SJx74cX79yWXAYVY38dczHt9w1q87zskb8QWFv2WlJgFgGTIW7-JLPdTBzgi1p22HQ2KzkxeAy3DuM4K2psQICdWTT_rBwO3VRJNKgSZJ7LUnFKdwfHCieafEV_XY1j_2EJ2KFy5-aW4XZ7S9gov0XkQfiLcTMKfuE8yoxVY7T5yf4Jgq-tGJErzJaQqhwehPC-EFyfnulSRLGDqxQ.jpg" alt="" loading="lazy"> <p>“Delusion” column. Online there are, like, two religious denominations. The apologists of YC and pretty interviews and stories about what righteous people they’ve got there, and the second camp where you just need to make money, step on heads, and so on.</p> <p>So, here are the main signs of delusion:<br> - the person handing out advice hasn’t founded a single company, has no verified track record, and hasn’t invested anywhere<br> - all the stories and posts are written off internet articles, interviews<br> - another chunk of posts is based on “my successful acquaintances”<br> - they’re constantly dishing out advice / running webinars and so on<br> - in their head there’s a clear recipe for success or how to do it “right” without accounting for circumstances and timing</p> <p>Specifically here they’re postulating that being active on social media is bad. Probably because it eats time, and the assumption is that if you’re in a startup you should be grinding 24/7, which is complete bullshit. Nobody can work for years in the same mode. I figure for a founder the most important skill is the ability to focus at the right moments in time, and the ability to grind when you need to, but doing it nonstop—yeah, unlikely. Elon Musk no-lifes Diablo and it’s fine, even though he’s supposedly working 24/7.</p> <p><a href="https://t.me/chillhousetech/563?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Yeah right they say not everything here is like that. https://darkolorin.com/posts/562.html https://darkolorin.com/posts/562.html Thu, 25 Jan 2024 21:00:12 GMT <img src="https://cdn4.telesco.pe/file/EvkbidVyaWJvItx4DzGoVimebsYOQHaSosmTe6S2Bw4ZoDSm_CizhUX0ai1L0VVFxBY07R9wupAIMbHpY5qM2onTEY5y9h8Tw29wHrpNM0tLRPmJgXRGzK54CXyn04ZbzmQTZY6ScXwIbifVG9TtpiVsrBm-u5ZrfEj7M3-fpqvHZasgHvdCw2Ug2Sh1yqujrHNkW_aSYHczF_IIAIN1jHaF1egm3XxS2njEUv0krcikaACsj-f6VMLLN5B3_0Jh_1m32CAxPclFnwAE5n42wizmmN7cKHcXcEuMFUHUdIEkFCt5PbPCjoVsY8cf8ZK0kCymF7nrKZn9bSpxrMSq2Q.jpg" alt="" loading="lazy"> <p>Yeah right <br> <a href="https://x.com/nikitabier/status/1750592825060921353?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/nikitabier/status/1750592825060921353?s=46</a> <br> they say not everything here is like that. <br> And in general, update added: <br> <a href="https://www.apple.com/newsroom/2024/01/apple-announces-changes-to-ios-safari-and-the-app-store-in-the-european-union/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.apple.com/newsroom/2024/01/apple-announces-changes-to-ios-safari-and-the-app-store-in-the-european-union/</a></p> <p><a href="https://t.me/chillhousetech/562?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> New stuff from Matthew, this time about games, and it seems really informative… https://darkolorin.com/posts/561.html https://darkolorin.com/posts/561.html Thu, 25 Jan 2024 14:56:55 GMT <p>New stuff from Matthew, this time about games, and it seems really informative and easy to follow. I recommend reading the whole thing. <br> It’s about:<br> - what was good in 2023<br> - and what’s bad in the games industry, where the layoffs are coming from and how fast mobile game downloads are dropping (spoiler — fast), and revenues even more so <br> <a href="https://www.matthewball.vc/all/gaming2024?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.matthewball.vc/all/gaming2024</a></p> <p><a href="https://t.me/chillhousetech/561?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Continuation on hiring. https://darkolorin.com/posts/560.html https://darkolorin.com/posts/560.html Wed, 24 Jan 2024 23:20:32 GMT <p>Continuation on hiring. This time my personal experience.</p> <p>Background (I’ve never had a team bigger than 100 people). I personally interviewed either for my own team or helped while doing the main role and making the decision:<br> - hired developers (30+ for sure, maybe more)<br> - designers скорее no than yes, they’re usually full-on partners to me as a product person, but I worked with 4–5<br> - support, marketing, comms — I don’t think more than 10–15 people<br> - условные top managers / including CEOs — like 5–6 people<br> - research or hardcore deeptech — well 10–20 people probably<br> - assorted outstaff/outsource and lawyers — a looooot</p> <p>All in all, I think around a hundred. And about the same number ended up in my orbit, but I didn’t hire them myself.</p> <p>For me the following soft skills matter (writing about hards quickly is probably hard):<br> - ability to manage yourself — in a startup or where I work I can’t stand it when you have to sit with a person and write out every single task in Jira<br> - value priorities — once early in my career I went to a conf and there a VP of design at Facebook said people are usually motivated by one of 4P: profit, people, process, product. I fit with product and more rarely people.<br> - absolute flexibility with schedule — I’m ready to pay more, but I also demand being reachable more often (this doesn’t mean working 12 hours, it means that after 8pm you respond)<br> - willingness to learn and change together with the team and the product — meaning you’re not just тупо insisting on doing it the way you know how just so you don’t have to do it differently</p> <p>My hit rate in hiring is 50/50. Don’t know if that’s normal or not. But in half the cases I definitely screwed up. Almost everyone who worked with me built a pretty damn good career (Google/Amazon and a bunch of future founders). And what’s the takeaway? Dunno.</p> <p><a href="https://t.me/chillhousetech/560?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Longread. https://darkolorin.com/posts/559.html https://darkolorin.com/posts/559.html Tue, 23 Jan 2024 23:25:04 GMT <p>Longread. Or about hiring, from Eric Schmidt. <br> Former Google CEO Eric Schmidt on why you should hire the divas: “Steve Jobs was a diva” <br> “If you read any management textbook, it says ‘don’t hire the divas’ because they’re nothing but a pain in the ass. And by the way, they are. But the people who are the divas—who believe—are the ones who will drive the culture and company to excellence. Steve Jobs was a diva. I worked with Bill Joy who was my colleague for many years—he’s an example of a diva. And I mean this in the most flattering way: they expect a lot, they drive people hard, they’re controversial, and they care passionately. If you find those people, you’re probably going to work for one so be nice to them.” <br> However, Eric doesn’t mean you should tolerate arrogant jerks at your company. <br> In his book *How Google Works*, he uses the terms “divas” and “knaves” to distinguish between who you should tolerate and who you shouldn’t: <br> To be continued under the cut in the comments.</p> <p><a href="https://t.me/chillhousetech/559?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Purely of historical value. https://darkolorin.com/posts/558.html https://darkolorin.com/posts/558.html Tue, 23 Jan 2024 18:15:43 GMT <p>Purely of historical value. Mark Zuckerberg giving a lecture at Harvard in 2005. Kinda more technical, but it’s surprising how few people showed up for his talk. And overall it’s really informative to get a feel for what it was like back then <br> <a href="https://www.youtube.com/watch?v=xFFs9UgOAlE&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=xFFs9UgOAlE</a></p> <p><a href="https://t.me/chillhousetech/558?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile it’s always useful to know the salary ranges at top companies. https://darkolorin.com/posts/557.html https://darkolorin.com/posts/557.html Tue, 23 Jan 2024 15:22:39 GMT <img src="https://cdn4.telesco.pe/file/l_vdj8cgOu3XCZavgQ4QaOJXBjyqNUGXnGyTvXfVuXrBRs3LuxkovgfKMqCyhkycOvtFb-yrzBZnHzv_ZJiCriFz_L1sXF2IzxudBgsJz5jdZXJ2T0jdOkOhZ2bptyCkqSKbdHTIgKfBlZt2IvKUud3-iCRR_-QA4k-rC7YokbY_3NiMVL67GLJwJbcYFwLCTPwW1v-qhsHYQMWwAPbiwRw3aAg0VuN82arhLoAF4m8GFwBrcYutT8Qan-P-6emlidBOkwg7TmZ6oUV8fUHeWCCQdDxCCtBGY2igihIHmsngaoWCXd8kUfJuEeYF7Axr9BY96xibFx7jXC8SVLL4PQ.jpg" alt="" loading="lazy"> <p>Meanwhile it’s always useful to know the salary ranges at top companies. Some data on TikTok in the US just rolled in (pre-tax).</p> <p><a href="https://t.me/chillhousetech/557?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Decided to share a подборка of accounts I’m subscribed to and try to keep up… https://darkolorin.com/posts/556.html https://darkolorin.com/posts/556.html Tue, 23 Jan 2024 15:01:10 GMT <p>Decided to share a подборка of accounts I’m subscribed to and try to keep up with обход ленты, i.e. where the bell is on <br> <a href="https://x.com/joshelman?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/joshelman?s=21</a> <br> Josh Elman, a good friend, spent a lot of time as a product in LinkedIn and FB at the very start of the journey. Saw everything. More useful to look not at the posts, but at the replies. Now runs Discovery and search in the AppStore. <br> <a href="https://x.com/eshear?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/eshear?s=21</a> <br> Emmet, the very one who was CEO of OpenAI for a couple hours. Sometimes really weird thoughts, but insanely interesting. The guy who turned JustinTV from live streaming into Twitch. It was him who believed in games, not Justin Kan and Seibel. <br> <a href="https://x.com/nikitabier?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/nikitabier?s=21</a> <br> the main hype-eater of the product consumer тусовка about fast takeoffs is of course Nikita. Gas in Discord, TBH in Facebook. Spent 6 years getting to the first exit. <br> <a href="https://x.com/jacksoslow?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/jacksoslow?s=21</a> <br> Jack is a partner in the A16Z gaming fund. Very active and replies well. <br> <a href="https://x.com/polynoamial?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/polynoamial?s=21</a> <br> the one who’s on the front line of AI work right now, combines a game theory approach and deep learning. Feels where the industry is going. <br> <a href="https://x.com/karpathy?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/karpathy?s=21</a> <br> and yeah I hope he needs no introduction. One of the very authoritative researchers. Predicted a lot of things long before they became mainstream.</p> <p><a href="https://t.me/chillhousetech/556?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> They’re getting paid just fine. https://darkolorin.com/posts/555.html https://darkolorin.com/posts/555.html Sun, 21 Jan 2024 16:15:08 GMT <p>They’re getting paid just fine. The interesting bit here is that the guys are working specifically on game theory and deep learning. But yeah, a $200M seed round where $195M is basically costs for training the little models for a year. <br> <a href="https://economictimes.indiatimes.com/tech/technology/google-deepmind-scientists-in-talks-to-leave-and-form-ai-startup/articleshow/106997496.cms?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://economictimes.indiatimes.com/tech/technology/google-deepmind-scientists-in-talks-to-leave-and-form-ai-startup/articleshow/106997496.cms</a></p> <p><a href="https://t.me/chillhousetech/555?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> How many hours a day on average do you work effectively (pure focused work… https://darkolorin.com/posts/554.html https://darkolorin.com/posts/554.html Sat, 20 Jan 2024 14:52:06 GMT <p>How many hours a day on average do you work effectively (pure focused work time)?</p> <ul> <li>1–2 hours </li> <li>3–4 hours </li> <li>5–6 hours </li> <li>Around 8 </li> <li>Over 10 </li> <li>I don’t work, I get paid $$$</li> </ul> <p><a href="https://t.me/chillhousetech/554?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Lately I’ve been watching how the big founders of giant consumer social… https://darkolorin.com/posts/553.html https://darkolorin.com/posts/553.html Sat, 20 Jan 2024 14:40:42 GMT <p>Lately I’ve been watching how the big founders of giant consumer social companies started doing their new projects. Whether it’s Kevin with his Artifact, or the Vine guys with some SUP xyz bullshit in the middle of the NFT boom. Briefly. There’s this persistent feeling that consumer social is like…</p> <p><a href="https://t.me/chillhousetech/553?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Real culture in a company, aka the weekend post. https://darkolorin.com/posts/552.html https://darkolorin.com/posts/552.html Sat, 20 Jan 2024 02:56:09 GMT <img src="https://cdn4.telesco.pe/file/usdrjZLmfLsPIKcn6b7c_9SG0of4CJYQI8wnysJBA6pXN5BEfHGGuloW-gbTo3ku8u1dFclUAMqgVqMGP-lb_rKUT-8_O8f1qJ9obe0uFh-BfowmNGordPY8R6ePEXyMSL2hEPr6O7WxB5PTjmhAoW_WFKrEUyhDfDWJ2VclwWESAN6eAZylKCZD2_r6l6MWxfB33O9-CLLNXvrL9XgH5J6tFPe0_F4z7IrARSwgTvXbGYG3pROfg7S2FYuAaGkFsFNCMW_TbQZTXwhaoR8SugbO34xpFPBovEtQtYW2rPMs9lO0d-H4EG7Zts-KXbn7qC4GGbmfAvPh9uRQR1NWmA.jpg" alt="" loading="lazy"> <p>Real culture in a company, aka the weekend post. <br> I don’t believe in:<br> - constant compassion, empathy, support and all that for no reason<br> - some kind of work-life balance, if you dream about it go somewhere where everything is already super awesome and you’re like -10 levels away from the CEO and founders<br> - the idea that tons of meetings and a jam-packed calendar is work<br> - micromanagement or “let’s discuss the plan with literally everyone”<br> - a sweet, respectful tone of communication throughout the whole job</p> <p>If you ran into any of the above, there won’t be progress — neither yours personally nor the business/company as a whole. And pretty often I see this kind of communication style like Elon Musk after buying Twitter with their CEO.</p> <p>Edited: the comments turned out pretty good too.</p> <p><a href="https://t.me/chillhousetech/552?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, hope it makes it to England someday too, though I haven’t used a single… https://darkolorin.com/posts/551.html https://darkolorin.com/posts/551.html <p>Well, hope it makes it to England someday too, though I haven’t used a single helmet before that</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/551?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A minute of history. https://darkolorin.com/posts/550.html https://darkolorin.com/posts/550.html <p>A minute of history. Or the Shazam everyone knows was a company from 1999 and it launched on the Nokia of that time — like, you turn on a call and it answers you with the track that’s playing. The guys took a long time to get to an exit, by the way it comes out to almost 20 years.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/550?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Very often I get a question like this, like hey tell me what hidden trends are… https://darkolorin.com/posts/549.html https://darkolorin.com/posts/549.html Mon, 15 Jan 2024 09:27:08 GMT <p>Very often I get a question like this, like hey tell me what hidden trends are there right now and everything’s super there, but NO ONE sees it. The question:<br> Where do you think, in what direction/domain, the stream is gurgling/boiling right now. Something is changing substantially, but it’s not noticeable. What in your view is really undervalued and deprived of attention by the tech industry. Everyone’s predicting trends, discussing something mainstream. For example, it seems to me the unsexy directions are undervalued: religion/church or caring for retirees and how they spend their time. I remember how Google Play was starting out and it immediately wasn’t very sexy. Both as a platform and the products that were on it.<br> 🙂<br> Well let’s start with the fact that trends are a pretty fast thing that fades out, and since I’m a consumer-type person, for me there are basic signs of where a trend can be and where it can’t. So, let’s start with where it can’t:<br> - anything 30+ can’t be sexy anymore, it’s a complicated, tedious topic where breaking a user’s habit is really hard, and also 30+ (age) have extremely limited social ties, few friends, they’re way less actively social, so targeting that at the very least makes the road harder<br> - accepting the fact that 30+ is a so-so topic, you can immediately throw out all sorts of religion and elder care<br> - next, insufficient density: a trend can’t take off on an audience that isn’t connected internally to each other in any way, which is why trends are often in schools, unis, communities<br> And finally if you want to ride a trend, ride AI or whatever is publicly obvious is taking off. Because it’s easiest to take off on a wave and get attention from both the audience and investors.</p> <p><a href="https://t.me/chillhousetech/549?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> UX changes that an LLM can bring, especially on-device, since on a server it’d… https://darkolorin.com/posts/548.html https://darkolorin.com/posts/548.html Sun, 14 Jan 2024 09:05:22 GMT <p>UX changes that an LLM can bring, especially on-device, since on a server it’d be kinda heavy (like it does a summary of a chunk of text, gives definitions and explanations and vice versa, it’ll even write extra bits if it’s still not clear)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/548?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Saturday news on “so where do they actually use LLMs at scale” https://darkolorin.com/posts/547.html https://darkolorin.com/posts/547.html Sat, 13 Jan 2024 11:56:48 GMT <p>Saturday news on “so where do they actually use LLMs at scale”<br> <a href="https://www.theverge.com/2024/1/12/24036156/openai-policy-amazon-ai-listings?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/2024/1/12/24036156/openai-policy-amazon-ai-listings</a></p> <p><a href="https://t.me/chillhousetech/547?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A bit about what’s been bugging me, or please write in the comments why the… https://darkolorin.com/posts/546.html https://darkolorin.com/posts/546.html Fri, 12 Jan 2024 14:08:29 GMT <p>A bit about what’s been bugging me, or please write in the comments why the hell it’s like this.</p> <p>I don’t get all these fundraising helpers, investors and so on, who can’t say who exactly they worked with, who they helped raise how much and in what timeframe, what the terms were, etc. Okay fine, terms are whatever, but why do all these advisors not have a single name, not a single check size, not a single link.</p> <p>All I ever see are lines like: my good friend, talked to some founder, had dinner with an investor from a well-known fund. Like, damn it, which fund, what friend, and what startup.</p> <p><a href="https://t.me/chillhousetech/546?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Next part of the Prisma lessons. https://darkolorin.com/posts/545.html https://darkolorin.com/posts/545.html Fri, 12 Jan 2024 10:54:24 GMT <p>Next part of the Prisma lessons. Deeper in meaning, or something. <br> I think the main thing I was missing was the ability to listen to a second opinion. On which exact questions was it important too:<br> - yeah on literally all of them :) in general you’ve gotta use your own head first and foremost, and having one piece of advice was often “the answer” for me, and yeah, that second opinion is me. A lot of people really underestimate their own decisions and lean on the market/advice/experienced founders. Know how to listen, but also know how to use your own noggin. The simplest advice in this regard that I’d give myself: take a pause, pin every important decision to yourself (so nobody except you can broadcast it), think for a few days. Even in an environment where I had to make hundreds of decisions a week, it’s possible.<br> - focus, yes yes, running in all directions means you don’t know where you’re running and it shows. Any normal fund understands that and feels it instantly. Where to run to build retention—there were no answers, well not “no,” there was one but it was cringe — we were making our own feed where it would show you photos in the feed by proximity and you’d unlock new styles by taking photos in different places. Like, viral. And don’t try to work two jobs.<br> - learn to be жестить, toward your own team too, fire people, don’t like it — the door’s open. Only this way can you get a culture of efficiency at an early stage (you won’t manage to hire only top people all the time).<br> - if your market is the US — go to the US, all of you (no explanations)<br> - networking has to produce results<br> - close deals fast, dragging it out is like the death of the term sheet<br> I’ll pay special attention to hiring and people. But for now we still need to pick other questions.</p> <p><a href="https://t.me/chillhousetech/545?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The next question was hard to pick. https://darkolorin.com/posts/544.html https://darkolorin.com/posts/544.html Mon, 08 Jan 2024 15:15:26 GMT <p>The next question was hard to pick. But probably this one:<br> “If I were building Prisma now, what would I change?” A question about retrospectives and lessons learned. You have to understand the following starting conditions:<br> - I had zero experience managing a company, before that I worked as a manager at Mail and Yandex<br> - for 25 years I was жесток and critical<br> - 2016 is very, very different from 2023/4 as a market<br> Given those starting conditions I did a lot of thoughtless things, obviously those are exactly what I’d need to change:<br> - The share and shareholder structure — the actual agreements and how it’s all set up — sorry, you need to do the classic, template-based approach. There were a lot of problems because of a pretty weird ownership structure and preferences. Back then Dmitry Grishin gave me the advice to do it without bells and whistles, but I didn’t listen to him. Look at how the docs are done in YC templates, ask your lawyers for a максимально simple and clear contractual base and shareholder agreements based on templates. Ask them to explain any additional “inputs” in plain language and spell out what consequences exactly they’ll have. Ask experienced founders what happened when you do it нестандартно. Ask specifically: “do we need liquidation prefs here and why and how much.”<br> - The ability to sell the company is super important. Not understanding the upside and the moment when you should’ve sold. This is important in general. Almost always an exit is what you need to do. And don’t try to squeeze out way more, especially if you don’t have a strong plan for what’s next. Back then, I’d say ours wasn’t that strong. Often this leads to shareholder fights, so think it through заранее.<br> - People. It’s important to keep balance and not hire too fast — and we managed that. Here I’d say that’s one big plus. But at the same time, if you’re doing a Series A, think through who you need and why.<br> - Consumer is very hard. And the wave of organic starts to fade fast. In our case it wasn’t as rapid as it usually happens now (a month). But at the same time there was no clear understanding of how to build retention.<br> I think I’ll continue later. Big topic.</p> <p><a href="https://t.me/chillhousetech/544?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, let’s start with one, as it seems to me, more personal question from the… https://darkolorin.com/posts/543.html https://darkolorin.com/posts/543.html Thu, 04 Jan 2024 14:05:24 GMT <p>Well, let’s start with one, as it seems to me, more personal question from the ones people submitted. Especially since I rarely talk here about any of my own decisions. Well, since the holidays are still kinda sorta going on, then sure.</p> <p>“Why did you give up on being public?”</p> <p>In reality it’s not like I straight-up gave it up. But overall — yeah.</p> <p>Main principles:<br> 1) It takes a lot of time, if you don’t know how to crank out content without straining. Content plans, editors, videos — all of that is hard for me. The most I can handle is a channel and Twitter sometimes. And even then, Twitter requires way too much attention to grow.<br> 2) I’m an introvert and I just don’t need a lot of communication. Networking often turns into nothing for me. There’s almost no benefit from the fact that I personally talked with a ton of billionaires or unicorn founders — or I just don’t know how to extract any benefit from it. And yeah, it’s important to understand here that networking works like this: on one level of connections it’s useful, but when you’re way weaker in track record, for me it only works if what you’re doing is potentially, or in the moment, highly attractive.<br> 3) I decided to dedicate part of my time to personal life and just resting. I keep playing online games, sometimes I watch movies/series. I don’t have all those stuffed calendars and I don’t understand why people brag about that on every corner. How packed your schedule is doesn’t say anything good.<br> 4) On top of everything, I think you should share what you understand really well, when you can give a substantial comment. Unfortunately the quality of knowledge in the industry is very low. Most content isn’t backed by anything. A lot of it is just rewritten without understanding the context. I was taught at university that you need to understand what you’re saying, and even more so what you’re writing. And that knowledge costs a lot of time.</p> <p>That kind of profile is called low profile. And once, traveling around business in China, I noticed that almost all the people who actually decide things prefer it.</p> <p><a href="https://t.me/chillhousetech/543?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Happy New Year, everyone! https://darkolorin.com/posts/542.html https://darkolorin.com/posts/542.html Tue, 02 Jan 2024 11:46:17 GMT <p>Happy New Year, everyone! Decided this week to do a series of Q&amp;As on whatever you’re interested in and whatever questions you’ve got for me, and I’ll try over the next 7 days to pick the most interesting topics and answer them here in the channel <br> 🙂</p> <p><a href="https://forms.gle/M9hJXbfpQtFfP4wc6?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://forms.gle/M9hJXbfpQtFfP4wc6</a></p> <p><a href="https://t.me/chillhousetech/542?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Who doesn’t love video concepts by https://darkolorin.com/posts/541.html https://darkolorin.com/posts/541.html <p>Who doesn’t love video concepts by <br> <a href="https://x.com/laurentdelrey?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/laurentdelrey?s=21</a></p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/541?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, for those who like podcasts and that kind of stuff They cover… https://darkolorin.com/posts/540.html https://darkolorin.com/posts/540.html Wed, 27 Dec 2023 00:06:21 GMT <img src="https://cdn4.telesco.pe/file/nCi_BF75XZF4jyNQWdGmUx5b2BA7GSoBsms64X8OUrNGB87Ad-qSexesYbGJSPQzTxJBdfXqYy4BOz0BP735WhF2r3E2ikn2LAwuk_2Gwv2zkF5dPusm8r7Tr5YgEidwmPoAAU7o66L5GpDI4sT17wJZYZmPcRLv9C1qwjvmMKeOV11PUUUghG8hvEyhPue4FLuXcj5002DCBHDDnaJ2gWp80Ib9bnY5RdCgvxoOAANUTZMRtI_9uTJsYyLjo09MNE5ynAzr5ZE7SQzA9h8hP0NckjwyBpzKq3BVtJ3U8vSVE9WkPyNe2M14Ndb_cSclgEKmAKXLVyBxSYcfQYOahg.jpg" alt="" loading="lazy"> <p>By the way, for those who like podcasts and that kind of stuff <br> <a href="https://www.youtube.com/@allin?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/@allin</a> <br> They cover events pretty well. The most interesting thing I’ve learned from there so far is that Chamath <br> <a href="https://x.com/chamath?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/chamath?s=21</a> <br> invested in 23 funds and the biggest returns came from funds sized 10–70M, which he says are small, and he thinks that’s where all the money gets made. There’s some confirmation of that in history too: one of the most successful funds, Benchmark, in the early 2010s was exactly “small.” <br> In the picture, returns from funds: a 100M fund returned 43x and a 220M one returned 12x (which, in his words, is already a mid-sized fund). <br> You can read about the metrics <br> <a href="https://www.crystalfunds.com/insights/tvpi-vs-dpi-the-measures-that-matter-for-private-equity-investors?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.crystalfunds.com/insights/tvpi-vs-dpi-the-measures-that-matter-for-private-equity-investors</a></p> <p><a href="https://t.me/chillhousetech/540?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m sitting here looking at a picture like this and… nothing. https://darkolorin.com/posts/539.html https://darkolorin.com/posts/539.html Tue, 26 Dec 2023 23:55:13 GMT <img src="https://cdn4.telesco.pe/file/rer8O_HRyKMzjDWXNPcRxOaBPNlp4vcqeKO43_JSaidA5ACAyj2dtZ9n93XARdhL7Bf8d2cw6fqkd3DJUaCNNDcyhAxxUc7E87JXDCUeiSWmvRLvCdwyqILAbiTUpop2ki_FY58JlollERW_ZFKI_UAXm2Pq9dWIYnWrPoqvf3fWSSEPajgDdQEUtpYb7-SxnrEwFewc4rJGAcJWjAJTIY87dE42BbM2yrNn1gyOOKIuDbiwuswgq9fqmvK4vK5N9JFM7IT5C_WdPP1fPrLgZQdfed9nVETTpYKSIHcucDn84lLsBBJWFAvNQ0nd60nBpU1UkSnwr2DzTzhfuy7C9A.jpg" alt="" loading="lazy"> <p>I’m sitting here looking at a picture like this and… nothing.</p> <p>At one point me and my team were at the origins of using neural nets for apps, or let’s say the end user. And since then it feels like they haven’t really come up with anything like that. With Midjourney I think they’ve got the biggest image dataset with user feedback. And I think that’s the company’s main value.</p> <p>My bet is — maybe Apple will buy them. A company that doesn’t collect datasets in-house, just buys them off the market.</p> <p><a href="https://t.me/chillhousetech/539?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Caught a great text on Twitter, it resonates really well with my thoughts and… https://darkolorin.com/posts/538.html https://darkolorin.com/posts/538.html Sat, 23 Dec 2023 03:55:36 GMT <p>Caught a great text on Twitter, it resonates really well with my thoughts and with what I usually do and what kinds of products I’m trying to build. <br> Consumer companies make up the majority of $100B+ technology outcomes, ever. The most important company of pretty much every VC cycle since the 70s has been in consumer. <br> So why have VCs been so reluctant to invest in it? Most VCs tell me that they &quot;don&#x27;t do consumer&quot;. <br> Continuation in the first comment.</p> <p><a href="https://t.me/chillhousetech/538?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Remember this one? https://darkolorin.com/posts/537.html https://darkolorin.com/posts/537.html Fri, 22 Dec 2023 23:35:26 GMT <p>Remember this one? Didn’t take off. <br> <a href="https://www.bloomberg.com/news/articles/2023-12-21/hyperloop-one-to-shut-down-after-raising-millions-to-reinvent-transit?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bloomberg.com/news/articles/2023-12-21/hyperloop-one-to-shut-down-after-raising-millions-to-reinvent-transit</a></p> <p><a href="https://t.me/chillhousetech/537?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A redditor said China is introducing new rules and restrictions for online… https://darkolorin.com/posts/536.html https://darkolorin.com/posts/536.html Fri, 22 Dec 2023 12:56:51 GMT <p>A redditor said China is introducing new rules and restrictions for online games. <br> A bright future for online games awaits us, because this is gonna roll out to everyone <br> <a href="https://www.reddit.com/r/Asmongold/comments/18o71lz/china_is_trying_to_implement_a_new_law_about/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reddit.com/r/Asmongold/comments/18o71lz/china_is_trying_to_implement_a_new_law_about/</a> <br> Some of the coolest bits:<br> - Forbids force pvp in online games.<br> - Forbids &quot;daily login bonus&quot;, &quot;1st purchase bonus&quot; and &quot;accumulative purchase bonus&quot;<br> - Not allowed to sell gacha (random pull) service to under 18s<br> - Encourage developing games that promote socialism core value, promoting the advantages of socialism, promote revolutionary culture, promote traditional Chinese culture, promote the advancement of technology and creativity, promote games that good for mind and health<br> - Limit how much an underage person can spend based on their age.<br> - Forbid underage donating during streaming.</p> <p><a href="https://t.me/chillhousetech/536?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> About PMs. https://darkolorin.com/posts/535.html https://darkolorin.com/posts/535.html Thu, 21 Dec 2023 13:33:08 GMT <p>About PMs. A big discussion blew up on Twitter about what the core skill for a product manager is. And surprisingly I liked how a lot of people consolidated around what, to me as well, seems like the most important long-game skill: storytelling. <br> A few points: <br> - this doesn’t mean the other skills aren’t important <br> - storytelling is about the long haul, it helps you explain what exactly you came up with, and it’ll always help <br> - caught myself thinking it’s basically part of selling (maybe selling really is the most necessary) <br> <a href="https://x.com/mlevchin/status/1737582782090461319?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/mlevchin/status/1737582782090461319?s=46</a></p> <p><a href="https://t.me/chillhousetech/535?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Midjourney v6 is out. https://darkolorin.com/posts/533.html https://darkolorin.com/posts/533.html Thu, 21 Dec 2023 13:29:11 GMT <img src="https://cdn4.telesco.pe/file/KBYNy7S76SYNvuooaWPi1pLyW59QpTTfI2bs_R_Jn5MJVL52ABUZkMpqH-ODTJHCpdHeDDNcPcoaNTxT9MdBN2zjAycndrZlsbT5D1TyPOY5FwJWeSrBoFUKYxbE29Ase4fejTBSQHCuKElYP_3VJnsG-p4tqpjsdw_P9qS0llmAWmskmluMX3xhO1pG53ZDyw6v4r5um9ZlNals7vfdG61eH6lsb2E-ujROu-t6WOd4wsgYbBwLnvTAQJs7whPyodQVW2oFhYDQ4SKerpFiFL7Pnm_WPf7wg2tkbUwfYAB-g2fypdZYCt2uHYFg2MQMUHIbJ2cA78-c_6TL1rn9_w.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/B3iBTRC5Sle_peURQ12GK6X_7P72cSzqzJmvhXWWJe3PtcbCAsE0uNDKFshbGvxXj92kFhK7lBi_MPhjxVV6KOqgyv0n6GE34WoehpGvTQncVxlF5k2FQN7DJcOS8X-dUxvYlTNkuEIr0Sb4uPNVYy8hSno_FLRsMTLDXqOSP1c5zjxX3TuH9bV_du3LpTOxJhSxxPggUM_m_90FmHhL7j10__l8RST7wvWoFtToUVsVD1QkoiR3-zFdfY52lNga0EjbZHr6DxPVxCFxvBH2BvKdNzkn3QKiXochazGru3JZP8db0g3RVifTybTQcACSYH1EENGajg5bpHEd7Mm5pg.jpg" alt="" loading="lazy"> <p>Midjourney v6 is out. Which of the two images do you think was made with the new version?<br> And overall I still can’t figure out where this company is headed.<br> I still think the main segment is b2b. Like placeholder images in chat avatars or on Medium and in presentations. Big services like YouTube will just build their own thumbnail generators. So where else can you even apply this tech (as a product)?</p> <p><a href="https://t.me/chillhousetech/533?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> That very product that doesn’t even know about itself https://darkolorin.com/posts/529.html https://darkolorin.com/posts/529.html Wed, 20 Dec 2023 23:38:52 GMT <img src="https://cdn4.telesco.pe/file/U7p7bqTy4f5ZsCUI3fVbzkCDYxjwG8rrbIv3J4_VM4gaOLVgT5QmTqQpBoSbuusOTP5s9givBo7Z-EZlumyRkJIe2Up2EeKfujNY3H5L1kkMd0ay7nBSik-2sDd9dOL8MOvjg9LDYPNnPK-y-Yy87Xc7cl0A_J5JB_Vuesyp7KP2YRZMtXhPyuvAKxILOjhjesujLSx_XpNoWW-8GTGAVL8-upQlfsBNI5w5qAo1hH9G56ZNM4aybwP1ylCYr_rA63jnsP0t5O4CI8dmVc4Sd4sdnyFi3gBWC2VeLvpGdS7uD_v9SnuBjDpbi2pDmGk7NsUnJT7U5ybZ5gyBnf8tLw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/AFSKA5bZHagnTvqGDBBkMaf-P0SXJoIKnUybicgqyq_nf4VbMxEKwgmQVK4zOzQSc7L1yzne0L_c8pfYzrUJQCKDlZZ9Zxo2kV0BawUdtv6ZgoA8ZAuAIgHIQzHlCFKKqgPYfkYSQO1wtfizzY41IS-OsXE_biFth5bDG87s2gr8vrghDiTEL01m1T_KHpmevbGkwgzynWEOhRy7fpCUk47LZ0g_S7T0brtHuvNiAyQH4WshRn_EKiXS8kktmFNF6xTnT-zgZtwOKUXYx9a9BuKWHRODk6DePCy1r9v7JsOONl7MaaMldlrbz4ms87muhRPQTN4WnhXb5Ebl0BGRxg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/YGJhrg96AOCiWJKxR1PHjpYMG2ugod7Ig68-1udAeiiTOkXADmHWwIt-D7LYUl8MIRde7QT-ZwqejrE9ofRUF_jmdzSrWVYaMtXdjt2rex0mZBqWhxmFhcUuShIrONkWfFaKvHMtCZ7AHMAjO9B3ZWgBqnRIs94pefLUjgWB0uQ2aBkK26OHmYpoykjk3QwFV6uF8IW9F_NeEcMQplaa-CVFhSPesfTCgBP5TgFupShdXGEJDpKPKT8iLTanFoOjw6i7ra-QTKMKERKFVa8_qxz3M_3iKE5nN64xyWFSHDZwnUvUElXnniIJ_3c81LFzvcTOBSNqb1B4ttShMlgkBg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/mODVOBCmaa-Ld1K3hyQZKLSjnJyPgO7B8crzWr2CjXZ0Gl3f9tb3rJdQKrXh6XqVpuSBKuVDMnM4L-YpYnHiIdpvADiHDd_GLgKDCv9LjFtvYgpMoc9LXf_RDk6Cg6eFUmfp426195Qrc4HsFnZqof7Tq1U4qBdfCy79OQuqZvG5i3UvYn8mygnHrxyRwer0Hhn4J9H-cIHfY0mI3YmtrHgb5MWZEbokkCN5UAj5MxBxhHgzwXs3Z8lwOslPsZhgB8V_tOe9dCiIxzKzMxEd7azeHAHSTzp7jTo_qw5DbmO0h03mAVtUbTmb1S1o6Q-WuW_rpAsQ6EB6Tremjk-M1w.jpg" alt="" loading="lazy"> <p>That very product that doesn’t even know about itself</p> <p><a href="https://t.me/chillhousetech/529?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> How do you make a million? https://darkolorin.com/posts/528.html https://darkolorin.com/posts/528.html Wed, 20 Dec 2023 23:21:41 GMT <img src="https://cdn4.telesco.pe/file/AeBMipFrcoX5QPU5zQf88w9iOrnB9WLB-LMBCKnY9GJ8gx9VQt3sLg_6j9QKzJfKDZJo9dagFJJgI-Ir7c9gM_Fk_HBdEn04AXBnOAW53s_9WbQY9iOcsy9Ujy14Suh1HJGLzPKv1LUvULBLNZy6T3jQdvrHow3ySCZ7zkq2fcKax_jAX4MogyTyzIpauKlyG5ZuDgb-xlpFs3P9S3XJuBNWdgDEfotpoNbWc6zxDwZT07rQdlFeBim1f_CKZz1VRrK9HmSf3beqpE9KmV8_otw2BIzWgs7ez7SmszAdpUvtO8arUj8P7CcyuB0hHQ9Lrapa2r2AFCBSYfBm_T24tQ.jpg" alt="" loading="lazy"> <p>How do you make a million? Which niches should you pick? Start right now. A podcast with Ayaz—oops, Ben and Mark.</p> <p><a href="https://t.me/chillhousetech/528?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A bit of politics in the AI world. https://darkolorin.com/posts/527.html https://darkolorin.com/posts/527.html Tue, 19 Dec 2023 16:33:47 GMT <p>A bit of politics in the AI world. But more like religious politics. <br> Ilya Sutskever pretty bluntly says a very weird thing (well, imo): <br> “predicting the next token well, means that you understand the underlying reality that led to the creation of that token” <br> Which goes against classic mathematical statistics, where for a long time they also thought that being able to predict the next value equals understanding and causality. But turns out that as of today that’s a pretty outdated view of things. <br> Links and details in the comments for those interested. <br> P.S. I obviously don’t share Ilya’s views.</p> <p><a href="https://t.me/chillhousetech/527?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One-liner: Adobe isn’t buying Figma. https://darkolorin.com/posts/526.html https://darkolorin.com/posts/526.html Mon, 18 Dec 2023 13:12:27 GMT <p>One-liner: Adobe isn’t buying Figma.</p> <p><a href="https://t.me/chillhousetech/526?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Lately I’ve been talking a lot about pitches and startups. https://darkolorin.com/posts/525.html https://darkolorin.com/posts/525.html <p>Lately I’ve been talking a lot about pitches and startups. And I noticed that a lot of people pay attention to the process of putting together a deck/plan and so on. They do “optimizations.” Which you don’t need at an early stage. <br> Or how do I put it—you should optimize what already works. And a startup you need to build.</p> <p>Another aspect is faith in what you’re doing and that you’re the very person who will actually build it or make the product or idea real. This is really important. I often notice that a lot of people have completely different goals. Make money, just be famous, or like, say what a cool founder I am. And a complete don’t-give-a-damn about what exactly the company you’re trying to build even does. I’m sure experienced partners at funds can spot that instantly and that’s a reason for a pass (especially in top tier). <br> A video about the same thing from Peter Thiel.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/525?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For all the AGI lovers, doomerism and the rest. https://darkolorin.com/posts/524.html https://darkolorin.com/posts/524.html Mon, 18 Dec 2023 12:39:35 GMT <p>For all the AGI lovers, doomerism and the rest. <br> There’s this paper up on arXiv: <br> <a href="https://arxiv.org/pdf/2304.15004.pdf" rel="noopener">https://arxiv.org/pdf/2304.15004.pdf</a> <br> And it looks at properties of emergentness, like when certain metrics grow like hell as you increase the number of parameters or whatever else in an LLM. Like, it straight-up claims there’s magic at scale. When hundreds of millions of people use ChatGPT it gets lazy. Well, no, that’s not how it is, as the authors say. And pretty often they basically invent these properties themselves by cherry-picking metrics that go up when you crank up some parameters.</p> <p><a href="https://t.me/chillhousetech/524?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> who’s out there shipping paywalls? https://darkolorin.com/posts/523.html https://darkolorin.com/posts/523.html Sun, 17 Dec 2023 02:47:43 GMT <p><a href="https://www.paywallscreens.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.paywallscreens.com/</a><br> who’s out there shipping paywalls? Here’s a solid one for inspiration.<br> And if you’re actually serious about testing, you can mess around with Adapty too.</p> <p><a href="https://t.me/chillhousetech/523?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A reasonable piece of advice, even if it’s toxic. https://darkolorin.com/posts/522.html https://darkolorin.com/posts/522.html Sat, 16 Dec 2023 17:25:09 GMT <p><a href="https://x.com/nikitabier/status/1736071454830051562?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/nikitabier/status/1736071454830051562?s=46</a> <br> A reasonable piece of advice, even if it’s toxic. But the point is, what’s stopping the team then from being bloggers and posting viral TikToks ;) <br> And yeah, in today’s reality it’s a very reasonable question about distribution, and if the answer is “we’ll post a TikTok and everything will take off” — then doubts about the team start popping up. <br> P.S. If the post got deleted and the preview disappeared, we dropped a screenshot in the comments</p> <p><a href="https://t.me/chillhousetech/522?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A bit from my favorite funds on the future of AI Lightspeed is making a big… https://darkolorin.com/posts/521.html https://darkolorin.com/posts/521.html Sat, 16 Dec 2023 13:09:23 GMT <p>A bit from my favorite funds on the future of AI <br> Lightspeed is making a big bet on legal. That’s reasonable — the lawyers market alone and all sorts of services around it in the US is about 50–60 yards a year if my memory isn’t failing me. <br> <a href="https://lsvp.com/legaltech-x-ai-the-lightspeed-view/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://lsvp.com/legaltech-x-ai-the-lightspeed-view/</a> <br> My favorite a16z is handing out forecasts for the year <br> <a href="https://a16z.com/big-ideas-in-tech-2024/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://a16z.com/big-ideas-in-tech-2024/</a> <br> And Ray Dalio’s Bridgewater is musing on how all this hype will actually change productivity <br> <a href="https://www.bridgewater.com/research-and-insights/assessing-the-implications-of-a-productivity-miracle?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bridgewater.com/research-and-insights/assessing-the-implications-of-a-productivity-miracle</a> <br> Light Saturday reading, so to speak</p> <p><a href="https://t.me/chillhousetech/521?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, finally the Zenly founders started making Zenly P.S. https://darkolorin.com/posts/518.html https://darkolorin.com/posts/518.html <img src="https://cdn4.telesco.pe/file/tgiDX4gfKR5FXQmiVakQfjyTnKDwhtu_ZBGD2ZIS7ODqOQTc3xXLqtRbK3BjWG98ctToUPIbywPYVFRkbCrR5UvpZJ-G9cktTtilGzg7Ga85mMsCkPrQjEdZfXr-9DcArKEfbLouPULbSzMxnWqQEEZYXeyhavvuUAGwUGsCUvUZkpTgiEEDq5HkfNW_2iIi9hMcgzCavANSw7gakdmwRJMfgQ-DJFq4iDwYPgV0wvRCpPjLxIdn1e4VR1AZ_51PFegsUgmd9qkEtOziXdbmwLbCoxpKxVLNUVDwLjAjFH5T6jBW51DFCczisnrtCPhc-ZHpPFffIhKGPewUHADCTw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/aBDaV-3EEIopUfmm4PV0OLWeLH83kN3t6_SUda1cooto0LG4Z5MGIUsQx6OsczexETL2fw3HfD6lT7xdMPZxD8CaBG6Us1-rXvjgahnLLNXmdch6qIALMI92DtDth0_AEMVKFPO8XScg8NBu3OWAgHJ5eyB_iTnGUon_4xMpsMq2X5ojmfvnbQibksp5lXrGrfWZDNJqsx3y459zeujA8FJ0TbxnB_h4z7rnP5XTpVgK2CwrFL9txmePN24BQuSQDaeSHiP2HT_X-bfTUhRV29yBSwF3O5kHCEebPwK9sRNwtuiua7_4lQJCi4BuSjoMTNhEoqLoin9-DirYqjZ-aQ.jpg" alt="" loading="lazy"> <p>Well, finally the Zenly founders started making Zenly <br> P.S. For the launch they picked the biggest market for the original Zenly — Japan.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/518?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Little by little this whole topic with NeRFs (neural radiance field is when… https://darkolorin.com/posts/517.html https://darkolorin.com/posts/517.html Wed, 13 Dec 2023 19:44:55 GMT <p>Little by little this whole topic with NeRFs (neural radiance field is when you’re walking around a 3D scene with movie-quality) is moving forward faster and faster.</p> <p><a href="https://smerf-3d.github.io/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://smerf-3d.github.io/</a><br> here you’ve already got scene streaming running straight on a phone. Google’s doing it, btw.</p> <p>What I’m getting at. The context is that everyone, one way or another, is slowly but surely gearing up for next-gen VR headsets and the Apple Vision Pro release. Well, I think by release we’ll see a couple of actually cool apps.</p> <p><a href="https://t.me/chillhousetech/517?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> GoPractice put out an adaptation of Andrew Chen’s article about referrals. https://darkolorin.com/posts/516.html https://darkolorin.com/posts/516.html Wed, 13 Dec 2023 10:39:33 GMT <p>GoPractice put out an adaptation of Andrew Chen’s article about referrals. Back in the day the guy was top. Now he often posts old info. But here it came out solid and his experience matches the market (even if he’s from Uber-of-antiquity).<br> <a href="https://gopractice.ru/channels/referral-programs/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://gopractice.ru/channels/referral-programs/</a><br> What I noted for myself, same as in the article:<br> - social products don’t work on paid referrals, even though he says there’s basically no such thing there at all, but in reality of course today’s “link a book and invite friends before using” is exactly that, they just found how to modify it<br> - low LTV == throw the referral program in the trash and it’s like, finally someone wrote it straight (hi to everyone who makes money on “drive traffic” and here’s a subscription screen)<br> - doesn’t matter what the message is and all that other stuff, you need to test the basic logic and the feature itself (it may totally not work if the target audience just has no motivation to share or there’s no acute problem)<br> You can read further on your own and there’s a link inside the article to a Dropbox deck.</p> <p><a href="https://t.me/chillhousetech/516?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> God, I’m so damn sick of these Product Hunt launches that are juiced up via… https://darkolorin.com/posts/515.html https://darkolorin.com/posts/515.html Wed, 13 Dec 2023 04:21:20 GMT <p>God, I’m so damn sick of these Product Hunt launches that are juiced up via mailings / “ask a friend” / “buy upvotes”. <br> Like yeah, I get it. But when you open a product and there are 50 comments and all of them are Russian-speaking by origin. What’s the calculation? Like other people will come in and go “wow so cool”. <br> I’ll be the marketing-backward dude for a bit. <br> - yeah I believe a cool product takes off on its own, without a hundred upvotes and mailings and begging and prayers to vote <br> - yeah I believe Product Hunt is a useless platform populated by Russians for like 30–40% for sure <br> - yeah I believe it’s better to go to Reddit and get honest feedback there, if there’s even anything to get feedback on <br> - yeah I believe you don’t need any courses, tactics, hacks for Product Hunt <br> - and yeah I believe the platform is overrated, same as LinkedIn and its fine-tuning</p> <p>PS: I’ll tell you for free how to become top1 on Product Hunt kekw</p> <p><a href="https://t.me/chillhousetech/515?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sergey Nikolenko has my respect, I used to go to POMI RAS in St. https://darkolorin.com/posts/514.html https://darkolorin.com/posts/514.html Wed, 13 Dec 2023 02:34:39 GMT <p>Sergey Nikolenko has my respect, I used to go to POMI RAS in St. Petersburg back in the day myself. A pretty solid series of posts about gen stuff with pictures, formulas, and all that. Not hardcore, but midcore, and it boosts your understanding quite a bit. You do need linear algebra of course and stats (better yet some slupas for those who are in the know) <br> <a href="https://synthesis.ai/blog/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://synthesis.ai/blog/</a> <br> And overall I’m seeing just a downright ridiculous number of AI projects that have zero idea other than to use the OpenAI API. In part that’s fine, but on the other hand — it’s hard to believe in the current generation.</p> <p><a href="https://t.me/chillhousetech/514?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> What kind of content https://darkolorin.com/posts/513.html https://darkolorin.com/posts/513.html Tue, 12 Dec 2023 17:16:15 GMT <p>What kind of content</p> <ul> <li>Whatever goes now</li> <li>Big posts with formatting and all that stuff are better even if rarer</li> <li>Add some kind of summaries</li> <li>Yeah whatever, main thing is it’s useful</li> <li>Your own suggestion in the comments</li> </ul> <p><a href="https://t.me/chillhousetech/513?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m curious: there’s a bunch of channels out there around all sorts of stuff… https://darkolorin.com/posts/512.html https://darkolorin.com/posts/512.html Tue, 12 Dec 2023 17:14:49 GMT <p>I’m curious: there’s a bunch of channels out there around all sorts of stuff from smart, educated, great top managers. Everyone writes so much good stuff there, like damn, these posts are something. I mean they prep content and all that. And I’m the only one who writes just off the cuff, whatever I found, that’s what I posted — like random observations or bits of some conversations (though sometimes I think it’d be cool to write summaries). Point is: should I keep going in the current “however it goes” style, or is it better to prep, like, one post a week but with substance?</p> <p>P.S. If I’m being honest, I don’t really have time to work on content. But it’s useful for knowledge. Since we’ve got a cozy audience here.</p> <p><a href="https://t.me/chillhousetech/512?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile, on the battlefield with Google, Epic Games is the one winning. https://darkolorin.com/posts/511.html https://darkolorin.com/posts/511.html Tue, 12 Dec 2023 14:47:34 GMT <p>Meanwhile, on the battlefield with Google, Epic Games is the one winning. 3 years of Tim Sweeney’s trash tweets, a pile of nerves and other stuff for the guys who make Fortnite and who I think basically lost a ton of users for good. <br> <a href="https://www.theverge.com/23994174/epic-google-trial-jury-verdict-monopoly-google-play?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/23994174/epic-google-trial-jury-verdict-monopoly-google-play</a> <br> P.S. As usual they’ll appeal the decision, what exactly and how they’ll punish Google for this whole “wow turns out” monopoly will stay somewhere in 2024.</p> <p><a href="https://t.me/chillhousetech/511?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Salaries have rolled up for AI folks P.S. https://darkolorin.com/posts/510.html https://darkolorin.com/posts/510.html Tue, 12 Dec 2023 13:59:23 GMT <img src="https://cdn4.telesco.pe/file/jx3AFCDEmU_tFO_jqyNVhghryHRGFq80kIPEPiqUAJc14AWEjff0r2gze6fKSuKCZPsBgNfgPkcUE186vKUXfDBleCX6YVmo0QSNficC2KYXSlq5I2Zk8k69FjUfg25n9A3s0QccoLqqMrDFcSCeQqDc2PP2EsNPj112BneWdv5lqzcx_0TbEaqwnF9f6xtz8ruHRCJ-5W9b3riOGsXVGN6v-3iax7-PulbECBvgoBCgZp3ERhWiBCzL9gJqT8-W1w00PYFhwufz6p8NtCUuffPqM2Lo8LeHmRJfTJUZ7dK38hDQDgyzpyxoG_QQ4-kVWdCSFxwfKkdbbNxVE9pDJA.jpg" alt="" loading="lazy"> <p>Salaries have rolled up for AI folks <br> P.S. AI Researcher Role</p> <p><a href="https://t.me/chillhousetech/510?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Now this is proper winter onboarding https://darkolorin.com/posts/509.html https://darkolorin.com/posts/509.html <p>Now this is proper winter onboarding</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/509?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh no, AI is gonna take our salaries and our jobs. https://darkolorin.com/posts/508.html https://darkolorin.com/posts/508.html Mon, 11 Dec 2023 21:33:42 GMT <img src="https://cdn4.telesco.pe/file/n0GTvPAngqxGIdvYmlqnD1JUeti_6_U24RnZUgvestxuHVCPMlNGD0oubLtY4MPuJjpUydaLPJ6q5qs4IVCI6jjl1EZY2UQEQH9Gri2rtvtkYJnEo3z2v0aFjZr6ueUcV3e_UNz5glH01U25dgZNT5KSqH9QCxX9Yer97TWLL4OO4Xh8-o08leyo7T9xr_a75kjNNzDmAFJ12-X5fYvd4dBVVffsbH5KGUbKmHnJ2rw5YTLpanQ2OBM773NbtWHBWYx6rSZFXhltHJZ8kbRpIpfDqotY-tc9RntC1Iv0DFyqmumISU9nvasiAV57oWaiPQgtMzD8tBQ1UPulYHoJcw.jpg" alt="" loading="lazy"> <p>Oh no, AI is gonna take our salaries and our jobs. Using the example of some *very* interesting cases. But overall there’s bias here. Because it won’t take the job, it’ll improve the current one.</p> <p><a href="https://t.me/chillhousetech/508?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/507.html https://darkolorin.com/posts/507.html Mon, 11 Dec 2023 21:27:25 GMT <img src="https://cdn4.telesco.pe/file/N2wOmbk6Xx57kZSvFAhJKwivj-TFoyYdEoXgw9Esrq87qQfiZ7Xl7tjAwt7H6DydhY1NNjxZXPlnx0msX_hft1i5JTYeGPfzMkpM2IPMiWYv4HrwWczVl9TCzekiJc1gju23gV85v-ymUo7ja0Cy96rURtelcc4K1nDfe0xFF_4X_oKgq520cmeGIEFPPpP-vlBB-__wFVPigHyXn42QZarRcbB4YGOjAxOLa0ICuaC0eJpMXCoy3KfbqEmM4zmwl5P1blRu1zaDhJ_I6kYynyObedAFag0Ju1AAlDGNVNkkuSmHQi2lOrP1Rp-nZT0qHwpv8FnlnijgzjgFf8Goag.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/507?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, ready to go grab money for some new very expensive research that’s how… https://darkolorin.com/posts/506.html https://darkolorin.com/posts/506.html Mon, 11 Dec 2023 20:10:22 GMT <p>Well, ready to go grab money for some new very expensive research <br> <a href="https://research.runwayml.com/introducing-general-world-models?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://research.runwayml.com/introducing-general-world-models</a> <br> that’s how you build a foundation</p> <p>And also progress in France <br> <a href="https://mistral.ai/news/mixtral-of-experts/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://mistral.ai/news/mixtral-of-experts/</a> <br> new thing with an expert model. Curious how they’re trying to beat the current quality stagnation. For now they’re going with speeding up the current quality.</p> <p><a href="https://t.me/chillhousetech/506?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well the trends are actually interesting, I didn’t even know half the names https://darkolorin.com/posts/505.html https://darkolorin.com/posts/505.html Mon, 11 Dec 2023 12:27:38 GMT <p>Well the trends are actually interesting, I didn’t even know half the names <br> <a href="https://trends.google.com/trends/yis/2023/GLOBAL/?hl=en-US&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trends.google.com/trends/yis/2023/GLOBAL/?hl=en-US</a></p> <p><a href="https://t.me/chillhousetech/505?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Let’s wrap up the weekend with my favorite problem for brainwashing my own… https://darkolorin.com/posts/504.html https://darkolorin.com/posts/504.html Mon, 11 Dec 2023 02:15:31 GMT <p>Let’s wrap up the weekend with my favorite problem for brainwashing my own brains <br> <a href="https://www.youtube.com/watch?v=TSBOBJsdEuY&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=TSBOBJsdEuY</a></p> <p><a href="https://t.me/chillhousetech/504?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One of the properties of communities that I’ve noticed. https://darkolorin.com/posts/503.html https://darkolorin.com/posts/503.html Sun, 10 Dec 2023 22:00:13 GMT <p>One of the properties of communities that I’ve noticed. <br> I’m in a hundred different Discords, and in general I read Reddit fairly regularly (well, like once a week). A lot of gaming communities live for a really long time (I’m not talking about teams or guilds). They can develop for years. Well, as long as the game works. <br> On the other hand, I’ve also been in dozens, maybe even a hundred, professional chats. You know, those investor/founder and whatnot ones. Very rarely did those manage to survive. All those product chats where, like, the wind is already blowing through after a couple months. <br> So yeah, one of the reasons is that people who are more or less the same level aren’t interested in being together. Boring. They’ve got nothing to write. In gaming communities the skill level is usually different and people are constantly asking stuff or looking for someone to play with. From that a pretty simple conclusion suggested itself: want to make a cool long-living community — don’t make it narrowly professional, try to unite around something where there will be completely different people and experience. In short, diversity, to some degree, is a very important thing.</p> <p><a href="https://t.me/chillhousetech/503?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “Sharing economy,” they said. https://darkolorin.com/posts/502.html https://darkolorin.com/posts/502.html Sat, 09 Dec 2023 23:19:07 GMT <img src="https://cdn4.telesco.pe/file/Fh0XeF6it08OtIFUlvvq57d9gr8UhWA9-GcJNI1c3S1_vJA3pd5rphoY3_76rTUt2QuGKMnvXqL-yMG1MsANKBDzJ8p3zbXFY3e4EuoeOlMOz386YTbNRJz5CT6HuWYSBspk1sjz3c65jcmZ4n_yEl6xmhalTR_8T2ExVDeWGuE9vTruOfp89j34ewDCc3ocRfYOSM-cDDrlTCNt9owmVj00aTZI_r48Okkg54c4SX04EtLvhzCgetfnAH75d2_qay4APs378wDZW72TDvu2AYrHZBpiYVlWjZtcySBbyR-J_ABWamoXoSvtrC4Tbq4tfqZnDqN-ky3sVI93j5hafA.jpg" alt="" loading="lazy"> <p>“Sharing economy,” they said.</p> <p><a href="https://t.me/chillhousetech/502?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Yeah sure, this is gonna happen everywhere, and Europe decided to be the first… https://darkolorin.com/posts/501.html https://darkolorin.com/posts/501.html Sat, 09 Dec 2023 18:43:28 GMT <p>Yeah sure, this is gonna happen everywhere, and Europe decided to be the first to step out the window into the inquisition <br> <a href="https://www.nytimes.com/2023/12/08/technology/eu-ai-act-regulation.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.nytimes.com/2023/12/08/technology/eu-ai-act-regulation.html</a></p> <p><a href="https://t.me/chillhousetech/501?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> This is interesting, Korea is huge on Twitch. https://darkolorin.com/posts/500.html https://darkolorin.com/posts/500.html Fri, 08 Dec 2023 13:08:50 GMT <p>This is interesting, Korea is huge on Twitch. And I think this is yet another confirmation that there will be less and less global products on the internet and more and more local ones. Hello China and Russia. <br> <a href="https://venturebeat.com/games/twitch-south-korea-shut-down-2024/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://venturebeat.com/games/twitch-south-korea-shut-down-2024/</a></p> <p><a href="https://t.me/chillhousetech/500?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “Russian venture” column. https://darkolorin.com/posts/499.html https://darkolorin.com/posts/499.html Thu, 07 Dec 2023 23:45:22 GMT <p>“Russian venture” column. This is entirely about investments and not just startups. But the philosophy goes further. <br> Stumbled on a webinar about raising investment. Alexander Gorny is answering. <br> Question: where do you even start with looking for investments? <br> Alexander’s answer: with understanding how to return them. <br> A thought about banks and loans kind of crept in. Wonder why.</p> <p>P.S. I’d say you start with results and people. When you’ve got something—product, idea, deck, numbers, and sometimes even desire. But definitely not with a payback plan. <br> PPS I think this exact answer from Alexander explains a lot about the industry and the success or non-success of Russian-speaking founders.</p> <p><a href="https://t.me/chillhousetech/499?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So yeah, I threw together a small doc about content. https://darkolorin.com/posts/498.html https://darkolorin.com/posts/498.html Thu, 07 Dec 2023 22:25:30 GMT <p>So yeah, I threw together a small doc about content. Or rather, how to think about it — I’d jotted the topic down for myself ages ago.</p> <p>Often helps to quickly size up what’s what if we’re thinking long-term or trying to choose what to focus on: users or content makers.</p> <p>If you’re gonna share it, drop a link to the channel :)</p> <p><a href="https://docs.google.com/document/d/1d7NhucsWA2egT-keW8sYKh5G9mtOebsCFNIw4G2hK_s/edit?usp=sharing&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://docs.google.com/document/d/1d7NhucsWA2egT-keW8sYKh5G9mtOebsCFNIw4G2hK_s/edit?usp=sharing</a></p> <p><a href="https://t.me/chillhousetech/498?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The legendary Google co-founder, Sergey Brin, is a core contributor to Gemini… https://darkolorin.com/posts/497.html https://darkolorin.com/posts/497.html Thu, 07 Dec 2023 21:53:58 GMT <img src="https://cdn4.telesco.pe/file/RJGp5FsqTAlncuR-7yGt1lapDCQCmwh6YLqSWgvAfnNwX1raFYWwwHjP6sPnTJiwSvChHPTE2xFOkdPG-StjcEKO2RFFagFdYXDHkr0WtU7VmjvEOV8b8EzLSjAHTXilHOx8zLpt40VntyrTR1M4hG_tHbBO8SozAEZfRkF6S5LJClSf6LxTeTdtezGXjYbFNfqyz1DEg4slJJK4IYfAr8sWc05PS50Ho0a-U0kk4e-stYlEZxUPZrCAKYCa0oGnUD2iX6SZVAvwslDKvjh-7hA_2_pBqgyQYzssaGwhzxufqT-ljCqYN1nJH8Smapdn3IXI8PhoZ9iTh2xajm_XrQ.jpg" alt="" loading="lazy"> <p>The legendary Google co-founder, Sergey Brin, is a core contributor to Gemini, apparently coding hands-on for the project. No matter how much he is worth, an engineer is an engineer at his core, and nothing eats that away. <br> According to Aravind Srinivas, Sergey was at Stability AI launch party last year chatting with Ashish Vaswani, the inventor of Transformer. Instead of small talk, Sergey asked him very low-level technical questions like why he picked sinusoidal position embeddings and whether he had tried other variants. <br> Passion never dies. <br> Such things.</p> <p><a href="https://t.me/chillhousetech/497?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Uh https://darkolorin.com/posts/496.html https://darkolorin.com/posts/496.html Wed, 06 Dec 2023 23:57:21 GMT <p>Uh</p> <p><a href="https://www.darpa.mil/news-events/2023-12-06?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.darpa.mil/news-events/2023-12-06</a></p> <p><a href="https://t.me/chillhousetech/496?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Video https://darkolorin.com/posts/495.html https://darkolorin.com/posts/495.html <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/495?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Not bad, gotta test it, but there’s this device https://darkolorin.com/posts/494.html https://darkolorin.com/posts/494.html Wed, 06 Dec 2023 16:04:23 GMT <p>Not bad, gotta test it, but there’s this device <br> <a href="https://deepmind.google/technologies/gemini/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site#build-with-gemini" rel="noopener">https://deepmind.google/technologies/gemini/#build-with-gemini</a></p> <p><a href="https://t.me/chillhousetech/494?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well written, though there’s a lot of water, it’s missing specifics in places… https://darkolorin.com/posts/493.html https://darkolorin.com/posts/493.html Tue, 05 Dec 2023 17:19:42 GMT <p>Well written, though there’s a lot of water, it’s missing specifics in places, but the approach to the article is great. Recommend it. <br> <a href="https://www.lennysnewsletter.com/p/lessons-on-building-a-viral-consumer?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.lennysnewsletter.com/p/lessons-on-building-a-viral-consumer</a></p> <p><a href="https://t.me/chillhousetech/493?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A small list of who’s using ML in production and how, sometimes there’s even… https://darkolorin.com/posts/492.html https://darkolorin.com/posts/492.html Mon, 04 Dec 2023 17:13:25 GMT <p>A small list of who’s using ML in production and how, sometimes there’s even GenAI &amp; LLM in there <br> <a href="https://www.evidentlyai.com/ml-system-design?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.evidentlyai.com/ml-system-design</a></p> <p><a href="https://t.me/chillhousetech/492?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Let’s wrap up the weekend with some illusions. https://darkolorin.com/posts/491.html https://darkolorin.com/posts/491.html <p>Let’s wrap up the weekend with some illusions. Been a while since we had any of that random bullshit. And during the week I’ll finally drop a couple things about retention and content consumption.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/491?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> You bought a BMW, please go buy another one too. https://darkolorin.com/posts/490.html https://darkolorin.com/posts/490.html Mon, 04 Dec 2023 00:11:09 GMT <img src="https://cdn4.telesco.pe/file/LahNx8pa-VOGy1LyArmBZuxfC_0ZWdXedUwptu9woqXl47ulyqYZnEzzztbT4JNQccOkFeXmq24MZIqnx06sfbbiib6oea9mKG4_l6JYwrolhJ--9BytL3INNpYybhV-6CsxSJx3OC0uFJKK5o4bnbMsnE7Zw3sE0j9mIys07t0_jdwNZX3q9g5DajRtVAjxzRCZAISev464Q1_kTeqAEK5_RAuu8ewq4OqvMCIgKrqGRGdGKv7IYrd530yIxWgG6YHXsrO3QrepEKjciPRRZRRT1kdsCPRYMWE06gg4GjxHrmMWGq77OozUmiBSlZ4FbwoSJqDl65EkUh1eVRkf5g.jpg" alt="" loading="lazy"> <p>You bought a BMW, please go buy another one too. The power of collaborative filtering in all its glory, right? I bet — nope, it’s just channel segmentation. And then cosine similarity (well, or something like that) from there.</p> <p><a href="https://t.me/chillhousetech/490?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Telegram rolled out a feature recommending similar channels. https://darkolorin.com/posts/489.html https://darkolorin.com/posts/489.html Fri, 01 Dec 2023 13:56:26 GMT <p>Telegram rolled out a feature recommending similar channels. In theory, if I end up in those “streams” my channel traffic should grow. Would be cool if they also added a traffic source in channel stats called “recommendations”. <br> Waiting for the feed. Kekwait <br> PS turns out they did add it. In traffic sources — similiar channels. <br> Long live the competition for trafflo</p> <p><a href="https://t.me/chillhousetech/489?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I like how the guys write these blog posts — the restraint, the way they… https://darkolorin.com/posts/488.html https://darkolorin.com/posts/488.html Thu, 30 Nov 2023 01:11:59 GMT <p>I like how the guys write these blog posts — the restraint, the way they communicate, and the overall vibe. <br> <a href="https://openai.com/blog/sam-altman-returns-as-ceo-openai-has-a-new-initial-board?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/blog/sam-altman-returns-as-ceo-openai-has-a-new-initial-board</a> <br> And Bret Taylor is of course a cool guy.</p> <p><a href="https://t.me/chillhousetech/488?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Not a bad communication style https://darkolorin.com/posts/487.html https://darkolorin.com/posts/487.html <p>Not a bad communication style</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/487?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> BVP has a cool section with memos about investments they made (or didn’t), and… https://darkolorin.com/posts/486.html https://darkolorin.com/posts/486.html Wed, 29 Nov 2023 11:30:44 GMT <p>BVP has a cool section with memos about investments they made (or didn’t), and given that just recently Emmett Shear, who was OpenAI’s three-day CEO, was getting talked about a lot, here’s a bit of his story there. In reality he basically built Twitch solo; Justin Kan and the rest had already tapped out by the time he decided to pivot the whole thing into gaming. <br> <a href="https://www.bvp.com/memos/twitch?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bvp.com/memos/twitch</a> <br> Good guy, go read him on Twitter. <br> <a href="https://x.com/eshear?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/eshear?s=21</a> <br> Known for being level-headed and calm. Vision driven.</p> <p><a href="https://t.me/chillhousetech/486?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Been following the guys since the start of the year, looks like they finally… https://darkolorin.com/posts/485.html https://darkolorin.com/posts/485.html <p>Been following the guys since the start of the year, looks like they finally finished something, waiting to test it.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/485?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well and this You can hook Slack data up to it too, and there are several… https://darkolorin.com/posts/484.html https://darkolorin.com/posts/484.html Tue, 28 Nov 2023 18:10:03 GMT <p>Well and this <br> <a href="https://www.nytimes.com/2023/11/28/technology/amazon-ai-chatbot-q.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.nytimes.com/2023/11/28/technology/amazon-ai-chatbot-q.html</a> <br> You can hook Slack data up to it too, and there are several models in the backend as well. Anyway there’ll be maaaany models and there’ll be room for everyone.</p> <p><a href="https://t.me/chillhousetech/484?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A pretty decent cheat sheet on startup economics, revenue, and the possibility… https://darkolorin.com/posts/483.html https://darkolorin.com/posts/483.html Tue, 28 Nov 2023 17:34:13 GMT <p>A pretty decent cheat sheet on startup economics, revenue, and the possibility of an exit <br> Who’s waiting for the post about Retention?</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/483?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A short summary from a lecture at UCL. https://darkolorin.com/posts/482.html https://darkolorin.com/posts/482.html Sat, 25 Nov 2023 15:26:46 GMT <p>A short summary from a lecture at UCL. <br> One of the first interesting questions was about the growth pattern in fast-growing social consumer. Like Prisma, GAS, Lapse. <br> In reality I had to unpack the thought and single out the signs:</p> <ul> <li>it’s always 99% organic. Yeah, in theory you can get your first users with traffic, but under no circumstances can you scale it. Traffic at the start is almost always a mistake, because even if you know what UA settings to dial in, it’s weird that you don’t know where on the internet your users hang out (they might, btw, be everywhere, or like in GAS—at schools, and buying ads “on schoolkids” is a so-so idea, because conversion is low; high only for games and clothes).</li> <li>Your target audience is not 30+, because adults install very few new apps and their habit around current tools is way too strong. And you’re most likely inventing a new form factor for old mechanics (messaging, challenges, check-ins).</li> <li>it grows very fast, and it’s important to get, on the day of growth (it might not be on launch), more than 5–10k users at once, otherwise there might not be enough for a quick jump into the top.</li> </ul> <p>We dug into Lapse in detail as an example of the compound effect. That’s when you do 5–6 steps in your main flow (ux), each of which increases the k-factor. But doesn’t add new features. Basically you bring a 5–6 step onboarding to perfection to start. Where each step boosts the power of the previous one.</p> <p>Then there was a question about retention. That’s for the next short post :) <br> Is this format ok?</p> <p><a href="https://t.me/chillhousetech/482?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/481.html https://darkolorin.com/posts/481.html Sat, 25 Nov 2023 02:14:54 GMT <img src="https://cdn4.telesco.pe/file/EwAyG3pzxODRoofQi_iEhUPUP9Oy1aQXbK25RD_8RrwRHyqQ70kNwkn568pxn9_ln8ghKJR3bveCXf8XRB1SZJRNPFvTealgr45WfsEMcSlBG56gmzGAz9zo29rm2wmxab9n8t3fMJ1TCeAbi9xBRfsbkcLPBMPA03DTlqB6l9_Engr3fl1wA6AP06vs7sutJhEiFcSyBL7w44xm08E1b9JhkzJrtiv94mVGCo0WUmJP8Sb1U5TWbFUio_lP3GVwje8q9jlUGsbV6HGEBU27XpWZSJ_jsNcP5eSuhXFq4pvi0N-_agTnjVveawZdfRJMRDmFENXUoS6DOr0gBAEnhw.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/481?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> CMO, CFO, CTO, CEO https://darkolorin.com/posts/477.html https://darkolorin.com/posts/477.html Fri, 24 Nov 2023 23:38:48 GMT <img src="https://cdn4.telesco.pe/file/TbpKH6sNrkQPl9kfqnlmZUtPX8W-w4PhkCx0t7vgmk9abCkfHhTSxCJguEvHjn-pTtV-7dTaiNzIqsFkgyKsu6FRCGdALuqry59uxHBCq9rGm48lZdldO6N7RH_TXWj0u_wFcFSwiel24nhxRe80LgV0rrnlIk_T9XV3VjD_R5NRJ-m7OhSuHqN9mYCP4ZfNK7VlCkY7wNoHC3lUPbjs6ZZ6ZLbtmxQZ5qO9C1BsLvFJsj0qi1ePsNADvY-vODHbFtz9B2GowhFVSJmwIltsBFtdKTLGlp4MPwXgyjA8sjVjolSNg0AEZg6EGc6Lt2WogJvE3jh1LVTzxmusZtwQLw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/D5kDUSRVrjsj34jtiIAlgBEBm3MwMLojM-Rt3IyFMhlNkg1T11bPDXj770UX-5y-Yk2hZG87EDnvBC2sjmjAQu4qo1Pz7VxR_vQOLzvmVQGy4it9dmzb-1Y-6HwG3w_MM1Ony-pNnz0AvAxqwXpNvFNlr_DgXVUYMxT-1rV65n4MxxwhqP7rWmTOPWKcHqoWl9gFGjNhJ2L7azSKLuHMNqbcRUrOl16rqfiG5MFLLsvWpogt8xzFw91SM6-d2dEIZlI6syDSMF7sfzMUaOhmpBvg7Y08dxMLN2oRUTpKc-sO1kmZDU3ZcOvl5cSH-8O8Krq_Mo3olt89kFrlAws5SA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/k0Lj1uX26wlgtraqbs84-dCuhZRs6gMGezmEBhbYR_6WK3tSHoVuzg5Z_LZ5Sa1n8DclAPJPJuHOXAqok0YbN8-jOZmG8I1qWpos2COF4GY6J1xv5pksiZErAj0O5E4Pvbp3IUZWpSjWtLtYGcVXmGCnUP4uLoeFrSlj4p2YbA8byeNKok231yDudpSGzvcrPJbNRjCu-rYot9d9AxTBf0yl2_mHowmSpAlgdcMqmif77_7NvWNOKsrvVHcZ-KPWDtekJVyzg123KvX8-rP4zln3PeFdVxQzWFx-WLwCpueKF0f_4BWZu9kqbqxwYTHM0d5pcPB62GsuEMJ25YzAcw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/ihRCz_b1PSJdlbF9c1KUB3ilWog4gKGBxfNdu1h4-TZ3iPizHqo-pAjhjelmoWoZnvGexXkC2kKxvYbVy8njvqyQLfZSoWN_1d_HG_SyL3Ibgyt8OxlEkhxbSz4Z1Qfx06C35GQZU6VH72CM7qChIMdT_Cd5n2URrNAOJ9IYQwI0TFTmGmy1Iw4AUtyDh3CN9wBogYhIcFNyYYsvWPiq_p1z2DVAhAxu6dstd3Au67zZUP_PgaGca_ex8cNzcd7GhcSpNkVS3MTit-7f69EagiuTphCVQoZ0HOkPX7jE7pCPRbk6JWosjEnx2WF7aXYuFOVQWLRpyYaap5TyQhPsGA.jpg" alt="" loading="lazy"> <p>CMO, CFO, CTO, CEO</p> <p><a href="https://t.me/chillhousetech/477?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> People create culture, culture drives people. https://darkolorin.com/posts/476.html https://darkolorin.com/posts/476.html Thu, 23 Nov 2023 19:09:35 GMT <p>People create culture, culture drives people. A good note to his people from Jack Dorsey. <br> Full text from Jack Dorsey to Block employees via Insider: <br> I want us to build a culture of excellence. <br> Excellence in service to our customers, excellence in our craft, excellence in our respective disciplines, and excellence to each other. <br> We want to help everyone achieve excellence here at Block. And if that&#x27;s not possible for any one person, we want to acknowledge that, and part ways without delay (which is a perfectly fine and honorable outcome.) <br> Our current &quot;performance management&quot; practices do not help us achieve this. In fact, they are holding us back. Some have described them as a &quot;denial of service attack on managers&quot; given the time commitment versus the benefit to our people. There&#x27;s also a perception that we allow people to &quot;rest and vest&quot; throughout the company, including poor managers who oversee great and promising individual <br> As we kick off our &quot;annual performance review cycle,&quot; we&#x27;re going to make some changes.</p> <p>Under the cut — continuation</p> <p><a href="https://t.me/chillhousetech/476?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m putting together a digest from yesterday’s conversation. https://darkolorin.com/posts/475.html https://darkolorin.com/posts/475.html <p>I’m putting together a digest from yesterday’s conversation. Turned out pretty interesting :) <br> There were a lot of students who are actively building different apps and projects, including offline ones. <br> There were also London startup folks. The topic was — ahead of trends. And they asked me, what’s the next trend? <br> I think the next trend is search. The most classic one. Basically ChatGPT and LLMs are a really good index and PageRank, and I like thinking about this tech exactly that way. And I’m expecting specifically search scenarios. <br> And in the meantime, here’s a GENAI tech use case I like</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/475?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Today I’ll be giving a lecture / or basically being the only guest on a… https://darkolorin.com/posts/474.html https://darkolorin.com/posts/474.html Wed, 22 Nov 2023 15:53:00 GMT <p>Today I’ll be giving a lecture / or basically being the only guest on a discussion panel at UCL in London on how to stay ahead of trends. I like sharing knowledge. I’ll definitely drop a short rundown in the channel of the most interesting questions and my answers. For now, wish me luck.</p> <p><a href="https://t.me/chillhousetech/474?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sam’s in his place. https://darkolorin.com/posts/472.html https://darkolorin.com/posts/472.html Wed, 22 Nov 2023 11:31:02 GMT <img src="https://cdn4.telesco.pe/file/rnGU7utyGeCCLjtR6PiJ2TSoLXU0NmWSzHt4TNdGz9nnsAJjdfkR0LK26bkdNDawLUkVJ4PElgS0zx7Tf041R9QJwOnNO380MqZE2lA4vqqEGzBXj81yhNz6DryrocdFWYckRXNgCUkvUFtvAi2fjRDzppPNuzVaYAoizhXtnkMQ6g1xxa6ovDE9brBik7y-Ma7Fq741nzw4wN91xFHAvjwczSd9TXRJlaudK8wcWHyI8drUr0Na9M-2XFlyqMKlZDQaU0Em4ELaIIZyJThq7i_REr6W5A6YwFEBdqZcVwq0C1C4Swt1LKjNaoXEUWl_jHMu45zKNbkm1XE0Eium9g.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/PD0X4dCZlW_wgcR71yXNSarQLUSfnofWsXEUTd3xcpCd1fXiaZvkt_9n3WDxtXtqimSY2dij-m0Hoz09c8Snr4dYsQ4wnDqVq8ASZ30O3riuh77biJ-XzeQQPCYXTKQ1az6HQF5ciubsA5rWVneH4RuTSrw2r1tvl_UIHG5E2qKLW_93l3IzIp8X0K7JXEkx5pPDTAHcDeEADQMwkoEMf3cBzwFYmjg-A-7xFhxUnyvPjo9Rmh3W3bjL0fNtTbLJtDTLXJuLTOi4ouwKf-gfS2Ti7myEARZG2Ur1wurAFGNRTg6pk5uz459SAbxL8cVL2203mu7Z5_aRgCLCYdtS8A.jpg" alt="" loading="lazy"> <p>Sam’s in his place. Satya, I think, got a bit sad. And Musk posted some interesting emails that were sent to him. <br> Why everyone’s staring so hard: <br> - standard drama, understandable inside the тусовка <br> - hype around OpenAI and, accordingly, almost the most expensive startup right now <br> - just to be in the know <br> But for me it’s more interesting how the fans of YC / the religion of how to “properly” run a startup and what to do, and the worshippers of all this, were in the process saying what a god Sam is. I don’t share that cult. And what’s more, I’m внимательно studying how this тусовка behaves in scandals this big. And it looks like a total mess, which says the approach is kinda bankrupt. But it’s hard to draw conclusions “for now”.</p> <p><a href="https://t.me/chillhousetech/472?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Now *this* is a proper commissions deal with Google, not those “startup… https://darkolorin.com/posts/471.html https://darkolorin.com/posts/471.html Tue, 21 Nov 2023 17:45:41 GMT <p>Now *this* is a proper commissions deal with Google, not those “startup programs” with turnover around a mil <br> <a href="https://www.theverge.com/2023/11/20/23969690/google-spotify-android-billing-commission-secret-deal?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/2023/11/20/23969690/google-spotify-android-billing-commission-secret-deal</a> <br> P.S. if anyone else has something like this too, DM me, I need an address / who to go to for pulls like this, because I’m sure it’s not some unique piece of shit, it’s just surprising they didn’t do it for Fortnite (guess there’s no money in Spoty)</p> <p><a href="https://t.me/chillhousetech/471?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Putting it in second place https://darkolorin.com/posts/470.html https://darkolorin.com/posts/470.html Tue, 21 Nov 2023 16:08:29 GMT <img src="https://cdn4.telesco.pe/file/JtF04xj4aqam85p5LOoA6pxdMevt59Pg0P59lwSat6wj4j_PaxfxOqZ6hZURmRbbk2c24S3mQ49Vs7qUzkSNCCJiudhWMLc9eV0rRGUYTtyEGo22OPBOyLecA-bCxOKhADPs97cFjjCKfWHpMSn-TIwBHQC3nhxdKO4O9k4fBd7Fwf8q-nKaF18KoO5PYGQqYAQjlclRp8L8dGtUlzPjyVtTDlFfZRnaT4JjSLTlzNZ6qF_CPZWyxili9JpVndan5u3Lizxcz8ye9DJFdv0Dd3-n82ubjd_Il8GbSYmXwjL3Op8VTyBxMbj_bahb2Xcalwfr7rE2hZbY_76UnJvNMA.jpg" alt="" loading="lazy"> <p>Putting it in second place</p> <p><a href="https://t.me/chillhousetech/470?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Found myself another keyboard. https://darkolorin.com/posts/469.html https://darkolorin.com/posts/469.html Mon, 20 Nov 2023 21:26:57 GMT <img src="https://cdn4.telesco.pe/file/eBg_JyXpw1MhhvcDkcsDJJ-sPHWV7aGdylwGTF_ij6oCMHJn8QMCEYuFy9p3Xn-sfDQIQKhRCaNjH4FeiA4i79nTI34rBSnki4zUXyhgZJ93ucLxTOtgl5ScD2aJE3lVN6mFwgU_iEbqf7N6ODJAn95NukYLbMeFgYuTWjM-oozVL_nLpOBD7kbXzH3T4RC_RbTPU0gO-MmnoAEHFd_emX_o-y8yakVjWfz9LRxrKX5tUjtOQcRTTlC-rehrY9WY2q8BetQkhMNRmUgp2XG9g9qB_JW-QPUsM0nG5WI9Sj-09erdVz9aQUCjj-UhDOZgpjQbFVfkwCA-Yivuco3rkA.jpg" alt="" loading="lazy"> <p>Found myself another keyboard. And the layover turned out like this: basically they gave us the airline hotel and it’s straight-up God knows where. And they gave me a suite. Thought — not bad for economy.</p> <p><a href="https://t.me/chillhousetech/469?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Hooli style https://darkolorin.com/posts/468.html https://darkolorin.com/posts/468.html Mon, 20 Nov 2023 21:24:34 GMT <img src="https://cdn4.telesco.pe/file/ds6VZ4McyYrKXaBfoxgrR4922r0aohmFGVv_CBVmLjV_s2Uoo-6YysRT_6132CywB1seEAp70jISZOTbUuWKXNGw-DMGfMcDxr6nl2uSELckkiAEQiMisO26OMjyRnA7K9Nm8jDo2HteI4d8KiVd88_7Lp95-hTNCZXk5PwRL2tZrGlFsRKSC2gKrt82y1QQHaK-b3aHBnYQCABIH8a-Ip4v-Ofpi_cLcaRULgiO4Yg93-g3XXZw18Y56tvBYd60EPJvYzq5Yy_IE2aWfPZ4IDQ4FRRuUS1DMRCJJ7YYaQNdP1k-FpODvlHf2_Ny97AGVDuRshd9GGZ2O2XnfRq9QA.jpg" alt="" loading="lazy"> <p>Hooli style</p> <p><a href="https://t.me/chillhousetech/468?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> While the endless Santa Barbara at OpenAI keeps going, I suggest this kind of… https://darkolorin.com/posts/467.html https://darkolorin.com/posts/467.html Mon, 20 Nov 2023 14:05:46 GMT <img src="https://cdn4.telesco.pe/file/oCZjW08YStoD0NNQar8rmxuEatAr59hBzbH0IW7HPN54R_rA8SY2dCq804GSmOl-9lPQyQnXkW3r883Eu2pIvhIoKkHVBIVZpX31gDmqG7SBwGRytJ_6Cnd8GeKc-sZC9_dPr8IRzZnMCBnxjbFJZeKCsqGb7e94WFfj_hTnyqhjcK0OiFl36m6HYQsw3HTGhlVEAYFaOzA7xT7A5LKDv1lpY8Va1aHtMVGoDiLVbOohvW-4ZDChqVBr7ByiQ-_3RM6kDKtigqqqqMbHOW4yKV7OaTvIPNLtYCVs-ac6bwAu8neuP7Tjsa0gdvW5XVKJmv98dKvbNVMTX429eTSYSQ.jpg" alt="" loading="lazy"> <p>While the endless Santa Barbara at OpenAI keeps going, I suggest this kind of feedback form for your apps. They say it converts like a charm.</p> <p><a href="https://t.me/chillhousetech/467?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, that’s the finale. https://darkolorin.com/posts/466.html https://darkolorin.com/posts/466.html Mon, 20 Nov 2023 08:35:53 GMT <img src="https://cdn4.telesco.pe/file/ZIWLVBgkNa2hj478DguD97Env2ypjZuYDeSnMDBR5HNHcV-VMMKMWX7LWTuX2-AfRTv8OGr6jmHeVvmmP7uZqRp78Q8UykdWMXckmWk0dTjohPnPhJ2dDIaOYRr_GCg2RhRMPmuU0hqA5f4-JNX_X3wFtkQYAZaI0TXq0u28EoyOf7C-pSRj9MTpDH7k3_jTfMFceSt0aBiA7__HctbAYwBbpF4L2enUznjgRqrKiUuPk5sp1W0IhdIJTVr1k_rlt_j0j4jiH61DCIqxoPrRXXL2X2J-WAceNUJg0XhDFbvoohqkMQFnka-4rbMS-xAH9AEr_KlIv4LuhYblGz_n4w.jpg" alt="" loading="lazy"> <p>Well, that’s the finale. Sam just went to Microsoft.<br> P.S. For all the hardcore AI fanboys: the last paragraph of the new CEO’s tweet<br> <a href="https://x.com/eshear/status/1726526112019382275?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/eshear/status/1726526112019382275?s=46</a></p> <p><a href="https://t.me/chillhousetech/466?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh yeah? https://darkolorin.com/posts/465.html https://darkolorin.com/posts/465.html Fri, 17 Nov 2023 20:55:57 GMT <p>Oh yeah? <br> <a href="https://openai.com/blog/openai-announces-leadership-transition?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/blog/openai-announces-leadership-transition</a></p> <p><a href="https://t.me/chillhousetech/465?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Next wave in social. https://darkolorin.com/posts/464.html https://darkolorin.com/posts/464.html Thu, 16 Nov 2023 16:33:38 GMT <p>Next wave in social. Noticed a lot of attention on collaborative editors/spaces. Kind of a multiplayer thing. <br> In short, game-like interaction formats are starting to burst in. Remember this post.</p> <p><a href="https://t.me/chillhousetech/464?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Amazon? https://darkolorin.com/posts/461.html https://darkolorin.com/posts/461.html Thu, 16 Nov 2023 14:41:37 GMT <img src="https://cdn4.telesco.pe/file/c5jOuSgKdvq7dDFJC5CbiAhoX4sFFboL_DCKBplce1adDkinACj7-n8bRfco3KBsj46SSpkDCgtuFVoDNEVUeBApTR4JxQIFpQ3jru2fp1PTfrybKo1Dm8Fe7Kmj78KnSDXyMG3jDRipXdgXksn8OAFPfRLyJKP8BU0CgmvfRbINfLnyYNytGoDRMb5LL06BMCe2-HZF67NNUREBTqWJWj8o71nl2qGLztoK0HX9Msm7LvwzwRxum0G2CmqVliwrgnc7oUqEriKwi1L6y28DToE0nIbzQD4iCTM63LvYEuckhagAfRhTp6UTpR8_pxrPdQuC1219OJJvAI7nHsst-A.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/qHsJKj1Aoulgw9U_4-YeEeGbyyzRq8nJU53lCjNJftjTGQsy7lZAiGghrfs6MZqK9Vg9cjVpg_YfyfOc9g5ugTm_xJv2LegQKfi20HhHhmCPuu3M_KGUsXMKl58AlxtcSlUnEVkvDtHYI_t3RGLAp5i6ziK3QRi5xv3blQ3BkSTVIgA75BO98bFZLhaaw9b_Xm_mv96VidKiL1OO1JW2pPwP2VKrCrh7nR2f5o2vi8b4ms6BOvFmZKBvaVL2-L_H6Fn5IvXK7gFIVqqYB4kYOSIJvmQRfitSD-tyQM89kREqj42IQsdRlMnRoZYPgEBA6fFW-Nnl74YFo6cxejDpEA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/DpsPzrjLD_Ks0WNU6wHZ6kS0017wmEyUHbo6srZT-1_-g3KJDOii0aNsrnWC6G9eUZsfXygYj8K24iijebTM45b5Z2IVzxDAr1skXX1fhnVoDhzRPmVEPg1jyTTMjvY5tD7vLgU80LZiojH9ZfLx4gq4pRTDefoZkeJl17YkC6ek8Q-4bEZ8QZAOH6DBUgqgbALs4SbSlvNCC7NP8VroHhdD86fOdXZ8R8DlxMUa5W0g-sexLojfOEvevu_ykFCLkIjhzPQ1VB9p7JDzxCw2TjxtvHx3vYLZidx763LLKtEXKnQ-7yQsIq6FlhF5kS2UDONAEGfSGOWKh9vPVtrTbg.jpg" alt="" loading="lazy"> <p>Amazon?</p> <p><a href="https://t.me/chillhousetech/461?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/460.html https://darkolorin.com/posts/460.html Thu, 16 Nov 2023 12:00:43 GMT <img src="https://cdn4.telesco.pe/file/tErEeE0m8vz9X-qtny-_GL66g7XPO3PCc5R1VKA0J6UxeS1-Rgjo0RCe-UEE1atuuogE7gUNhysdkE7MLV-GUWTYY7Xfe8NSscZbGlRMovJh1W1hp3PXYOFIXOJ4QEAkpCo2c7xu_vVh7gFObxDeKyEoG26X3mOGwvV0kVUG8JFCtSjvsFrDbO5BoF2AMpgz-Wc5nK6bzX5KARRJoqah1fMtKJuRbP9WrL0gOi7YrautYB5kpsf3IU5G2gDRdiY6t1FIFwzi2gxBtWMjy0RDXYnHUUZCDn_rcf1rclExyx5-E1Q1gHIDoHp_D_G_4-kMlJc2BIPhnVLNSM-UnuN6nw.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/460?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Buy a spot in a traffic jam on the canal — 4M USD https://darkolorin.com/posts/459.html https://darkolorin.com/posts/459.html Thu, 16 Nov 2023 11:08:26 GMT <p>Buy a spot in a traffic jam on the canal — 4M USD <br> <a href="https://fortune.com/2023/11/08/panama-canal-congestion-record-4-million-skip-line/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://fortune.com/2023/11/08/panama-canal-congestion-record-4-million-skip-line/</a></p> <p><a href="https://t.me/chillhousetech/459?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I love watching how censorship works in the US. https://darkolorin.com/posts/458.html https://darkolorin.com/posts/458.html Thu, 16 Nov 2023 08:08:57 GMT <img src="https://cdn4.telesco.pe/file/fyDoDQBGo4dd0Z8QxyBIYwv27iSsqmtut_bdJWFSJKaN883SgSjZmoDUtw8sU8wpqZgBppWywE-wna3Nh41s-BKuK-RBPsgZKFQxycW6nMC6KOY69Pgbjzn2art3sxNqObfv_mwdzL_-wnU7ob2QcYFKoDYihf4aua5px1RIo-3AR4XEj5BwyVEmKh40f3x2BRMRf2byvERG1zvIwFQYqmSV6lLjPLyETdaPy0MC5infAshxzTivMA3mMWKCrdAwz1B00vavDJZ5kY2yBfSk3M1DIW2uxMMOOjwp5qOLRgctO2GdBMvKgaZZ2FqYkdWBYny7bS3NFKJlz-Hy9EqV_g.jpg" alt="" loading="lazy"> <p>I love watching how censorship works in the US. <br> <a href="https://about.fb.com/news/tag/coordinated-inauthentic-behavior/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://about.fb.com/news/tag/coordinated-inauthentic-behavior/</a> <br> I’m predictably waiting for reports like this across literally all social networks. And where can I look at the Russian ones?</p> <p><a href="https://t.me/chillhousetech/458?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Fresh Series A from Benchmark A spatial editor, kinda like Figma. https://darkolorin.com/posts/457.html https://darkolorin.com/posts/457.html Wed, 15 Nov 2023 19:38:40 GMT <p>Fresh Series A from Benchmark <br> <a href="https://www.bezi.com/hq?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bezi.com/hq</a> <br> A spatial editor, kinda like Figma. A good bet if the headsets take off.</p> <p><a href="https://t.me/chillhousetech/457?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Xiaomi’s car just rolled up https://darkolorin.com/posts/456.html https://darkolorin.com/posts/456.html Wed, 15 Nov 2023 13:06:27 GMT <p>Xiaomi’s car just rolled up <br> <a href="https://www.bloomberg.com/news/articles/2023-11-15/xiaomi-ev-closer-to-production-as-china-approves-roadworthiness?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bloomberg.com/news/articles/2023-11-15/xiaomi-ev-closer-to-production-as-china-approves-roadworthiness</a></p> <p><a href="https://t.me/chillhousetech/456?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Put Ohanian’s thread about the American dream into one post. https://darkolorin.com/posts/455.html https://darkolorin.com/posts/455.html Tue, 14 Nov 2023 23:11:57 GMT <p>Put Ohanian’s thread about the American dream into one post. And yeah, this really does happen. The guys got together, got turned down, but still did it. And yeah, it’s not a unicorn/billion-dollar story, but the very fact—what kinds of stories happen. I’d like to read more of these. There aren’t enough.</p> <p>How to turn $4,000 into $100,000,000+ <br> 💸 <br> They couldn&#x27;t get a single investor to take them seriously... <br> That&#x27;s when 4 friends decided to each invest $4,000 of their own savings to launch Chubbies. <br> One day... 4 friends noticed that there was a lot of room for opportunity in men&#x27;s fashion. <br> They asked themselves the following question: <br> What if we reinvented men&#x27;s shorts... and made them *gasp* VERY SHORT? <br> They pitched the idea to multiple investors... <br> 👇 <br> &quot;Picture this - tiny men’s shorts as far as the eye can see…and we’ll call it Chubbies.&quot; <br> 🩳 <br> 5 minutes into their pitch, they were politely asked to leave. <br> They couldn&#x27;t get a single investor to take them seriously... <br> 🙃 <br> But that wasn&#x27;t going to stop them. <br> They had a vision of men everywhere breaking free of the shackles of overpocketed, shin-length shorts and they wouldn’t be stopped. <br> So they put together $4,000 each to bootstrap the business. <br> 🚀 <br> In the early days, they worked full-time jobs while building Chubbies on the side.</p> <p>Moved the continuation into the comments so it’s not so much text <br> <a href="https://x.com/alexisohanian/status/1724473217770033269?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/alexisohanian/status/1724473217770033269?s=46</a></p> <p><a href="https://t.me/chillhousetech/455?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For some reason it’s doing pretty damn well in Asia https://darkolorin.com/posts/454.html https://darkolorin.com/posts/454.html Tue, 14 Nov 2023 22:54:42 GMT <p>For some reason it’s doing pretty damn well in Asia <br> <a href="https://apps.apple.com/gb/app/vibes-widget/id1660118341?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/gb/app/vibes-widget/id1660118341</a></p> <p>But the most interesting thing is that in Japan the next most popular social app (number 5 and already number 2) is, I don’t even know how to describe it, photos and a social network around plush toys <br> <a href="https://apps.apple.com/gb/app/%E3%81%B5%E3%82%8F%E3%82%82%E3%81%93sns/id1464870286?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/gb/app/%E3%81%B5%E3%82%8F%E3%82%82%E3%81%93sns/id1464870286</a></p> <p><a href="https://t.me/chillhousetech/454?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Telegram squeezes quality. https://darkolorin.com/posts/453.html https://darkolorin.com/posts/453.html Tue, 14 Nov 2023 19:39:35 GMT <p>Telegram squeezes quality. All PMs, take note.<br> <a href="https://www.lennysnewsletter.com/p/when-they-hired-their-first-pm?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.lennysnewsletter.com/p/when-they-hired-their-first-pm</a></p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/453?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Remember that Google pays Apple for search. https://darkolorin.com/posts/451.html https://darkolorin.com/posts/451.html Tue, 14 Nov 2023 12:23:35 GMT <p>Remember that Google pays Apple for search. And on top of that there’s rev share too, and it’s pretty damn decent. <br> <a href="https://www.bloomberg.com/news/articles/2023-11-13/apple-gets-36-of-google-revenue-from-search-deal-witness-says?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bloomberg.com/news/articles/2023-11-13/apple-gets-36-of-google-revenue-from-search-deal-witness-says</a> <br> It’s interesting that basically the same kind of deal exists with Samsung, only the volume is several times bigger given the preinstall of services. <br> <a href="https://www.bloomberg.com/news/articles/2023-11-14/for-google-play-dominating-the-android-world-was-existential?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bloomberg.com/news/articles/2023-11-14/for-google-play-dominating-the-android-world-was-existential</a></p> <p><a href="https://t.me/chillhousetech/451?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> You can kinda sum up the release too They announced a presentation, a $700… https://darkolorin.com/posts/450.html https://darkolorin.com/posts/450.html Fri, 10 Nov 2023 13:19:54 GMT <p>You can kinda sum up the release too <br> <a href="https://hu.ma.ne/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://hu.ma.ne/</a> <br> They announced a presentation, a $700 price tag, and another $25 subscription for OpenAI services you can’t skip. By the way, awesome model — you bought a device that doesn’t work without a subscription. Waiting for an iPhone in the same style. Very solid theme. Kekw. <br> As for the device and the presentation: <br> - an omega-innovative approach from the company. Well I mean the level of risk is huge. And I’m sure you need to move into consumer hardware and moreover the combo of AI and hardware is the most powerful thing. <br> - cool team, but damn, old. I’m watching this presentation and it’s like some old dudes walked out, said their bit, and left, feels like they’re in a different world. A product from old people for old people. Sorry. But I can’t imagine a 15-year-old girl buying herself one because it’s cool. And that feels like an important factor to me. <br> Drier facts: <br> - hangs on clothes, pulls down the fabric of light wardrobe items (you can see it on a T-shirt) <br> - what about rain? <br> - what about visibility in the rain? <br> - what about charging? <br> - and if you clip it on a jacket like in the example and then just toss it on a chair, everything ok? <br> - what about kids’ hands? Different hand size and so on? <br> And yeah, I’m not really seeing compatibility with this, even though it’s not required <br> <a href="https://www.interaction-design.org/literature/article/design-for-the-future-but-balance-it-with-your-users-present?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.interaction-design.org/literature/article/design-for-the-future-but-balance-it-with-your-users-present</a></p> <p><a href="https://t.me/chillhousetech/450?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> People often write asking what an investor update is supposed to look like. https://darkolorin.com/posts/449.html https://darkolorin.com/posts/449.html Thu, 09 Nov 2023 00:13:59 GMT <img src="https://cdn4.telesco.pe/file/QlrFgLfTPHncvl1iW5MkNl8YyE4s_4y3iJxJI-1K3onm9k5WrFmd5QNHKZb0dAZ73m24SOZYjHxZRyTQ2PP53SyxlqHyG6ltyt2LxKAQl4KjTBhFQd1A2pjXTlgjMcdU_guFKiSA4NzOM3zXNaqY1bsc-p1slqE3MIj55IEJYN_GDXdh0niCjcP9WsZCtTvf06srBuDfKy09AsvEvhca3fDbH_VqEgl7P52KJOmtwnlEYiM1ppS0Wu7pm0S-NPjt4GQKR-n6dtUZFVLrYXK7_3dlbOuy2MSiunav6tSBLcEbi5FbSXM3DEPdSq44Mhez7EeOWL1A4zQTPQcV1VIkBg.jpg" alt="" loading="lazy"> <p>People often write asking what an investor update is supposed to look like. I think a format like this is great. Whether you write once a month or every three depends on the project and the stage.</p> <p><a href="https://t.me/chillhousetech/449?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The hardest stage, really. https://darkolorin.com/posts/448.html https://darkolorin.com/posts/448.html Wed, 08 Nov 2023 20:11:27 GMT <img src="https://cdn4.telesco.pe/file/geaU0LFr3rA5aJa5QZTOj6uFeuUJ9Gh90YxMMMa4y3PYX28OIT5zeDiJJzE1QAfS_MOmRqBVr_SGeXYdDan-zWOdR61q_b0Cpwx_2uUCxJYQMNrhVZ7aESBfESRQnMxtRNFFIUS2WAhPVlQkH1MEtRkssESGuvU11eQHStWojxF5_ntvAbde_mnTPbg6b7GSsqsUrl7WFNIIlDnmZbmxOCBnMEXLyTP7tYDyRNyUBT_xyXPtLNTtZ_veb7URqbxVLa0Bogg5ThT581Tf02c_6JnWjRLWp4pg0hdE7ht8mA7lu5a2IMJIGxHBtynkZfoU_qkn--VVouYkym4OOVQepQ.jpg" alt="" loading="lazy"> <p>The hardest stage, really. Because it’s easy to invest in an idea.</p> <p><a href="https://t.me/chillhousetech/448?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Yeah, still way too much, the fund is trying to jack up the valuation at least… https://darkolorin.com/posts/447.html https://darkolorin.com/posts/447.html Wed, 08 Nov 2023 17:33:46 GMT <p>Yeah, still way too much, the fund is trying to jack up the valuation at least somehow <br> Coatue Cuts Value of its Stake in NFT Marketplace OpenSea by 90%, implying the OpenSea&#x27;s cap down from $14B to $1.4B</p> <p><a href="https://t.me/chillhousetech/447?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> OpenAI is a new AppStore? https://darkolorin.com/posts/446.html https://darkolorin.com/posts/446.html Wed, 08 Nov 2023 13:31:40 GMT <p>OpenAI is a new AppStore?</p> <ul> <li>Well yeah, like fuck it is, what else</li> <li>Yeah, but it’ll be a small one</li> <li>Dunno</li> <li>Of course not</li> </ul> <p><a href="https://t.me/chillhousetech/446?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The last step you want to take when job hunting (if you’re even a bit past the… https://darkolorin.com/posts/445.html https://darkolorin.com/posts/445.html Wed, 08 Nov 2023 01:04:19 GMT <p>The last step you want to take when job hunting (if you’re even a bit past the entry-level stage) is “repackaging your LinkedIn.”</p> <p><a href="https://t.me/chillhousetech/445?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Yesterday we already started building it, we feed video and frames into GPT… https://darkolorin.com/posts/444.html https://darkolorin.com/posts/444.html <p>Yesterday we already started building it, we feed video and frames into GPT Vision and on the output we get a video with commentary. Why the hell do you need a commentator? <br> P.S. Not me doing the building :)</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/444?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, TL;DR Overall the guys are gonna keep paying for products on their… https://darkolorin.com/posts/443.html https://darkolorin.com/posts/443.html Mon, 06 Nov 2023 22:36:30 GMT <p>Well, TL;DR <br> <a href="https://openai.com/blog/new-models-and-developer-products-announced-at-devday?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/blog/new-models-and-developer-products-announced-at-devday</a> <br> Overall the guys are gonna keep paying for products on their platform. But for now it still looks like an Apple-style monopoly. I think next up will be a device (an earbud most likely).</p> <p><a href="https://t.me/chillhousetech/443?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The new Jobs is trying or not really yet? https://darkolorin.com/posts/442.html https://darkolorin.com/posts/442.html Mon, 06 Nov 2023 18:09:45 GMT <p>The new Jobs is trying <br> <a href="https://www.youtube.com/live/U9mJuUkhUzk?si=NYUMAHPS8341x3my&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/live/U9mJuUkhUzk?si=NYUMAHPS8341x3my</a> <br> or not really yet?</p> <p><a href="https://t.me/chillhousetech/442?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Ilya’s good, sure, landing big deals. https://darkolorin.com/posts/441.html https://darkolorin.com/posts/441.html Mon, 06 Nov 2023 16:31:01 GMT <p>Ilya’s good, sure, landing big deals. Kicking the door in to gamedev. <br> <a href="https://inworld.ai/blog/xbox-partners-with-inworld-ai-to-build-generative-ai-tools-for-game-development?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://inworld.ai/blog/xbox-partners-with-inworld-ai-to-build-generative-ai-tools-for-game-development</a></p> <p><a href="https://t.me/chillhousetech/441?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Alright, now it’s clear what this looks like. https://darkolorin.com/posts/440.html https://darkolorin.com/posts/440.html Mon, 06 Nov 2023 15:51:45 GMT <p>Alright, now it’s clear what this looks like. Turns out the end date selection for the giveaway is bugging out. <br> The news is here <br> <a href="https://telegram.org/blog/giveaways/ru?ln=a&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://telegram.org/blog/giveaways/ru?ln=a</a> <br> You can pay for a give on the Fragment platform with Ton. Didn’t find how to do it by card. <br> What’s more, it’s weird that all subscribers are auto-entered. That’s something new for pumping bots into channels.</p> <p><a href="https://t.me/chillhousetech/440?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Media on Telegram https://darkolorin.com/posts/439.html https://darkolorin.com/posts/439.html Mon, 06 Nov 2023 15:49:56 GMT <p>This media is unavailable in the public Telegram preview. View the original post on Telegram.</p> <p><a href="https://t.me/chillhousetech/439?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Media on Telegram https://darkolorin.com/posts/438.html https://darkolorin.com/posts/438.html Mon, 06 Nov 2023 15:49:56 GMT <p>This media is unavailable in the public Telegram preview. View the original post on Telegram.</p> <p><a href="https://t.me/chillhousetech/438?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So today Telegram rolled out giveaways like on all the streams, and I’m… https://darkolorin.com/posts/437.html https://darkolorin.com/posts/437.html Mon, 06 Nov 2023 15:30:47 GMT <p>So today Telegram rolled out giveaways like on all the streams, and I’m curious to test it. I want to see if this affects pulling people in the same way it does on game streaming platforms.</p> <p>How to tell if the giveaway even started — I’ve got no clue. Paid for a pre-order on subscriptions and now I’m sitting here thinking :)</p> <p><a href="https://t.me/chillhousetech/437?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Hadn’t noticed before, TikTok has an option to slap an “AI generated” label on. https://darkolorin.com/posts/436.html https://darkolorin.com/posts/436.html Mon, 06 Nov 2023 01:22:09 GMT <img src="https://cdn4.telesco.pe/file/TidyWB6DgySsKgy_o0C6bzTtMrcrGWXRx7ywZz0gHeSJkA3FCWBISkiNR4tLu9IpZjmEr-2IX7Lq58lNyXUPmTypEJEOs8W9dUqRcEFUn8rTKZT3f74qJXGPmj_T2SypQT03_9MxbXSicZn0NfGU1p1Xz7mHmdZMYF1UFzE3343OWsz2nxb2LOYbyjOIDBUCoT30CyG6nOdtaL3DEqPUOJjwNGKFFCNbykIFphS1hGcEcQJBh7u7LqSi07lB1c2105y90SEvjc6rl706qAsKRsJl1nyrtZBto4-Z5tc2rkfmoYzQJ2_KYwfHYeronT1RBOl48KStAFjDYcK2riHPCA.jpg" alt="" loading="lazy"> <p>Hadn’t noticed before, TikTok has an option to slap an “AI generated” label on. Pretty clear where the guys are headed.</p> <p><a href="https://t.me/chillhousetech/436?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> RunawayML dropped a new little model for generating video from, obviously, text. https://darkolorin.com/posts/435.html https://darkolorin.com/posts/435.html <p>RunawayML dropped a new little model for generating video from, obviously, text. Here’s an example. <br> <a href="https://runwayml.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://runwayml.com/</a></p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/435?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ll just leave this here https://darkolorin.com/posts/434.html https://darkolorin.com/posts/434.html Sat, 04 Nov 2023 21:53:49 GMT <p>I’ll just leave this here <br> <a href="https://pessimistsarchive.org/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://pessimistsarchive.org/</a></p> <p><a href="https://t.me/chillhousetech/434?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> UPD: added pics, editing and sending with no sound Waiting for the bot on… https://darkolorin.com/posts/432.html https://darkolorin.com/posts/432.html Sat, 04 Nov 2023 14:54:22 GMT <img src="https://cdn4.telesco.pe/file/SogJrHsbdSGbwHnZPA6hb0CBmBMnUz9TURUeFVhO0wJ0tIH2r-RfTM2aYKdqNvCzGVeqVMPcM_N859Zx5blHF5e0OQeghfd0ffSaAoPt72YwrIGsiqXOV1IOsOqxQtJsjh1K1oja2mno0vQ6eC4h8XIaBVgTnAwozsMBoK7WGhkOfF--TBoxzO2GRrALCgNzjdcXg8qAyGnaVZm4R3hbM-tz5mqFnR-d_vuydlJTjN07R9hvLhzIcjuXxLxfdjYzGLtndODnKdPCZyeBvdc7ysen6iY-KXBtXgQhhOx0VYkEs9jygnlbwswhXawYWSK-KIsXk1oIoMYgLwfJ8YjZBA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/RrNcIlCpGupiUKA6Ml3WLuyBzXpdikmK5l86T7mrdh-HWT8XPIXXZzfX-8kzRKEUQVEUEy9xMnCrN2h2kucF1psdhOhJmLiGghUnIqkH1RDmAojeWPibGr0AHyNYOQ22eSYt3y7L0E2_5_XVh1DOJCPSuU7EUAFY7TokIXHTyTUjoz7N6vkY3hvWRza65d7ZgaEQdmY-ZqmLP9b-poHb9YRXjfBYeIbOMyPYuDenGPL4fBjsWfriJpgRmvk73tGOdbSPhyzpJgFiX_KBMVxzYu5dsyXnP6VszWvYXoluYFuB6jjIbkf3NcGTpwi59cG2qn_5ltHwf0pmb0t0vp445Q.jpg" alt="" loading="lazy"> <p>&quot;&quot;&quot;<br> UPD: added pics, editing and sending with no sound<br> Waiting for the bot on Twitter too<br> <a href="https://x.com/elonmusk/status/1720645900471554146?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/elonmusk/status/1720645900471554146?s=46</a><br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/432?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Finally, someone did it https://darkolorin.com/posts/430.html https://darkolorin.com/posts/430.html Sat, 04 Nov 2023 14:47:49 GMT <img src="https://cdn4.telesco.pe/file/IYsG6kUof_TR1wvl3ZSsr92tIKReIHOtl0DAfN-jfyD9rKSHAcpVaP_xd8JWb1NYaj23XHKSu6rTF2uXaFYrn_l2DFs_G13VbZKWnwAArA45EGoTymboy00ZfkBPOnoyyEkgbgYwIiJY2Tg7qPf3QO77TczL4a4KLePyQu4khoxHCkE9E6NxxPJxcgxejLwX4SREGQ8vbq2JoJNQtDxgv7lY_PenQxYoCH3_15cqfsSr15BH9l579zT_WKovkyiqCB0rm0XLzMVbPd0WGhQBVTdFACHUg7XCYxmQ59bQg9PHjTjJkxFH18yQu7vYnU9wzEpuT2YPq9-CLMZtrXCP4w.jpg" alt="" loading="lazy"> <p>Finally, someone did it</p> <p><a href="https://t.me/chillhousetech/430?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Good thread and a great conclusion. https://darkolorin.com/posts/429.html https://darkolorin.com/posts/429.html Fri, 03 Nov 2023 14:46:38 GMT <img src="https://cdn4.telesco.pe/file/niPwTp4LJYKU6OdOiNLVFFScIEffX6wesGjxBIB_p2dzzZArbwDV_4JZKKdnHkuzwLPnYO69Yb8e8Bv1myB_f9iK9MWLl8B90cexr9eAJwgyNJW6kwnYEn1juNF6Gbr-DhOHcF8PgCTDcv0wTk9nlvnoOgV1xc9YovvlBUJJZgvEaQuAIyP0fwTefKup3CuYngy_IvevfaLh278n3RAl7GWOdXS1wiYBCg1udZT2e80iE6YfI8unkX8N1SZ5Agbj9W23mjBF1We66aztFU9yZt7qBz3i4SdoEZVLw-sknqQcpmtQoc4PEpZfxmTDHvqCRXyyaqLTgaeIJOzjBfm2-g.jpg" alt="" loading="lazy"> <p>Good thread and a great conclusion. Which, by the way, fits a lot of things overall. Food for thought. <br> <a href="https://x.com/halvarflake/status/1720015225036956090?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/halvarflake/status/1720015225036956090?s=46</a></p> <p><a href="https://t.me/chillhousetech/429?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, that AirChat app from Naval Ravikant kinda isn’t really taking off… https://darkolorin.com/posts/428.html https://darkolorin.com/posts/428.html Fri, 03 Nov 2023 03:12:36 GMT <p>By the way, that AirChat app from Naval Ravikant kinda isn’t really taking off <br> <a href="https://apps.apple.com/gb/app/airchat-better-conversations/id1469462351?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/gb/app/airchat-better-conversations/id1469462351</a> <br> But I guess they’ll pour some traffic into it soon enough</p> <p><a href="https://t.me/chillhousetech/428?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile Midjourney is stubbornly going where I once wanted to take Prisma. https://darkolorin.com/posts/426.html https://darkolorin.com/posts/426.html Fri, 03 Nov 2023 01:11:20 GMT <img src="https://cdn4.telesco.pe/file/i8mcrRiRECPez7DNBmhSCMZ5-2m-GpHF6N5ctDpManEdu5xonzXBmFlP3wxWcp0I-s0VnLL9Sv1urbkfsn-9V4Ncql8AHX_D7VtXWeHpkXZ1rny-gXcNWtrdpD5DsLpAjMZQWl-Rk5KXytWJnercqAmOX7QIHvhqoUKDI3pDj544cavCvO1eK4RXfb2FBHFyhS2te5vesOGVqoW4DpWQlh1mwlhWXL8ttdZjVqKPixR-QR_vYV7j_2k60dvXEvXMEjLkF0JSD61xDTVGFlc5EFbMfhendpNuZsvZ7niQqeczNn27b_5iBYrZNiNf5i_zL-X_xFm8WVWN42v0N37Nww.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/GJYJUgYlp5zqjkzvRcz2h6saVZ-cGkHEZHEyre01puNMOIF7wBCQCNGdoddw8qkrtz_uLWbKIxyfIOElo0T_s5LTL_oeHSYHpLZyVyFRoS82mM6zfEc3q-lsWeF0ZLJiljNJrjj9bppHME5DPHnDuTRTYmmlURKTs6HCxdPiYkSUVCHnHsg6t_ahWRWPyQcjack6wy4EORzKNuFdFKgzQRQd9Y65MOI51gTbe4AKmmcbleJzQ_I0HOPwdw7lGAnR-m8HJ-q9f7FMrAWQBTN7texLUCv9s0D4naCGXGp4C8OKkBhLTgGCKuONTY-WCq_eWlOXWY-VdKLK4wob9R66hQ.jpg" alt="" loading="lazy"> <p>Meanwhile Midjourney is stubbornly going where I once wanted to take Prisma. At some point, talking to the Apple guys, I realized that one of the biggest tasks is to make photo and video stylization kind of, well, autonomous. Like, you press a button and it just does it so you’ll like it (even now it’s basically there and it’s called the domain of computational photography). But there’s also processing and copying styles with context taken into account — that’s a big topic. Here’s an example.</p> <p><a href="https://t.me/chillhousetech/426?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I get sent all sorts of projects, links, and sites pretty often. https://darkolorin.com/posts/425.html https://darkolorin.com/posts/425.html Fri, 03 Nov 2023 01:01:21 GMT <p>I get sent all sorts of projects, links, and sites pretty often. And I stumble onto stuff myself. Sometimes there’s just nothing interesting there. But there’s an idea to start a “roast” column. I drop you guys, my dear friends, a link and you vote or even comment. How’s that sound?</p> <ul> <li>Let’s go</li> <li>Hell knows, depends on the project</li> <li>Noooo, better just news/opinions</li> </ul> <p><a href="https://t.me/chillhousetech/425?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The most loyal startups are promised support (c) ahahaha. https://darkolorin.com/posts/424.html https://darkolorin.com/posts/424.html Thu, 02 Nov 2023 18:46:45 GMT <img src="https://cdn4.telesco.pe/file/dgdgy1_U7lwOa_5NPReDLoX7vtEWpiSa_G8Ecfg4IUbQFwjjIX1BGk-ItBcGXnSxlFfGQfUgAOvAs8Rs51jIgDmTBBVf50JjHNO5yUpJf_gKQso7_oO9KPtq0jIQprGjf-oYjhsLZvsxtpKYz7eTHzyQZPbTpanivBJZ3TyXw_5Kwkpz3-61Z0AN_pRYBTEXNPpVwPStO2Qbldmk1yWvq3DB-kfJINfCHWzKIShzjYCs2IeHk7VUp1po3xq3kmUnzBmfhZQSQquS1r6c7lPRMvrA0Skph4FM12nvp62lfBaO-wv-6FjFGVcQ9vcdVtb3T7bQD4CyI4_8LCkyB4OvTQ.jpg" alt="" loading="lazy"> <p>The most loyal startups are promised support (c) ahahaha. Now *that’s* the pinnacle right there.</p> <p><a href="https://t.me/chillhousetech/424?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Another interesting social calendar. https://darkolorin.com/posts/423.html https://darkolorin.com/posts/423.html Thu, 02 Nov 2023 17:16:53 GMT <p>Another interesting social calendar. Grew up to top 10 in the social category. According to the founders, they hit 2M downloads and a mil, as I understand it, over roughly the last three months. <br> <a href="https://apps.apple.com/gb/app/howbout-social-calendar/id1477248221?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/gb/app/howbout-social-calendar/id1477248221</a></p> <p><a href="https://t.me/chillhousetech/423?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> People started posting info in a gif, so it’s apparently harder to put a deck… https://darkolorin.com/posts/422.html https://darkolorin.com/posts/422.html Thu, 02 Nov 2023 16:19:43 GMT <p>People started posting info in a gif, so it’s apparently harder to put a deck together or just grab the pics, but anyway I’ll also attach more pics in the comments just in case</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/422?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So when should you join a startup https://darkolorin.com/posts/421.html https://darkolorin.com/posts/421.html Thu, 02 Nov 2023 00:01:21 GMT <img src="https://cdn4.telesco.pe/file/h08j09G29I7oW4SobX34skNz-sgIx1ntKhPk_XWz_IOezvwUKgfjYcN1S7tyHmvSz5mp-RwGn3JXWB7gf_HqvNemYJPI_ozHb-o8fRe_VnPYWxeAeVYnoVuTBhthe7nl9CGFkBY0V60uoa-hAdeeApWmg-X9ygz0IkFLhA8bE6tR8m8D6GQhjqlSBKCkBKS7K64RVD1mz0DfLZdXVVnDsW9nMIByrIQXZDUz3scztSFVRmtQuOL7qxorjY6LMVgkUsVbUSblQR7NXjSLlMuscry7TD0iwkLHxUXTUNtvb-zuk1kvybltMFDIegoL5PIJ-YcF-nRoYTsyc9nnXHZuxw.jpg" alt="" loading="lazy"> <p>So when should you join a startup <br> <a href="https://www.joinprospect.com/blog/which-stage-startup?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.joinprospect.com/blog/which-stage-startup</a></p> <p><a href="https://t.me/chillhousetech/421?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Acquisitions have started trickling in a bit, and I think this deal isn’t… https://darkolorin.com/posts/420.html https://darkolorin.com/posts/420.html Wed, 01 Nov 2023 13:47:40 GMT <p>Acquisitions have started trickling in a bit, and I think this deal isn’t small, back in ’19 I remember the Series A was at a 200+ valuation <br> <a href="https://www.hubspot.com/company-news/welcoming-clearbit-to-hubspot?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.hubspot.com/company-news/welcoming-clearbit-to-hubspot</a></p> <p><a href="https://t.me/chillhousetech/420?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well there you go https://darkolorin.com/posts/419.html https://darkolorin.com/posts/419.html Tue, 31 Oct 2023 20:26:13 GMT <p>Well there you go <br> <a href="https://www.wsj.com/articles/wework-plans-to-file-for-bankruptcy-as-early-as-next-week-1fdcb6a5?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.wsj.com/articles/wework-plans-to-file-for-bankruptcy-as-early-as-next-week-1fdcb6a5</a></p> <p><a href="https://t.me/chillhousetech/419?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Man, the media has gotten so bad, and you’re still asking whether they’re… https://darkolorin.com/posts/418.html https://darkolorin.com/posts/418.html Mon, 30 Oct 2023 21:54:10 GMT <img src="https://cdn4.telesco.pe/file/AbYFxdVH4Twub3jIulHJ9fo6O8fMHZCyr4IE49eqUsfOJZIpgKUYwosFLVF-cYk_iqhRVN06kLlGiQWEwf5gYqHcZu7hwXzAU6dLPF6S18ZXmSlEGdSjEyQj2qESV5RinRLEmxzZRwpTEjsVHS2YOu5fx9TV4soV-Jdn1qLn7mUrhSJGRnRJRIDSyDzUrV4l7t-6PenS8GawgudikSa8ml5lmfrywVdLL8wks_utVhrE4P_B_GefEKuUQ6OI-sO5E9OV6kWh0nQE_ish9miKRFwh3zw90qTM9XXre9h0Zv2BL6XgNxQmS4zN7Wg1QFOWYa7Mw00lJwiqkmH2YLhO6w.jpg" alt="" loading="lazy"> <p>Man, the media has gotten so bad, and you’re still asking whether they’re independent. Here’s your “independent.”</p> <p><a href="https://t.me/chillhousetech/418?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Heard of these guys? https://darkolorin.com/posts/417.html https://darkolorin.com/posts/417.html Mon, 30 Oct 2023 19:32:40 GMT <p>Heard of these guys? <br> <a href="https://www.medialab.la/brands?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.medialab.la/brands</a> <br> The dudes, so to speak, buy up whatever didn’t sell, for the traffic.</p> <p><a href="https://t.me/chillhousetech/417?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> How often do you use ChatGPT or something similar? https://darkolorin.com/posts/416.html https://darkolorin.com/posts/416.html Mon, 30 Oct 2023 18:34:00 GMT <p>&quot;&quot;&quot;<br> How often do you use ChatGPT or something similar?</p> <p>- Daily<br> - Once every 2–3 days<br> - Once a week<br> - Once a month<br> - Very rarely, more like in spurts<br> - Never<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/416?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Need more gold P.S. https://darkolorin.com/posts/415.html https://darkolorin.com/posts/415.html Mon, 30 Oct 2023 17:27:02 GMT <p>Need more gold <br> <a href="https://www.theverge.com/2023/10/30/23938283/facebook-instagram-ad-free-subscription-eu?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/2023/10/30/23938283/facebook-instagram-ad-free-subscription-eu</a> <br> P.S. Regulations, and all that</p> <p><a href="https://t.me/chillhousetech/415?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Interesting feeling from subscription movie theaters. https://darkolorin.com/posts/414.html https://darkolorin.com/posts/414.html Mon, 30 Oct 2023 12:30:38 GMT <p>Interesting feeling from subscription movie theaters. Basically cable TV, where you’re like hooking up each channel for cash. Nothing changes. <br> Monthly cost - US streaming services (ad free) <br> Hulu: $17.99 <br> HBO Max: $15.99 <br> Netflix: $15.49 <br> Disney+: $13.99 <br> YouTube Premium: $13.99 <br> Peacock: $11.99 <br> Paramount+: $11.99 <br> ESPN+: $10.99 <br> AppleTV+: $9.99 <br> Starz: $9.99 <br> Prime Video: $8.99 <br> Discovery+: $8.99 <br> Total: $150</p> <p><a href="https://t.me/chillhousetech/414?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> From the author of that earlier popular intro/presentation post by Sam Lessin… https://darkolorin.com/posts/413.html https://darkolorin.com/posts/413.html Mon, 30 Oct 2023 12:24:59 GMT <p>From the author of that earlier popular intro/presentation post by Sam Lessin about free speech and all that. Interesting. <br> <a href="https://x.com/lessin/status/1718810964060156243?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/lessin/status/1718810964060156243?s=46</a> <br> TLDR: elections are coming up, TikTok needs to be banned because no reason it should be.</p> <p><a href="https://t.me/chillhousetech/413?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There’s a good answer from Steve Jobs at a 1997 conf that really resonates… https://darkolorin.com/posts/412.html https://darkolorin.com/posts/412.html Sun, 29 Oct 2023 23:47:28 GMT <p>There’s a good answer from Steve Jobs at a 1997 conf that really resonates with what I’m seeing in the AI world right now. Nobody gives a shit about use case &amp; UX. The only thing that matters is showing what has become possible. <br> &quot;I&#x27;ve always found is that you&#x27;ve got to start with the customer experience and work backwards to the technology you can&#x27;t start with the technology and try to figure out where you&#x27;re gonna try to sell it&quot; (c) <br> Full answer here <br> <a href="https://www.youtube.com/watch?v=oeqPrUmVz-o&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=oeqPrUmVz-o</a></p> <p><a href="https://t.me/chillhousetech/412?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Put together for you one of my favorite presentations by Sarah Tavel from back… https://darkolorin.com/posts/411.html https://darkolorin.com/posts/411.html Sun, 29 Oct 2023 00:29:43 GMT <p>Put together for you one of my favorite presentations by Sarah Tavel from back when she was still at Greylock, and now at Benchmark. On engagement, usage loops, and why cohorts. Must read. <br> P.S. if you’re gonna forward the file — ask them to subscribe</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/411?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A brief sketch about the role of a CEO in a company https://darkolorin.com/posts/410.html https://darkolorin.com/posts/410.html Fri, 27 Oct 2023 22:28:49 GMT <img src="https://cdn4.telesco.pe/file/QjuJvFVJjwcyENevuYFR1yHCQjeF1FXKsDSBvX0dSW1jCzMShcm_e7LeCKduO_QJyTY9HQrqCW6UOEcewTv2gAuCP1agdKrW4896nvB-07EVOO9WbkpuG83kpPu4YmckiSYSOahtlutDdnthCgA3UdNrs5P79_b52J37osboRxRmnoudV43zpVEdIv6VpRNcI29fxKfwxt6sa_8c5sKj8W6akApgpRMfx1cKhTh25huqdnYb24n_wToeq85P5OmbLJHxQDY5jtMCDPDorQaUvAigGjqIbCxa7J-xKqem3e3bFFaL0Ynu8rn0JQG3JOQMeCCr6RIes50-h1qjK9HaJA.jpg" alt="" loading="lazy"> <p>A brief sketch about the role of a CEO in a company</p> <p><a href="https://t.me/chillhousetech/410?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, if anyone’s into digging through various papers, I use HF for… https://darkolorin.com/posts/409.html https://darkolorin.com/posts/409.html Fri, 27 Oct 2023 16:48:35 GMT <p>By the way, if anyone’s into digging through various papers, I use HF for myself <br> <a href="https://huggingface.co/papers?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://huggingface.co/papers</a> <br> Overall, sometimes they even have an implementation posted, and it says how to run it, lists the authors and so on.</p> <p><a href="https://t.me/chillhousetech/409?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well Yandex put out a good report, by the way. https://darkolorin.com/posts/408.html https://darkolorin.com/posts/408.html Fri, 27 Oct 2023 16:41:10 GMT <p>Well Yandex put out a good report, by the way. Predictably, I think the backbone of all the growth is search and the portal. Trust me, without search and internal distribution the other services are gonna feel not so great. Report’s in the comments.</p> <p><a href="https://t.me/chillhousetech/408?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Document https://darkolorin.com/posts/407.html https://darkolorin.com/posts/407.html Thu, 26 Oct 2023 22:20:55 GMT <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/407?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Don’t think they’ll die anytime soon:) especially WhatsApp’s growth since it… https://darkolorin.com/posts/404.html https://darkolorin.com/posts/404.html Thu, 26 Oct 2023 19:57:08 GMT <img src="https://cdn4.telesco.pe/file/M6qN3UxOzXgg1OCJuJ10M2_h7a_ineshfBBqbBYlwjBOQarwzw-3hPjGHHaLBNJs27b2LN4kq0fbmZhGE_ukhWbJ_oZXNAVFEd-I4F_4eZ6lNuHSxOpA0PbSc65DlS1tg5DDvpyhODgwI2n3fO0zOFBRMNppUMcsCXGJ_05tHMSDR-qW03hIIQrhjW2B0KBiCIwkqzXBNmmioEJLMvf0I1Khl9e36ujjqVYseamnE27Bdqo69YlFj7GilfR761ff2FQ80_0NEFkDn5g-Rs7rNgGmql0WAef6mHvsSTc36oCfL5F4XNIGpv50t3Zh1eS-jKGmSrJPXj9LeLwKNBpW0w.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/Xd_3TrdSbA7DfzzIevPhF3NxDhmdYDqJBXcEE4NOHTwH2ocPTFl07iXrXkt2bh-94GlnHG5TlswZIcdCo2yH0e-TjODkqVhLSRwTbFSOtz3ifBHgtaKBx5pSuO1rlII4CURAcBzxwVLVW0O8aVoTxKmS8_5XqFgM5lTO1h_v21KkKWNy9Qdr7kzv0z6U0RLz3SGbicCiMC1AfkPf2zK4EHwhJfE2u8Jq-ruHE2se16lGXiacgYffVjrpkWCW2VzPsvPOxikuuLdB1wVl6a70-3cYFNCxCsdfQxnu09e9quUtV3VzH1GGIY7_T_-J7d9c_rbDiyBr4FEWSn2R86OLNA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/q918mNHe52FAvjrmV9biwUEiWzh3MLd3fqCe-o8SvVFjsopAcgzE0IAyuvCotozkghK0VUUHvKV7E9sCyqHEKaQypwaZUeZmVLkHn0qXd_OHOV8atrqYthg_IuhnWUAcGLRfDtbap7NSRZQ7gC1qqjhhY2NCvwlVFYJGLi6ob-Uv4fd5MceoLV53nRlILYb20KBg_qGx5vCCmVErLt6Z9ZKO6Qm6KHibo2d2r-ju8B8xAOJfGMz9o5fN6epSDYXY39qAnah7ucLytn-9DOObgM99yeBRQEWb-3tqGCAwC8PGHa_eNegiyElXwGV1IMC9ndIv2qgcE8pD8qw-dffuWQ.jpg" alt="" loading="lazy"> <p>Don’t think they’ll die anytime soon :) especially WhatsApp’s growth since it was bought in 2014 is impressive</p> <p><a href="https://t.me/chillhousetech/404?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, YouTube’s starting to grow up, makes a lot of money and ahead of… https://darkolorin.com/posts/403.html https://darkolorin.com/posts/403.html Thu, 26 Oct 2023 13:49:24 GMT <p>Well, YouTube’s starting to grow up, makes a lot of money and ahead of expectations. Pity it’s not clear how much the subscription brings in. <br> <a href="https://variety.com/2023/digital/news/youtube-q3-2023-alphabet-earnings-1235766877/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://variety.com/2023/digital/news/youtube-q3-2023-alphabet-earnings-1235766877/</a></p> <p><a href="https://t.me/chillhousetech/403?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If not for this line I think they wouldn’t have raised it: Our long-term… https://darkolorin.com/posts/402.html https://darkolorin.com/posts/402.html Wed, 25 Oct 2023 17:57:36 GMT <p>If not for this line I think they wouldn’t have raised it: Our long-term vision is clear: one engineer, one artist, one designer, multiple AIs — all culminating in a singular game experience.<br> <a href="https://venturebeat.com/games/harmony-games-raises-3m-for-mobile-game-tiles-delight/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://venturebeat.com/games/harmony-games-raises-3m-for-mobile-game-tiles-delight/</a></p> <p><a href="https://t.me/chillhousetech/402?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Love this blog This time it’s about corporate strategy of parallel bets. https://darkolorin.com/posts/401.html https://darkolorin.com/posts/401.html Wed, 25 Oct 2023 15:02:35 GMT <img src="https://cdn4.telesco.pe/file/De7VQ3OA5_7_UYY1YwR6z-bRYMJVUynyTkmkzWnzWMF8QFNGZbHgGzOeYpTR0Z27VflZxaarM_sftsMXUxCSxWHV1OWhTdsOF9BhbQqDnYaE6vxZcvokj5-b3caVXjYNk8NgRzhpOaGqyrMrBj1B0Db2VcL9JU-95DNNoZlFbeRvbIVjCoaa9Q4M_voea37ZtklD5hKne3gfQT0HZYqY0gvv3fQrWmiL127HQW5IH87VULM_RuVvr6Dj73iHDYiRNihh1IDvKE22gH7eMDJdgAJeBViVsPmEkGb4Pc7kTdK10PYx8zhnOWm-hx2BKxbX3bksqkSw_c7DdkbPlsA2aw.jpg" alt="" loading="lazy"> <p>Love this blog <br> <a href="https://www.matthewball.vc/all/parallelbets?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.matthewball.vc/all/parallelbets</a> <br> This time it’s about corporate strategy of parallel bets. Recommended. And a meme as a bonus.</p> <p><a href="https://t.me/chillhousetech/401?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Pretty interesting leak. https://darkolorin.com/posts/400.html https://darkolorin.com/posts/400.html Wed, 25 Oct 2023 12:46:44 GMT <p>Pretty interesting leak. Google Stubbs. Some kind of no-code builder for AI apps. Reminds me of scripts for Sheets or Visual Basic for Windows. There’s a pretty clear strategy behind it. <br> <a href="https://medium.com/@bedros-p/gemini-is-coming-to-makersuite-so-are-stubbs-32248f3924aa?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://medium.com/@bedros-p/gemini-is-coming-to-makersuite-so-are-stubbs-32248f3924aa</a></p> <p><a href="https://t.me/chillhousetech/400?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Mark Pincus, Zynga’s founder, is betting on Snap. https://darkolorin.com/posts/399.html https://darkolorin.com/posts/399.html Wed, 25 Oct 2023 12:42:38 GMT <img src="https://cdn4.telesco.pe/file/TfecUiDa-JcVtkye688VqVvdYW_yTU-Jr5YDp9YVKZJH28oq64jOTE-aedrc9J7yJ85SdmU9eUFtmO3Cp9CuTbo0aVTUgzh2-__ODR58J_BNq3GRHBXP1Lejm41xpSOLlr1CAlx6FzH2gTGZE0jOC6zp-WQD7FAsxRxxQi5mzrKpbXN8UW0edQLW-Pb6_ideFQ_OqesTi1L3PMV5oBppcuBEo9nfB5WgwXbN5JdSyFmrOjGMwTI-cOBfP0ePvBrPIunAH5GX2nuWzBKiXIWfcrtSMvXpM_5YS5O4JcXqtNUWLUxKdikml0sMDx504cKRYkufKH7j6Uy9Or1SZOrYxw.jpg" alt="" loading="lazy"> <p>Mark Pincus, Zynga’s founder, is betting on Snap. I’m betting the opposite. 1/5 over five years and death in the form of M&amp;A. Your bets?</p> <p><a href="https://t.me/chillhousetech/399?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> TikTok launched its own shops in the US https://darkolorin.com/posts/398.html https://darkolorin.com/posts/398.html Tue, 24 Oct 2023 18:53:55 GMT <img src="https://cdn4.telesco.pe/file/R8mjr6n_KCOFJhbZyQQK2sokXJ-mwTYcgg7egSfPxXXV7F4wGDAjZy9xHivMnoE5UNWAg8P9lurfWldBk4sF4BwEJ5LSXWx6EFgYaFnQfaVt0MDtwIhJZM0Btl58tb-bgsjY7-RbDkRBjsWGEOfmxMLeJDntYfn9G_gVD3hBstyHHjFYk19XM5thS0oiljHjZ7HxwXKt6d1mVmhvKs7B4g0_Gtk3XpIdpFYyN-VzrftN26kfXDCP8qCl-ayPHau4mrhMdsNUWFJtWFacmSTjiL_cBWhlDDh0rLoEaVFyp7t8J770O2HyJvL2HcZHM1DMA2AMWDqvgYkHt4LJv0byrw.jpg" alt="" loading="lazy"> <p>TikTok launched its own shops in the US</p> <p><a href="https://t.me/chillhousetech/398?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A quick bench for valuing a company: Annual revenue x 10 x YoY growth = how… https://darkolorin.com/posts/397.html https://darkolorin.com/posts/397.html Tue, 24 Oct 2023 12:30:31 GMT <p>A quick bench for valuing a company:<br> Annual revenue x 10 x YoY growth = how VCs would value it<br> Annual profit x 5 + assets - debts = how independent appraisers would value it (409a)<br> Annual profit x 10 + assets - debts = how PE would value it<br> Annual profit x 30 = how the public would value it (if the stock were fully liquid)<br> This is just a rough sketch and obviously depends on the industry and the market (geo). But it does a decent job of forming a ballpark understanding.</p> <p><a href="https://t.me/chillhousetech/397?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Zuck, though, says stuff pretty close to what I believe in. https://darkolorin.com/posts/396.html https://darkolorin.com/posts/396.html <p>Zuck, though, says stuff pretty close to what I believe in. Your ability to build something new and put up with failures and mistakes is what determines your success in the long run. <br> Even though he himself said pretty early on that the best teacher is “immediate success.” Haha. Life changes.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/396?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Gonna throw in my two cents on analytics. https://darkolorin.com/posts/395.html https://darkolorin.com/posts/395.html Tue, 24 Oct 2023 10:06:12 GMT <p>Gonna throw in my two cents on analytics. I think that for early stages, you generally don’t need any analytics at all, and no deep user study either. Like, not mandatory. <br> Analytics, especially product analytics, is about optimizations. And I’m in the camp that you’re basically not gonna find product-market fit through optimizations. <br> You can understand that you’re fine by looking at your spot in the store, the basic metrics the store gives you now, DAU/MAU. Everything else is basically details. You can tell what’s going on by looking at UGC. At launch I most often look at one number, in GA or other systems it’s “users online.” If that number is growing and not dropping, then everything’s ok. <br> This approach just saves a ton of time. Criticism in the comments.</p> <p><a href="https://t.me/chillhousetech/395?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One of the best presentation designs. https://darkolorin.com/posts/394.html https://darkolorin.com/posts/394.html Tue, 24 Oct 2023 01:18:21 GMT <img src="https://cdn4.telesco.pe/file/Zr_fwjNGbl0w-hBBhkyto2wNGqC3_4Q0MYn1eHkaBDqtyUZ6MrubFViUwGZhkU5D5uLQwLBgTJyugYZhem0raSJ3w61ScSWLujBBA6Rg3H4z0rdrdxQEDNOuyJ3SgKf7D-X7ue0iNte7XzDmsojJdXfc2trS354mFZkjVLEJcXNiZD28fbTIGX-edeCAqHG2rz0Xr3LCBZuxPf_lba5cZfxEZ8cXuw91mffDDJ-_oUZpK0lItopbw-1-cEPtiGEw43Tk6buBo7zlA7LiRTmtgiKI7ttN48TwWqHPtpchULSXq2xWAEReVdWgYB4fr9iyZw-AHPIcjiDPJ6t565-54w.jpg" alt="" loading="lazy"> <p>One of the best presentation designs. This slide is gorgeous.</p> <p><a href="https://t.me/chillhousetech/394?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Huh https://darkolorin.com/posts/393.html https://darkolorin.com/posts/393.html Mon, 23 Oct 2023 21:55:56 GMT <img src="https://cdn4.telesco.pe/file/TQUjDuHQSiAGevk1Ov-V9O8MwDmpkHuVwln5syNzOhKEoaHK4nMXqgrZJbE5qHXQpe_Vh-dvRAdExu1HZcbOCMntXNMKzs2hF9p6znne0QTvAE2HrebLpGPWGAwmdKJfONI9oPa5Wnt8Om93Y6Fc4PsfUdoiXF7jDW3cE9Nk8rQbGc6-0DieHVE4sAL2pCmRzVhuPDZUk4wTxayL4w4d0hLsZK862FU9puQXy6i1VZ3O03mQdZVtLAQEXLcHUAO2zj9CgyhWIX5m8cSqt9dPivv1BqrZKTP8-zwsHEyyd-rodBgLnSZTdGf2VtSDvMv_Ycb8JJX3xe4vD8S6nigLxw.jpg" alt="" loading="lazy"> <p>Huh</p> <p><a href="https://t.me/chillhousetech/393?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Recently I had the honor of talking to Emmet Shear, aka the co-founder of… https://darkolorin.com/posts/392.html https://darkolorin.com/posts/392.html Mon, 23 Oct 2023 18:36:55 GMT <p>Recently I had the honor of talking to Emmet Shear, aka the co-founder of Twitch and later its CEO already under Amazon. We were discussing the following question: what do you think needs fixing, or what’s more broken.</p> <ul> <li>Money is being invested into shitty projects and we need to stop letting it get put there</li> <li>There aren’t enough good projects</li> <li>Everything is fine, there are plenty of projects and the money’s going where it should</li> </ul> <p><a href="https://t.me/chillhousetech/392?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Love that Twitter account that posts internal correspondence between company… https://darkolorin.com/posts/384.html https://darkolorin.com/posts/384.html Mon, 23 Oct 2023 15:41:28 GMT <img src="https://cdn4.telesco.pe/file/GZAQXq1xPGkz7ORh8sFOGs8VZTo-7_T5uDw_IM3NCQ3PTTHZ0bUPjEjkhkr_U_J9Umd7nC7qUwX0Kt0_7Y6O8Wqb444q790nkPJPZV_PDnn_RSDCeTgHNMfj2ni6ws0tQZ1gUiJNT_iIKmfd8A1VZpG0fp554J7JTWx4weVHhrUg4sflCK-CXfwpfbjdnio0t88CwdoAR22U2PE1fwIu2LlIwUwaQ9xKZREan80tcW9ZOLa1c3D576a3QMWq_by7P1r_Ra0ptLQ5_o_Zm1iVJk9dKlgVL9A2V8xURq297Ze5BVSeRZZzXkz9R_aedDyeYGIKn3t19V56NG1u3w780w.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/DgvVAQkcsrgqwRWO0cA27KexfEeDkQ-QWQk3MZN4hgGfI0FqH1x17sYw5m5gjvhECdPjqNf8eOM0Bq2rIVkE-tg7GnGI_5SQJf4BMSErB5s6371e1tBmHfmB5Ob_cV0jBWLjb5RrBWiVXXazo94JiSA2g6CXwtRrf3pMTtSwKibZ_UaUjBjPe8ZcFMTuit2KEiz8iHnYbQ7AmQgLvods4yMK6kyjXrqoNpQS1XQUcssEbzxgu4HmO6rR-jm2KwprrSgiPVXYnMARG0qOel2VxrjEUIpbTMluFVmGr_zsH1Og4O2kkG5jaQnf46gMdO860LEGpfYxQslT_xVIkHYrcA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/j8uCTJitjHZrY-sS-ZcOkABJkJuNxSgPeopWmQYIgp8UN9h8621g8e4V3LRc1owyzQMt2lUpt2rzoIQSJ6bMIYIb8cj-O15FUgqXi2UKbgtbLL6uTnKUHMhXpFng_qZyrPVPFMIbRGlEVS_3t698w2H0sZvwH_XiaYigTlazGD-fSzlYd-PjlPJsLCmJ3oIYmYBSvLejy8zy-N7VPVdMUasI48thd2BqztX1lnu1v1fdNIrWFfAD-WqkRob2g_vcz5fBCb5DvfeWKHGoEFeB8mekAV8KtrSZoPIjY5f42ExKT-gnHCi_6lxO335Miai3QQ7TEGFsWcn6ybOCeabsiQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/rgxuMz6CcwsY4FNHMqxB3ckvKTeqYh5UPuTxsNPp7l5-JXYIHfNKHeRFD6kY7sfhGbq1KXSFHF6ds33fdsQNdU9Wwe0eYz2cdh5FyfUIkFynph-bos8iQEU4C7pC9wdhfpggycyzuHXnt1pVXQoR5xuYNZTahfsF-WPmnXK3cRzPpDfjtCHdz4exSKMJQifgAy5s23RpmWYTpXedrzAimQQp5deCe75rPRoU6sEXlUtsgSgudUx8vsBAGnfporR3VO3VaQoR8IHFIGdTktJpwsaRzdiIjfE_yB0OtwM2i6Jp09BhL1U5W1Cil-kWHdsMdKDHk4LfANk-rHRrq4gV0Q.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/VB65xahDr8W7-hDx1QTKQ0Nh2i5174MupDbzndr2SnemAqD-Sx-DbKNIvL7OhqAIiwUaQqQ04cA-xOWQVQmOAzETOdGQvBkZxtcLgany1hZ2DDn4grLvtAutgTiRh9UySBto2xWRWobXLy4uGfBZd_M1jyrojB9wHR3DXEWlDBXdEBlVLaV8_88c7K_nQWWCfTgBF0UlsnvgBdaWvbIVa-kox9yoNLVneksr1QOIZiqMPO48HhhLZ96k-tFtFqckL6XThbjZ5FYcC6V4dRfNRQyZxU96MRFpdLMsr6-UJSyyZS5Mqobv0TUtKYtaKNMMYUqfqy9HM0je5ohVzg11iw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/azNKJzvcO5OxMCfn_t5K77m6YK-nhnZW99nO5ZDoqEk7_2PnMwz3ffAnAkWgqZpU6RGEfiBkWCR2UC3hWmE5067B0BSQbo4T8VzKZ-A7CNFxR4VSHZMemU25kGwI3sa2sY54XJode80M_ekDnesghxB0v_WjHtNlte7TAGUpjF4xaUF6J8Kb9D6fFU-3JMiqabTnzl9OlzmlvyA3GVqP9NkeOLVKFokZxPl3e5QiXQSXCNAxuYV3dM4XJHpc6O73Ls6_Pj5r-brP06mPBgrWyxOzOYBWz_yzOoaFYs8K9l_dD-sGwxPZRdLVSCD9LVyEkxxbrW7MjJcj1EGcKzZ3Jw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/K3ZqpIX3l1xqf5MllULTo8_BrSU6lMPWXNBeRlKVwvmq0PGzxeJOYFdqXcd725lyD5Ex0_ouqvkxCL5Bkqx_r--Hfh7Eou9e6B4O4gkn6q2ro4rnMolS5aljE9L4-05wzhCY42SD3b38KOBxjTJHuiKEeIWvQaXxoJLth-egpGRgO-t20ynNuqE7bFS9U5c7_iL3RzOy2cPqPxHfUq4EOnvk_1FB5sPSADU55uKJ0m-3RhCq3VCGfeMgq0cJ5XwxXEuxvC8uJZq7WRBB73unvDgcWdkqFMFBNwQmMhr8-GPyhuQhlDUFmi0TZkDX_dE6xJGaTcL_KtrwwDSCa8Hskg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/eZ41AGKE9Jh0zEo73ScPW73Xiuh6UzKdvedIOIRZ55gqwL1z_xEp0nKQc5RcabooYOxzRBjGA4vaifhsdYeyREklpDeEUQQhOFIJWPwxeul_ro2CXE0Ic3MvC0i4HDUsX7qO5Wju250QkmEd9KybPq74UiUPI6DOlxizsBOvR6hKBlQyNpyZm34pG4kT7NvU7bGkJmJwMkJ7xC4n0_xWAQIIPf32BUtP_rG8OBfx5U8MUuLjqT4yu4t_UR7l9FNemJ8gfypWuedOz5qgLnG4RJAToQk2ELAK63K_V9ozvB7ED_oDuLSchc-FyjAewOKpfQBd5JG_b9ksDDfjBDzozg.jpg" alt="" loading="lazy"> <p>Love that Twitter account that posts internal correspondence between company leaders. Obviously the source is litigation. But sometimes it’s priceless lessons, like on pitching and valuation. Substantiate your $20B valuation. <br> Recommend. <br> <a href="https://x.com/techemails?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/techemails?s=21</a></p> <p><a href="https://t.me/chillhousetech/384?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Weekend question: what do you think, is he right? https://darkolorin.com/posts/383.html https://darkolorin.com/posts/383.html <p>Weekend question: what do you think, is he right? Was he right?</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/383?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Just a modest tech week event, that’s all. https://darkolorin.com/posts/382.html https://darkolorin.com/posts/382.html Fri, 20 Oct 2023 10:28:52 GMT <img src="https://cdn4.telesco.pe/file/dHYlIaUSk_ptoSUYqL9-6tAzji0F_L6D98Rf0BTE3r8VKLXLasPjm5xq2BtOzSCHdbiV70sKxTUlLUytCqOgb6qs0cvG2kLMn7CBoD3QucOob7OMhEZMBFEaUFMoeHGHABr0n_mzkNQkVRFn02Tdva9vldOv1npqCENJuzzC5njx-sxv8lTYGCj3YJW2qmtxXBRi4tIBDG7UOyWPSg6znvSXnHAxM1JAXo-GGjYChKQrLa0FScL0lhJwOkh5iJT8hW-WsLuSTcjeWdpxV2chCkPERJ96Ms7r2f8JXRBbGZKzP6iTGRW9ZnyDsaomqc7PgbFUfylDeu8QYfbQXLxd3A.jpg" alt="" loading="lazy"> <p>Just a modest tech week event, that’s all. You know, in a work vibe, talk business.</p> <p><a href="https://t.me/chillhousetech/382?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Yandex Search’s share in Europe nearly doubled over the year, from 1.8% to… https://darkolorin.com/posts/381.html https://darkolorin.com/posts/381.html Thu, 19 Oct 2023 15:48:10 GMT <img src="https://cdn4.telesco.pe/file/BWblqFeYMyouQTUL7B88eZ1JE_ReC08INZeaRdSBNSmNuPG_5gfkY44pOXKOwR7n6wxT1bCWvBglVJPJ2eRLafnYqgDZXYPARepTOzoN5uTQBvvNF0OuuzFz4kWi3i9fYcxgPKdmn32e73lNMSWq038PCP-O0H8FG5Is-w3webk7WemltwqtkdRWgdQzzIBq1niAFrPNV1aw-tEvDz0zBiW7xT2Wv1QHfODn4LHZitKP3EbvLnEr_zvpUdn2kILjG0z3jhAKXv51BM09aPLg-9t8Haq2wC_jDl3GvDDXNTNR4DW844RqDr27C9YdHmhqMOeV9qsoXJxEejrN8TNgQA.jpg" alt="" loading="lazy"> <p>Yandex Search’s share in Europe nearly doubled over the year, from 1.8% to 3.4% <br> But those antimonopoly “laws” are doing their job</p> <p><a href="https://t.me/chillhousetech/381?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Specifically for you I pulled Sam Lessin’s deck out of DocSend and stitched it… https://darkolorin.com/posts/380.html https://darkolorin.com/posts/380.html Thu, 19 Oct 2023 00:18:48 GMT <p>Specifically for you I pulled Sam Lessin’s deck out of DocSend and stitched it into a single PDF. Very very similar to my thoughts, TL;DR <br> - invest at an early stage and in companies that want to become billion-dollar, but not more <br> - startup factories a la a16z, pumping via secondaries and trying to “juice” things up to 10B are doomed, the model won’t survive and won’t produce returns, all those funds are living off their management fees <br> - new 100B companies are possible, but most likely it’s not AI, because AI will only make existing players stronger, and the only scenario for backers of those projects is exits, there will be very few of them btw</p> <p>P.S. if you’re gonna forward the file — ask them to subscribe kekw</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/380?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For anyone who’s interested, a super fast little model that does… https://darkolorin.com/posts/379.html https://darkolorin.com/posts/379.html Wed, 18 Oct 2023 23:38:58 GMT <p>For anyone who’s interested, a super fast little model that does text-to-speech and also accents, and clones a voice. Top-tier. <br> <a href="https://play.ht/playground/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://play.ht/playground/</a></p> <p><a href="https://t.me/chillhousetech/379?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Rich people cry too https://darkolorin.com/posts/377.html https://darkolorin.com/posts/377.html Tue, 17 Oct 2023 20:23:54 GMT <p>Rich people cry too <br> <a href="https://collabfund.com/blog/a-few-laws-of-getting-rich/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://collabfund.com/blog/a-few-laws-of-getting-rich/</a></p> <p><a href="https://t.me/chillhousetech/377?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A bit about the power of an intro and the importance of the essence of what… https://darkolorin.com/posts/374.html https://darkolorin.com/posts/374.html Tue, 17 Oct 2023 20:12:41 GMT <img src="https://cdn4.telesco.pe/file/SJnWLsy9RiY66F2xJHAnAH6EaOKXD7YqeqQd80S-JfVHdauipMTkYsFSKV_ZV3wKDaKS7lWdD5-LD-Zjo2mxZEZlT5SL87RFk8S3rmY6nMqJjyf-YbDRPQbnI0tRxWSnNj2cN4ZH3Dx2X7GZANKJNcLrIO6_oMcxbQjRbCzkyb6vocrR-wPVZXNalfi5h0LDvZ9xnzhEVWTeGt1ayJxU_0qWBCm5qcZhPhFUgsgixtbyN6arfQfgHvHWch-kw1kp5GXiNK_bocl0oYxcIcckSbj93DSf1J_eJj14Z43nVlVpgeMLAhEHXrTpSagSqEd6zQ4VKZGYbnkinnWZOMOUiA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/mn8jpqdvyqNljjKyhJdDtaYnddwdMf7lJEoDWaxKbFZ2mM9dPuElXTm7uBYXGFSbisO3uuOYOUFJS4mNO_GBfNEhdPf6Byq9o4H6k1vnlMNtJRRtiW_nrySdcHVX05HGZgPWQ5DqVNzO4D6H1jkx_0oqHMYD5nKDHzNtResX42q_pgJFGuB-VYQytpjIMF5AFbsoLLLRlWoKGOVmc6AKK3fm7WKKN15rpC7N74d3J7F_ezU_3_Kun5Fi_U1X_8GxUaURl3451apee9S_GT7mHzR5D67b8ATWy1EsXxKhgrsX-sMULqukdKeSrNb-7tYm7dCoY-V8VFZ4cd2KKqFKUQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/lCRx3GpnYqB8j92-nbhovSqG1PHI4nkeiJu0dsKrjFmPejFXOQA6emwFwltg6j3RvXOQHlG5thcUMtKXrpflUaZcJaUa0934cHJe_7UFum4RexDfbP9_5cm0e7SfkK2SiuWcSvX242-6OionDvNDBdHbzmA1OpRwYSUF8Ush50Q3REu68iBQDs8WjWeyUomMx0yleYO9A-apg_FmcwxSgclSHhvnaSRIG9kHWQHf-r4zvLnTvMhikGAIA79Bcsh8k5iywudD8TDpyjzbK5NHpYp-KK7R7WuYH7U0_hHmueBVb1uXvn5DEGgp7ylg3oHWqkT1f13_E8So570RSHbyvA.jpg" alt="" loading="lazy"> <p>A bit about the power of an intro and the importance of the essence of what you’re doing (it’s usually around zero). Sam Bankman-Fried FTX memo.</p> <p><a href="https://t.me/chillhousetech/374?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Waiting for them to put out a compilation and call it the Bible or what, is… https://darkolorin.com/posts/373.html https://darkolorin.com/posts/373.html Mon, 16 Oct 2023 22:15:38 GMT <p>Waiting for them to put out a compilation and call it the Bible or what, is there a better title?<br> <a href="https://a16z.com/the-techno-optimist-manifesto/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://a16z.com/the-techno-optimist-manifesto/</a></p> <p><a href="https://t.me/chillhousetech/373?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In the US Congress in 1993 they’re, dead serious, discussing how Mortal Kombat… https://darkolorin.com/posts/372.html https://darkolorin.com/posts/372.html <p>In the US Congress in 1993 they’re, dead serious, discussing how Mortal Kombat can cause waves of violence and all that. 1993. Like, for real, this was an important topic.</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/372?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/368.html https://darkolorin.com/posts/368.html Mon, 16 Oct 2023 21:37:38 GMT <img src="https://cdn4.telesco.pe/file/sOv679MUjZeTMirK-2PAUmfQEmM_3knPVJEAN5CezUXWm6xgw-smXadlcNtARhjdGBSgok_ct5oLdwp18HWO-ERU0Ax2ivWSGB64aFZXUaxL7KkFslaC0nO76DjMjkgs6gWFpzbbA3nV_Fo5bSkUq7OdVzAxj63g6edOWuTQnJbjmLFLgMvex1wDhiu2ZJx4_n6eejLNtqCiQeKLcwx9N4lUB4ZAGCW-6kQKEeRUGqjekCCaj6geip8lMq-FT5200R3rsvq2EG12TUBH1l2tS5h5CqTXL1qPXJLrcYeIrPd8-is9sleuONtMQD6A1v1hlVCFjYO6EJ8ViPEc2DQfJA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/pD_WK1jkaXHGhELPLSvxOMfQb9mASUg0GcAYbz5MF_rx320Hya-BmQELZla5XYU9wAKdu3ucke5P6quEMnIZGBrT9s2R0-QEOtUhCFg8la1---8yvzXVeRRnlAB2_8eojjG9DZzC8PwjejPIbQGoILgV0N5hYlKlDg6CTDs2atmIWv_4YRp27aultSr2lVeH-KsVD9Bjfqe_tdWUy3PHccZroegHeezS4XN6ubBmjGaYPuPQV-erh7GEVVjia3f-NATc5NFRTVkNygkxPKNqliVzPktqhXPtXOtTJygkGcf_sSyTfWpGWMrDuL8YhpBVOGz_EhqRtwQ3tZZjidgG8w.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/ELKFTyZRkAPQPATE0JiK2OJ_aVusAZJAtQBumkh49h_5o3RQlmodX-CEdYw2-2VZKx_lUWmKlvhIjnRQR-TYOe0h4N9eAc97fM70gVRWMlJi9E_3WrAN2hckxav5kMk4zUveqx2GjGfsI2XuK9rSezKbYlcayL833bNFTO0RkSNjTwzYkW1QDOLFBKKU8IM36NONDZ7rZtwhMBPLLzchmKTRS1FREhR4nzZGNAsrBWLhxkLMVS9wwony6PbXuamOSuh32lFzSBz0P2krzNwSo7SLF5SRCecnZbvBbYNk4qIzqLLW4gdGrwsN1vJLNVNA3ll70HN3WKszTnGxRYh1BA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/FVuv-dNStvu_MyWKTE0E3ZOw4D1yMEZgUacKnkHsSM1d3LRjYZ_ivXRZXXC56sRRGC14OkscCwuLMznuwEos4ETUEUwVSeCYakBe3etvvSrU17634h6jepKuFbMJ1rTlP8M1KFoiy5ZK2rxQlYImL2zCgXo8SJNE_DPI1WqN4iWDzI0tQSxE0yX3QeGa19wdUMHyzQ5ZumepXFeMHjbz4ATky6XiAguEUD9gtVy5-0MrIQG84LmVdAEH-SzjZ4uiiFlSED_hG0RIw-_Bim6RV-L1ogHBMweEiWV0qnCE5SLwjFVFeFj-BXXA7sTVxPnmQDAXuVgl94kOM0ZkbbJAgA.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/368?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A pretty interesting presentation from every angle. https://darkolorin.com/posts/358.html https://darkolorin.com/posts/358.html Mon, 16 Oct 2023 21:37:37 GMT <img src="https://cdn4.telesco.pe/file/C-kTheZ0IDzsQUI7gew-4sr_dr3ePZeaNL-WZKSDJBKLty6Lw4pclJM2uiY82pVZtkdoZzImGeJ9z8-kpFutFeqaWSNBZwWnggVW5dd4Rlel0UF9rIWreXzZR1WixH6scje1IE3tMqUP-RULxwVwl-7HvsJQ96pCLqoQ7-Gw0ny7rVrbw1chD54wR4q7gNL31ywcrQCDbRPFk9OGBMyIcM0MeJBMTmz8-WK88Fj5BtFH6MaJzs4gzVg2n8U7NsO0t9tnjX3mb4j1BsWJ6rqlT_sXALjJ350v6QvgbYUY3Yi_rcJIcqZs1NF08XbrlQSotlw7aVsSXK46ZDAo2k4o0A.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/Jt4nTSENRsd80qvDlTYxTq1RmugbvrWIuQIFAtWEGdpRGvuTRF-4d_IeXw0UGph8HagJyCgKz8NW0VgJOLYVVOAsASGkJEaLWehZrjZ37_PVIR4XxYDTRpLCVrF_EfFHAPWmwTJI4FdARUBA5L6yNHuFz4pHZTr-M7RzMWiRSlT457QI0rojsy9VQKk8G70nYI1c5KEwjDLyOILGWq0-jl24QyrPufn1eSGOwVkBBW9fyZyYUCxAhBkRh-zk4RxFLStIctDZHEQEwI_aLW3xtpaPT9HIoywBWgabzWpVvhENvhGe_SELmuN1I410Wnb0WKbHLzt0zTp1ueKrCKcauQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/UWDGLsDuyq8AIrgoNKU7nLQycT_gHmnaIcK-C4qpfBJNdfB2hmpm-9231q9xn7htmX6ggERKM9cdjx5G1U_1Pbjae6Tox5p6HvG78JlGCfTR7LfU3DkDZA4az6RdCOnBAm3gyVd6hpWtdWuVbiRAU9nssNSnbHMUg8LEvoH53PXC9eQi8PkWGHzas9lYK9oXPu9ec33cumC1jpOadMeCzhTvtL7JbBoV-aOW0hDazTidDL6Q-BJEm4JnD-BxJjIxQZ6_LLtRYqTNXOIzzqA11s-iaSnzt_iGa7uvHgqg4NoiF6ApFqVwRAIll1iU-vs6SF5RMr4kxcEe5rGOlVThyw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/LodJMFNEDwf-2DGBbnaxleoik6Fvb8POMeOee3rH3a-89N02Nk-iA9CvSk6TMJ8bHA4f-wDvDT1oVrE6M69qKOKKUkzK_cx1rCVuoJOdDEEkqHL5m8rp-rWyQWfDJgW6xnnK5SHx5ipD8wdkH2M28rw93tz0phRWFg8BHkCw1aqW-uY_GYDf3iQBz8dLc_Y8JjxoL3sy7Qp4vh00AkbyWy95UVyShbOw31lAfwH5plMWNtbj-J0LVQwcc53LdM8RsIyOCnWhLnoJp6jUManlHtHnjosclyplPlVycHjg4-ilUKmHKwXhg_f57_yTfe3fGK5jtZlJXBIzKCeyzCvWDA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/WCBJMr_jpVkYqS0OyEppkYeycYgXmTehuKprp50aviS98odCSzj9bEzetrT9Zgn8DRTtzIvjdYYg1IbyoRPG0uYXV8FRBd__j9z6rwMUQTrN9ge4WJzUBqisvL2qHbWvZ9rs_ZyLpEJA4fyRGLkOtt5GrRW2agWd3276ZuvgsUptuT_OsCJfOVN-gByY4OcVy7aag8m6yu63YkepsE0xwz68zVVFu4WadGjqR26CzyLbmTfITzKVFLtKRwRrt6Pq-opOqZX50VRQPpZDd008RpTkG3cexegbmIuXEVHvJaxY2PQGCrBiURrg2I4t496vawiAqiarLoDKecNrtnI76Q.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/sYWF7rqKu7AmY9HLf32_q2j5pxCDJPUQsGFybk51EDCOgJckVDLV1tNDE7cM6rBwY1tYN00z651pkhpE9Zw1KSDJjGR9W4uAuHpyUSOqCG7G4jGvZh9-DcsldUwGj-sulEKYYMhEWeAP-4FHqazmHoTdJh2nvIDbwFhHo0BKwmNWJ5a3tnvutCqo3xXOzefKl6E8sPtbEQl1B9uaxfkZKqnTgajtH8GQGRemokzK_nuxRJCAXa1KJx9Aq_ioNQUM0-62ZGbA5AQ49uxxFcQT8BcFDJy8MhzWBM4nnvGPOQzIq401SjfJp-k39vqHgdC3OzMYqflRVcEChSnw-3J7Sw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/JmBYwvrQEbuvaMCWr-WWcLOmqk-9mBTo9VZ5M4CIIwPeYS6OFfngDf9dY_xDOwxVULMAyUAJhQdbfOu82uAq0JsLtDM0y-lXJUb4SKTDoc7NKGBNFFlMgF1gjQjj3O5evevOwxs2TNoqTokJqxeLzuIx8f1RdR7xeGG0-x1nmrtP32hIMMIzUTzN41Yk3lDwtpGcFkzvg6Z014ivFoilyjPKus7WI2qOwzgGSgQ9FA25Fz5k01gixARBp-C4bNa2257gW37rfPqny9uiuhv3D7e5uCXO1nEVWBX5kL8Uh9e9rnR34sAq1q4nLDAX0MTxzshfnKZjyXD3jdnufLoCHQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/lI8cQEeOC1PMFnPTd1i1gSin3aMCDCPS7-l-GypPEpTAY9WV-ZKUg88rZBXRnn_-hiRXiU9Yl9jOvO0YxyNy2RlRirigoG2iLvaLORtjbFL9m8GradivfRHtwEt4n5UWirqWIEXV_nmLBmTsYeUSJIau99bCAtjIccVQ7UCEB9yP4Yw4bpX7dOfTvFMci_le2XPU5s_6X4DOZVLXAdJBs1YLBU5YGnrtou90GQpAQuBMp7PknMDoqhub2MhnEKWuyGMOs21V7DgTeZeDaVR4B6EBTA1hsD8N5RGMAIWkzLggpgXSPajyc3P0EM9UqJD9tDTEqRVXhhthgelAdo9VFA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/ODsIgCWsFxc2IMzdWWvGst3OPbI3w6iVgycF77nDwBCvwhhFLuym0nS-Wh0c927YZ_EMTwq9OXvqI4cCCkHknLTZSdryKU7UtNs-6YZkMPOuSR11duXD_8LU0TVicEvJTupmFxCrBMPjAU4qlCM-Rh2ki-ooAeR89NDJO-4KRS-i02F2ds3AFvdb7ANb7jEOEOPi_9ygBYtPNhJTmRhV5YDkWdkBRoNpMjn-gubkuek4F29-e_gLmJsJqIu0LOHAplOhSLHEqDLMjAuKFXpjEOSZJhIg1MSWX0uOLWoAMY6p_RdhikwHC3ShfqnqGXgRrSj1PhomKexZ-y9owdQaZQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/m4SXg6y8-C5hUMHn9RChMHAgh4AktHby3cRc-_-I6J6PA0sZ0H6lASRw-_C8IQgfSlZ5aIZ5WjUI5NWcqWKpHN11HLCpMAeY3jt41ZavKyXCLhuxsuYeGh4rLi5MukLhFX6_UzfRXrxf7w1DHqmaacJy2JFiGAugob7UBxcvYWg9WOvkjXhn7EZXT546PP-nSC-DkQAcFzGbdpXZJ8ioBYlMFWFbVtgrS0NO5DnuYHicb02pCO5NYwvG70d2TnOJVFr0yIpypOH2CLcaqFpip96WIS0Fqb9THmTd29ORzEaw4zhXiDrx8Y7D55_zGkNCHl88LDzJ_c_Ftsrqn26HNQ.jpg" alt="" loading="lazy"> <p>A pretty interesting presentation from every angle. Both the idea that there’ll be a lot of different new hardware devices and that Apple’s kinda old and the time will come to change course.</p> <p><a href="https://t.me/chillhousetech/358?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well there you go guys raised a $5M seed for manual recommendations on where… https://darkolorin.com/posts/357.html https://darkolorin.com/posts/357.html Mon, 16 Oct 2023 17:20:06 GMT <p>Well there you go guys <br> <a href="https://www.itsgood.life/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.itsgood.life/</a> <br> raised a $5M seed for manual recommendations on where to eat from trusted friends or celebs. Now *that’s* what I call a hypothesis. Though what the problem is with just asking your friends I’ve got no idea, but the celebrity angle is more or less understandable. You onboard them, their fans come in, pay a subscription for recs from the team or the celeb or the startup itself, to like, eat where it’s trendy.</p> <p><a href="https://t.me/chillhousetech/357?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A report that very clearly illustrates what’s going on in US venture, and I… https://darkolorin.com/posts/356.html https://darkolorin.com/posts/356.html Mon, 16 Oct 2023 13:14:39 GMT <p>A report that very clearly illustrates what’s going on in US venture, and I think everywhere else next. TLDR (I’m trying to run summaries through Anthropic btw, how is it, ok?):<br> - Deal activity remains depressed compared to the breakneck pace of 2020–2021. Deal value in Q3 was the lowest since Q2 2018, and deal count the lowest in 3 years.<br> - Mega-deals ($100M+) have become less common, making up 48.5% of deal value in Q3 versus 60% in Q4 2021. That points to investors being more restrained.<br> - Software’s share of investment hit a multi-year low, while biotech’s share grew. This highlights the squeeze between sectors amid the decline in overall deal volume.<br> - Instacart and Klaviyo IPOs in Q3 injected some optimism, but both took valuation haircuts versus their last private rounds. The public-market headwinds are still there.<br> - 77% of capital raised by VC funds in 2023 went to established managers, while debut funds are heading toward their lowest level in 10 years. New managers are having a rough time.<br> - Deal terms remain investor-friendly, especially at late stage. Insider rounds are at a multi-year high, and rounds with a new lead investor taking a board seat are fewer than at any point in the last 10 years.<br> - Non-traditional investors like corporate VC and crossovers keep pulling back. Their share of deal value fell to a 5-year low in Q3.<br> - While exit value in Q3 increased thanks to Instacart and Klaviyo, the number of exits stays low. The IPO backlog could mean more liquidity events, but the bar for exits is still high.<br> Overall, the VC market is still under stress, even though sectors like biotech and AI highlight the strengths of the US venture market.</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/356?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Brief recap from Peter Thiel’s latest interview: (1) Don't play zero-sum… https://darkolorin.com/posts/355.html https://darkolorin.com/posts/355.html Mon, 16 Oct 2023 10:17:51 GMT <p>Brief recap from Peter Thiel’s latest interview:<br> (1) Don&#x27;t play zero-sum games; play positive-sum games that lead to growth. <br> (2) Develop the ability to think independently about new technologies without pattern matching--that&#x27;s where the biggest breakthroughs will be.</p> <p>And in general, here’s the legendary lecture, I’ve always been blown away by that kind of thinking. Never once ran into anything like it across the former USSR, feels like people like that just don’t exist among those where I grew up. At the same time I’ve seen people like that when I was in the Valley.<br> <a href="https://www.youtube.com/watch?v=3Fx5Q8xGU8k&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=3Fx5Q8xGU8k</a></p> <p><a href="https://t.me/chillhousetech/355?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Also really waiting for the Zen rollout on Google’s homepage and a sea of… https://darkolorin.com/posts/354.html https://darkolorin.com/posts/354.html Sun, 15 Oct 2023 21:59:03 GMT <img src="https://cdn4.telesco.pe/file/uHHABQenR9RxnNPYtpsN4OO3XLNffLsGAKm0BrN1qHYrmTysxm3N0gTz8uva18bOX_61-Rxc8IldyHH4vj0bEWbjZPpkL6xfvAvVQH07in4rHQCMc1n0xEnBkJeiCO2OOBjl15ih_1-vM-ms-JqGPEgHoN3YjCXh3ohX84NM-3JN2NUCvPtHDFCoIjaTvU_m2hZUsAni8Xhui3y922AQc2duCCtmMoz8yv1nqeHtiMyOm7xNAsfgxi9KIqONSyqVWYvKdIPQQPg_SIw7bUFqX4fcAIJFH0wfHVU_hKflUCCSCBc0xvoKEY9rx0nL4X4w1-XMVS71Y2xkRufSrPS6TA.jpg" alt="" loading="lazy"> <p>Also really waiting for the Zen rollout on Google’s homepage and a sea of shitflinging on Twitter over the propaganda</p> <p><a href="https://t.me/chillhousetech/354?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Back from a mostly offline trip to Scotland. https://darkolorin.com/posts/353.html https://darkolorin.com/posts/353.html Sun, 15 Oct 2023 21:32:32 GMT <img src="https://cdn4.telesco.pe/file/p3wDyNu55g_9V9QjRzMlHGRy0s0mvmlHo5_fWzJicAWcwPOdH0itUZ2C0Y8C8--DkzHMQvNehRhi7pLwiuogO9xLK8YuC-ztvPGnfSV3dcZKnsj7d-6H3ENpUISonZEtzHPZs2ZCsKGOv_B3mmYmQCuNxy-WP7b5QBMZDiiTXTg0gG7EODPIRy7dRLuHexnbSbpNvCY8jW1IoiS0NUGGGvkwsbNyTMYGxwCP_0YYpbw6uPq4q1XXuKuGX9KHFwVroDputepbMjU5a8eB3QGTnap5LqeWuRP3y6GXzaHajJ1d5HFzjaM_xKFD7pDwhzSR6R3D-MP8aJRDCPoieXBIUQ.jpg" alt="" loading="lazy"> <p>Back from a mostly offline trip to Scotland. So now back to news and content. <br> Bloggers, I see, have in general gotten close to unicorn startups. And it seems the effort is basically the same and the time path too (8–10 years). A table with yearly earnings.</p> <p><a href="https://t.me/chillhousetech/353?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One more M&A Most of the amount is cash. https://darkolorin.com/posts/352.html https://darkolorin.com/posts/352.html Thu, 12 Oct 2023 14:09:34 GMT <p>One more M&amp;A <br> <a href="https://www.marketwatch.com/story/atlassian-buying-private-video-messaging-platform-loom-for-975m-ec97c8cf?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.marketwatch.com/story/atlassian-buying-private-video-messaging-platform-loom-for-975m-ec97c8cf</a> <br> Most of the amount is cash. A very good deal.</p> <p><a href="https://t.me/chillhousetech/352?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/350.html https://darkolorin.com/posts/350.html Wed, 11 Oct 2023 23:09:26 GMT <img src="https://cdn4.telesco.pe/file/L16pyN_Sq1ZzzIOgIblQV1ouUxjhNSJW22AAkMG9qM3RzRjNuww24zU7kxLlInfPjJeOAXtzYI0vIOw7s-oml7-Oi5Q6AzX71hN1SecywCoNNVCYFndaLnVVkkhQS5LRb5Ew1pwy8sSEnBo8fJD5HoFQgFb4OsS58eRPBU1katgDrPMtX72evNN1i3ZQgI8dD8q-qmzzQEcHuJK2f8KlwWVWjbwsc7aZjjlqOfu-mnBHyJA9_x2E91kf3zroIfQMXIr47bR08c3c2DAJcEO4kIPgRm3sMGYESRa6teBQk9a-d37zlZ3fQSil4hmXRxuuqHCZyvQByfb97b90uCe5_A.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/kstCvSuDgrhYA454enwVaqaRyuXUonisbkv6nXKELzjFTuc0FRn3W9WH8A87kOcz7_G3iGQexHqmq2WW53iJK7lXH9Rxab7pu1QgoHbyFrqW-5AfidrN3tp1bFUzJ2JBskZAQd-GgaLAzKS8AX-4j7Krl0CwmqB3cyNSvuP7T3GtMF69QqaMAYvK2gawE6ApL5FmQi6G_Hl7vAIYA7cOuWWsWn_GAM4cqHR-0K_S97FMs1_CqCiUU0uXyIAjxj0pJ-7yairilUGysuR1DZsAGKIfRlhM4V5WlRCCIKsQPiMWeLTqj8GVJ3kB8qCJOe4kHPeiXL8ZKpgAGXhbVYKKjg.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/350?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/348.html https://darkolorin.com/posts/348.html Wed, 11 Oct 2023 16:54:35 GMT <img src="https://cdn4.telesco.pe/file/lQnMmdeF1mwIjodqGq2Gs31touZ7kVRiuDl81-aUMFiO3KhRqx4OdaR0QRz70p_cnRjyu92bAyXGYVp7fSlI_PqEG0jPNzXsRBp3lbjw6PWPbqlfoyzqdlPtdscckDlVImFLX7lO_j8weBWm5pE2rg21N7uAm-X3wqJXl__qOmBPmALVBbENyuMZpfZWkMeVSZ7dSmjt_yGol_wFkRXLp75iwg18lgATVAvH00nNxYbzS1JczrQ7-MQIeaXW6VRN5rWWtgLArew7TGdwZBwCPP_0ALV2TkE3D-vNhGLk4PMlGPjNpCzzdbKi4yaeYsKsb-xD2NvwgL3pD3SYKOGIXg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/Z-WNfEZk5UOMlrr6tZ4BXNwMkuizCKIjDUZ3DjInbK5T0paEma4Z9Joup0oFMc_5WCAKpedqQW19Etvb2snbuPRrOhWsyqjMgoT8BNsmDqf9E4zAkcOBxbh7xXrWvCbNXWjPQ11yTtHn0GH_a3ZWzHXQZ4Ap8CmkUiG0wWOBWz4uQUGvK-a2zBoMpoqz-Rw_hS5fHcliKRSMQMKYtTq6xF32amRKZNC-2L_iFdM6x82Q6RQ-f0c1-yAVnCUnMdjz7NFlsgRDf4hv6Zzr68ncQKANYdML7K7A447gLZI_5wLDbeJOWI6sETmiXuCJwOyLJ3E4bmAmLw6pLjKxOey9OQ.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/348?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Eleven labs deck ~20m usd raised 100+ m valuation https://darkolorin.com/posts/338.html https://darkolorin.com/posts/338.html Wed, 11 Oct 2023 16:54:34 GMT <img src="https://cdn4.telesco.pe/file/ls-u7jLAXGThZuatb2vfT9ogDKLVmYy5R3eMXlnEi3BCKFJA3s3DITqGW9JY-PFSFvCu5qVoUgLDUEDaEUcrIz2mHB3bWRefW4cP1rHkIhQnVznpRyggEMyRksUXUIqmQK2CQR1JnnYRIWQ2F-nhxCq-xCdqTkf-bLyUHJXIJu0iMPUD3UUAsJHyWPgXCgDA_UH5k99j862BMXPMhLPsAwawzYv1SDS-eL7bgN_ToXuosDJSQTK8sRMb69StYO5IHmrB-heDyEo-GjASKzb2h6e-52Ka3EnSL2OEieAvlbIEAjtd4R9kmvxqtz6YH5EXu3jJX57R9ZGRhIOmSN2NGQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/rISlt6mPbsbK5n4pgGb3JatvW4xy1ulvosewxRHxsw_-HBGXvNQE9dB4x4resdx4-PMp7YOuH5YUAAW9ssW1sOi0PG2eB6hMJexlpgb5hxLyJTqFFQqjlt8OPA_CoBkXrHjE8vO8JkDQ9QNwr4tIXHpxRRVSKNtTh3Y4sl6VLMuaKBPIJPD-JiscK19vL26Mfv72-mHF-gy_IRrXWmUpr4DHaB3Wywc2Ar5QrHlVpNDCBOgA0HXWrwmBp2ogI-rPNJsHZVkqu6VHO-sjWK-cC0RRPPv2CkHP2aLANSXNTc44OnG6d-RcEwTcrEoVDoYx4H55EtopxqDDDrUGLHK0SQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/uSw_kTFLZ8c4ftnfTiyvUt-4px9Z_PaLOvdGXDHDuCctStGZEmKxoKbcNS4XDvWN3K-ioEhEhcIbGFrEley4JgZXvid0Ajpd9gKszsJhdQRyHWCQWtVtSGXT0qtY_pz3J3-QEzvhYrry_kr6CRqI6uf_mLGftw_BVj_PVyv0OwxGQdtoGIUV0YQWfeE3HpvBaKKc2Ew-cEtj-7-U2AypQz_ulVcTu5jf2Sl0q4kx1xgWlrCtD6eWrptFoCF0hwvSEJFx8bHP2NUqthMlTz3nOGNOOEIei6IgzCoZ89XQCmPyciUx7s7uHqbSeEuFBnnUjZt2JG1Nz_jTy0vK3EUBAQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/UnORiIpPBWWofbWVg-mJCZBoOwix3VdlibtnM1bzxGHzKGkkbZ2zsa0AmgSjJz3yQ-3AMMxcHt01339kk3GrEYPBfZC9zlEN7dnhhqDWht7JzlkdGgFLdazwNjn0or1DoMe4MCiWwnL4xSlOy_PM5P9jk99-uHIJC4WSMF8hXXmBpqbZRGz1WDvPzzrBhkPETob4EF86H_iWGbE81MTAz0v92RT0cccN3Q1Qj1FTEAWlROAbfpAneOI-Vuv33nRsjXWHd9Ev7stoLnuvAhSwzL-OCDnpW1J2eWa1PTP_LrfQoKx33dyCWlC10urcK5oIysM3s7IFtdU31halkT8_2Q.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/dHShF37-mx3wa7ekxNmxgFwZZoh86nZyx4xdlcVPn7X2B0WRzVaEqIIfESD-1j-M054sAsEyV6Tv1cmD15b7sngw2IlH3pLrzxc07SfzkvPMjnB5ZL0PmBlIYhlL1Gq-vD1Bh2mOy5_wR4uzsedL0Mfa4bQ67c08ejzc5gU35YakrKAWqd2rO8JKEud6cHxAjsSxn7mHRPhHnN-J75tPlz-Nc00el2I2kMRjMK4wZyvyQ2f3dkY9g25c3sNLRSjFcxGlyPZxGVT7Ztc6uCbut7lhaKTuuh9ZbCfnqgV2AmCMEfhSsOF0SeRR7UOW0BRIs2KtQE2taMzZFJreOGOGYQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/lBPHUJIBnITazGhg0HZOnX3RHRcqkSLlE3aeE0JnHplfnXVOkTxseXuBzTFJi5aeeX94B5X26dpX7eYx-NSGzoaIbpGXPZoQge7J5U4ecfeZSPu9Lt1ppSGQzHmqdDzU0BXTdbxWh0-ETZhDj79dHK6a83riGdQrG0GUFc8coUNYtajWRBPkr4jDuNZ8xKymUhZgrPwqr_wjXLvQN1k6NCkcoDb6BlyfTveXbPe69bX3gnKJME4eFuUKSW0qdJYak_gwzMsKX9oFTvhEQ5Wu18qEWpxpfR8RoJZ3qFbKtuTd2XZBaY5IBUi9VwM7c4ENvxjn4Gi4JpDSK9Hfi3cF0w.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/o46pL83f0gzT5FrXVQZ_frkjrCwe_U_id_araYxWoosC3ePP9rdO71_wjCZTyfU8FmdoXzTgT-g98mS8zQrhnnfI_IG58x1tj2dbIVizRUSwDWl6aopFM4uP_wrKVR_2s_hTQE8mCl1Qfwn1x65M3W-H0tQAJq8SKz4BsSdfL55BlJ1bZhUdt3GEIi8WrTHCYmq3ogfHTuKjcZZ5ArPzg2dVujulOwTVVKObclQEqws7YGNdTh259ebjIWfb71bvyECYEiEu2banCt18Aaj1ZXzrgv0PygiK5266sbjToIRsliEjZPJBA9Bb0VzVh9lXHSyMkx_Gz6jRWwh4ZytdFg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/D8dGGeQMipscwBqEy5IuySAcQNH6ap-58j3phn9kSW5y6LFnneC0J6PZGqj2lzVt6PpEORruDx0x0yQ2fCiWemfHoO6TEC-fMjhsQQDncTnDs8acf2r0X_u-djNe27UzBAlvr99zzlLmRaEWgrCuyROAj04jY1LdsfpfZmFdHG13LG1onaQHnG5VuKa0j_9aKy47TE9xAOlbdcBq33VEBQNXHw5LXqvZJK53afv8xI8shSSDb-JvAhQuH5kkioR6_OcHl4wx2DbQ6fuWEs4NTkbxaVtwajiHXbaJk-w-F8xmIcHqBa_Zi4zjwHjXCu1Dy2WmhtLK51jXF96gv-GHBw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/rYnJzpLN3bamGAnCA5MQ9B3YVg5a2ELDld_aYm7Lvoe3uZxFYIk7XM5tK-BHi2Mz0J7UQNoE1kfDa8DQOyJHdC_a_fgaiwhHPuuWdMTDc61ny28PfYO6ADWMHpSUmdHSJHKBgMDiEa8RFPiPlkplEP4aCtGbHlobgGla2sxVp8Af7tINvQ2323_Uyag6qU3J2DgdQ9ZceMwwzo6NX16w3wweJVdP_FV1yf-hMgALvPQyHqB8Q1yY4Z6Xo5Gy71hEmkk29GI2ClsXmImHcPw-YOZnkbAd5ijWzciTi_o0VE5H5jQu60TA4-3_LmnNt9OzEeCFX1xpYKhrzjHbEpcN8Q.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/nS785aO-1yY2ZFGhnEb1WIsnlpmJMoKtm86QMtjpj6maRN2p2zSVq6zrqLAAETn_ZSmgtyRUqeedN4rGZ4H8S5lKpd7Kpu3zlDDTNgGjfV4ehhosGjS7lan_ABx2-v6IGVmB6QMaGIkbGspG_VO2KssjgNuyA37a6MoWcAKWxgyIryGtin1xvB9vyNjfpW8gfiQ6w6EjpQQhkUuVoN7bXnaPsQZmSPakJXwZP-AjGlqK1ZZaeKrax5KnCceKD5TWRmp97KaaZJX_OgY1A3mBctRQ5-vYruF_e--1MaKyTEGffWC8B6FuLIGxg_P2AXNMCCIjak1y91d4sAXQWntZqQ.jpg" alt="" loading="lazy"> <p>Eleven labs deck ~20m usd raised 100+ m valuation</p> <p><a href="https://t.me/chillhousetech/338?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Google pays Apple ~20B for their search to be the default there. https://darkolorin.com/posts/337.html https://darkolorin.com/posts/337.html Wed, 11 Oct 2023 14:42:09 GMT <p>Google pays Apple ~20B for their search to be the default there. But that’s profitable like, insanely, I think — can you imagine how many ads get shown there. <br> But the most interesting thing in this deal is that Google is needed by Apple itself. Because users will riot, and Google’s search quality is kinda higher, seems like (debatable, but the global coverage is definitely better, and here it’s exactly one partner for all markets).</p> <p><a href="https://t.me/chillhousetech/337?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Pretty interesting material and its transcript on how to make it so that AI… https://darkolorin.com/posts/336.html https://darkolorin.com/posts/336.html Wed, 11 Oct 2023 13:39:59 GMT <p>Pretty interesting material and its transcript on how to make it so that AI, in general, actually helps <br> <a href="https://www.youtube.com/watch?v=-vsYtevJ2bc&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.youtube.com/watch?v=-vsYtevJ2bc</a> <br> + <br> <a href="https://www.effectivealtruism.org/articles/paul-christiano-current-work-in-ai-alignment?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.effectivealtruism.org/articles/paul-christiano-current-work-in-ai-alignment</a> <br> What’s more, it’s interesting when they were saying all this 4 years ago and openAI employees</p> <p><a href="https://t.me/chillhousetech/336?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And back in the day at Revolut they used to say: use your head, not A/B tests. https://darkolorin.com/posts/335.html https://darkolorin.com/posts/335.html Wed, 11 Oct 2023 01:42:34 GMT <p>And back in the day at Revolut they used to say: use your head, not A/B tests.</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/335?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “Report: GitHub Copilot Loses an Average of $20 Per User Per Month” Here we’re… https://darkolorin.com/posts/334.html https://darkolorin.com/posts/334.html Tue, 10 Oct 2023 13:48:08 GMT <p>“Report: GitHub Copilot Loses an Average of $20 Per User Per Month”<br> Here we’re basically in for just reducing the number of parameters while keeping quality in specific verticals. I think GitHub is still using a very general solution. And overall this direction is top.<br> Not long ago an a16z partner DM’d me looking for who’s doing what in genAI for SMB. Anyone doing anything? Anyone know interesting directions?</p> <p><a href="https://t.me/chillhousetech/334?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So yeah, what I was waiting for at uni when I wrote something in vim was… https://darkolorin.com/posts/333.html https://darkolorin.com/posts/333.html Tue, 10 Oct 2023 13:39:32 GMT <img src="https://cdn4.telesco.pe/file/P6wDJwnhk2wncMIG4W5TmmcjNY_j2i0JHvJ0mMdL1J-zHK8lNMdv6IxN4BxPUCBgg5bPkWEjqO3oOULLw2no38dPrXGYWD1USCiwgfJGUan9AqSaFuWhNhsFXBRCqlk4SrxFJRjhMGa0vhxzBFZvxAKFpwQdX9MDGzW67wqrjWgo2DvL-Y-j-HiwYF3j2XBPIh-MVhzKu172-JvGSx6VLc26J54lNLcrS05WxvTdug5zH3bfLOdJ5or0_xyOqlMhIZcHLoIYK4z-V8Z5K4qSZUYJEpv8HAkaaxGIr-8jZef8PY8K_Ls6InmEv12z2NwbURCZZKikY2k89RY4N4JkFA.jpg" alt="" loading="lazy"> <p>So yeah, what I was waiting for at uni when I wrote something in vim was automatic bug fixing. Although of course, judging by the architecture, fixing from scratch won’t work out particularly well. <br> <a href="https://ieeexplore.ieee.org/document/8804442?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ieeexplore.ieee.org/document/8804442</a></p> <p><a href="https://t.me/chillhousetech/333?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> welcome to the world of app distribution from Instagram https://darkolorin.com/posts/332.html https://darkolorin.com/posts/332.html Tue, 10 Oct 2023 12:58:12 GMT <img src="https://cdn4.telesco.pe/file/iWD9W_enWrbcL5ZowfkQqJN8TxGGFPQ4m-fGpzfrT_NEsnWkLaevDjZir_nNI9YXRdSJ011hDTqYpuLHgaVutCYDzhO2M5YnWty5XNQldkbxvYCxPqrbsbPwOfAnZ2USUXkP352zxSedfCFnpEnmiRMdQV8t9MRn2mwZdUOEf04f3zKKfmmPNH5lBZVDQ-5jA3HRgrs2PTs2LEgl4Ot1O-i3mCtTxZeDAEszQCjjj5dTZjTNpo5PmUDyNk00k78Fgu7fb_qh5Sn2mUnBonsiaOkatljzqnH-BJ9LHrCNjpkm-8ggqd30b3py9zwmymkMkRWD4wy4cg7W9RjoNEzvZQ.jpg" alt="" loading="lazy"> <p>welcome to the world of app distribution from Instagram</p> <p><a href="https://t.me/chillhousetech/332?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And back in the day at Revolut they used to say: use your head, not A/B tests. https://darkolorin.com/posts/331.html https://darkolorin.com/posts/331.html Mon, 09 Oct 2023 20:22:09 GMT <img src="https://cdn4.telesco.pe/file/L_Yqoff5BQSjRk2iACe51-Jh_DnYhB2-dOxDq68PrCP-ww8c5MIPN3jPcalX6jVYJogvnM6wo1oX5HSmt1vDGk0qBbLJrIvR-xj0B9u0cDOikj8se0624JBz5Lfg9WxqwKNfuZeNqYhy2R4HR05vrYFKlxTjaXYuwR6YGgO-Oy0_Q9Jb-nP1VIEzj78ZVzJ3DyfE280reCGBvUa2OWhmwMraH4xr05bqNsATdX6qXJ5XIVQ79ElgVSwjafhTUKGnFy2ON-O1rvplPJ2mYdv100eTfZJ9Sq_b-1SkGaR5hak8hF2DqzOPj_K7v1BXIlyziUqqAyi3g2DOEtl48nJ2CQ.jpg" alt="" loading="lazy"> <p>And back in the day at Revolut they used to say: use your head, not A/B tests.</p> <p><a href="https://t.me/chillhousetech/331?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> 15 years and still relevant https://darkolorin.com/posts/329.html https://darkolorin.com/posts/329.html Sun, 08 Oct 2023 15:23:32 GMT <img src="https://cdn4.telesco.pe/file/OSSpqV6ouP3ARXzdmhiOyIqLFHtQezF274FwjCHQamYNe5pJmdQKOquDLywRElKw4h3gF1voB_53a54D4xwVrVGN12epeWYuqUDkSh14dsOEGj-cl8ezCZN8IfAO4zgcRmqQitOrCMTCu6lMoBc7C_QhG-X0GtbNTJpzjUHnqf6aGDI9ZyvsFouy0ixHf99dHvcspvDZy4Eo3slpHI5nqOssjf8Xwj_4T4IIeZvrRN4aOLC-wkDMpnw3oX8Ccu56UUYqZAVHP6n5JgufGe7MP42sYPXJ5e1VVxLBbuOnpLu29Vy-TK5w6jbl1cnUCmk6jWN5upqaFEIU6Z2ooHuT5g.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/h8jPRouAVf59We4Qo8U0ReKFlX9tetjZ_1mEs2jnv86I3u3uQ-2VzmlJYGk9t3xoUAFN3yFxcXcJEZQf0fjYXxZb7Uo5YaAPVBY575I9Um_87f3XFFhEJgWQPMZM3OYRM0Baa-xIKirAMTamEqAfE6IjgwZr_BwI_lw0IaMVIUo-oTKwhBjfJus_CjjPrvk2t5SJDH2zI76vUQYMdVDLvKX7e1vJD4VH0CtZ9BhKXdkWwz5RVgGp8Tcv3mKy2eL3wugmMjTYTBSaftWcf297Tzk-w4s2JEuCR-F5ghohiIhZnxgUWiP7o7wF5EA3Q1P0-Ra_zXkVDg4wxPdOpN7oqQ.jpg" alt="" loading="lazy"> <p>15 years and still relevant <br> <a href="https://a16z.com/good-product-manager-bad-product-manager/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://a16z.com/good-product-manager-bad-product-manager/</a></p> <p><a href="https://t.me/chillhousetech/329?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, I mean, decent startup, but like… why though? https://darkolorin.com/posts/328.html https://darkolorin.com/posts/328.html Thu, 05 Oct 2023 14:37:43 GMT <p>Well, I mean, decent startup, but like… why though? <br> <a href="https://x.com/alexisohanian/status/1709939631490506770?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/alexisohanian/status/1709939631490506770?s=46</a></p> <p><a href="https://t.me/chillhousetech/328?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ve always been curious how founders sell this stuff, man, the bullshit they… https://darkolorin.com/posts/327.html https://darkolorin.com/posts/327.html Thu, 05 Oct 2023 12:34:11 GMT <img src="https://cdn4.telesco.pe/file/KclZS1RLPCf0Nbg-sUWDQK_MtXxvTzzPqiZHP1SdQ7oUz0cTp59bpWHWdRDwtLPcb-F7cVwJ4DFKp64SJaKG3CFb4EpQNgxywoqO7G81IbyhzIBqaNnIlHPsii263fEdkbSgjzt3SuvpbID-1UVfzI5LqNAZf3Ru-vTC1ECdis5EkCG4F8eL7gRrZb9V7ceH4a2BVQwU-jPxWf3ETvTZJcNIY1OH1026pjGim2T76h506lo_iRP-a4FhNKONc-ArnErnICh_uP3vg3n62yjVaj8UylgXsMYH8dAE_y1UFdElM42PEXlPh2CabN2VW-DKuzoRRr1P4jEZjWhhOa7eeA.jpg" alt="" loading="lazy"> <p>I’ve always been curious how founders sell this stuff, man, the bullshit they hang on your ears there.</p> <p><a href="https://t.me/chillhousetech/327?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A whole cohort of players has moved into UGC Games: At the same time… https://darkolorin.com/posts/326.html https://darkolorin.com/posts/326.html Wed, 04 Oct 2023 20:37:06 GMT <p>A whole cohort of players has moved into UGC Games:<br> <a href="https://www.ign.com/articles/fortnite-creative-20-x-100-thieves-bank-heist-game-mode-explained?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.ign.com/articles/fortnite-creative-20-x-100-thieves-bank-heist-game-mode-explained</a><br> <a href="https://dev.epicgames.com/community/fortnite/getting-started/creative?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://dev.epicgames.com/community/fortnite/getting-started/creative</a><br> At the same time, Bladeball in Roblox hit 255k concurrent players.<br> <a href="https://www.roblox.com/games/13772394625/Blade-Ball?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.roblox.com/games/13772394625/Blade-Ball</a><br> (now it’s like 130k).<br> Trend’s doing pretty well.</p> <p><a href="https://t.me/chillhousetech/326?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Calling it now: every big market will have (already has) its own watchdog —… https://darkolorin.com/posts/325.html https://darkolorin.com/posts/325.html Wed, 04 Oct 2023 12:54:16 GMT <p>Calling it now: every big market will have (already has) its own watchdog — app/site/data removals, full-on regulation, sharing everything and anything with the state. Goodbye, 2000s internet. <br> <a href="https://www.wsj.com/tech/apple-china-met-to-discuss-beijings-crackdown-on-western-apps-2219afcb?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.wsj.com/tech/apple-china-met-to-discuss-beijings-crackdown-on-western-apps-2219afcb</a></p> <p><a href="https://t.me/chillhousetech/325?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Great report on influencers and the creator economy https://darkolorin.com/posts/324.html https://darkolorin.com/posts/324.html Wed, 04 Oct 2023 11:32:30 GMT <p>Great report on influencers and the creator economy <br> <a href="https://linktr.ee/creator-report-23/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://linktr.ee/creator-report-23/</a></p> <p><a href="https://t.me/chillhousetech/324?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Dec https://darkolorin.com/posts/317.html https://darkolorin.com/posts/317.html Wed, 04 Oct 2023 01:10:07 GMT <img src="https://cdn4.telesco.pe/file/IOwumCUfhIJXF3zreXaLUH_xxzDFUj64ivzIuMm_0KuQv1sewQQWeLVd8IaMjlrItd3tkLWbEmkIcF84NALWPxQ9BaryjnOrpgH0XVcCO00pfyKVqm7PyGwz_bEvim-Sr60W9e8uvqfjpn9eGK9mW_KzCy4NgL9nuP51p9kw_UKWR1EBWr5dYhjo6Z9qeecXzQko0Zsp87eH8YQkZRvhUX0VaP3bbVtdiuzmalkqw7_0fc74j7-xJPccDHdQj1_PhQggbd2zSmpdr1c5JOCjNhjISVqestg09TXLS8VAoPY8u7fTUcCbDfz-oVxSNlZumMslhaAmxRD2oratlLmaRA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/n7xmQfiYn_ZpMoAEPcyt3GB3HIM3LD2_V5rkRUCzMyFo6f_qyx6D22govzEgR7kVrXikbf-Pcuen5Z0F_hYEStvJQqhNyhNBQ8Smsm2hNecT8Oafmbb23PBHOtmFNhEGUYDaYp6WqqznlgfXNcV9Oc0HYIjVOBTjMjkqKWNO7X7d5LRe0oixr2EIdRHo88sNJOeXJJDW4nEK_7B_WM8t5FfCCY5JkjI3vWSqmhDdlzkPHNBHvNw3voJU_ZX7roqukihr2Z5LSWmIY6eWGQTCz7EpUAWIYBKKq3tcRu8orZH_Y_AATGggUDD39fJLZnUdQ04DSwCyBGDBpW9XE1-1Jg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/HqHwHJqe1bzwLUc0R2qh_mSvsURYmbttrhmTQqP-dFk6QmVLOfr6ZGQ96_CAfxzvwzTpvMZ_jnwEQnimxogOiVvcYOLTZqhTZnwy_WuOFnMq1pQBeoyXG_HJoA4jxYlsPZqWorg2OZpHAKAM9VMYmG0GOO--cDDMtbWhZdj3mLM9F07nM4uRKbsTxlfTZOPVAT_8BJHN9xf5nm3N3PEBoeV8HB80i3zublRxgEsG1x9N_AsaprY8iS3XZbyNcmi_y2pEA6my02t9dWQeO0LIrwsQ9gG8dou0FPbL46-i4zv5YWZdAm74orCBuGGD9aLkud2gRzlBKFDzkMFOUqWLpA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/QgS19O7tRkBCbEQmeu-y40LTRRDlEUSD-uaZnAMWLdM4IAp3l54R4-mSUrNb6j9piDJIand5oQkGB3DGYZWfEI5D4YcignOHrYciAwavyy4vGfXLhjQbfI_vm79o6H_GvRHb3UgWY2eIGiCOWcZTQXv44cNzG-jb_slpaepqzRRDCWngOFPc1jphJtbg10ClqFnIEodiHKX8PqDctTSarp40ShC6Q2MTBpOvhGqm4TSNCR4zEy9Esi6iMQgy2Nu5K3LLV7MvzN3-QJRG-POIFAX5G80Z-RlCHoFfZLXp0Za1T4wj2BI7L5Pp3rW4sLtJ94f89_tm8PVtBpYachzdhg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/vKqmdsHpjVddKfUQiopbB4WOo6PXEfdz0ik8mLpnEF0-9BBnyZmGVRMaxcRmC2s82y5O_2G7395C0fOi5eIqbMHl57QrzkLXXWuBLLtWS6hP1Rws1UApGMNy6RLLcx_dSUBxAeWzX9ZOWQIaJU05oNnHgrSW-c0phw8Cp6KZsqYZMUyWkkpml-XBl-RHyFyRQXyxaHksT0hHmPdZBTKWzsBvJcfC6SPg86Fp8hGJneakrSQN6K0oeCGyFrbts6IXhEa34V2fkAHLi3eZTXbcFj5Aimp1wo1F4OVnXc63vrmERkvBY3IheH7IsGfciByOtSqdml-oooS1W-qXXpbtMA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/C1anD0eC6tSBgAJnuAQDig1x0iPGeJ69_zz1AHvfEFIMh-himu4HbGxyZ3893N5mXSZZFKDURJ7RLfs9quWil6h0S3Pl-r2JxMl2G7uEH2w49NVGsTOCQ8YDhcFkmU3dOFtUEksZ2D69i0SNrUI229rX7TG-V88Wsthma6xo_YiljWl00M0F0QBeTXySZE2kY-lYdyMUXO3Xg5ynz9q00pvn4hfby-kjilV_L0o3Wy_o9dORDiImrGuNGpqzHDObVimX4_B-96IcYF8XGzhUx1qDnJBEhjZ9-nbtN5-qJRFXItYpTWrPlE-rzGZuQhmYRAHq4AwZpsVfGPwrE77NOA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/rpJZaU33WFj5W20majjR2g-G3LM6AwcEt5anf6H8PBx4waZpITsbUl_t-LZd8ZGFwPwAfWEfjJwtftCvIvDgWG0iQ_pxZswZIX80mpxGCL2RG2dazolIzajrktrldOuJwFku7P9PiofSzdc2DdnV2DJZQ9ZS2vBHm4Eo04epOzmLO8r_eIGGp9W0qsuCPcc-5uHFVlJnmT60i5T5ilxTbRLhurdzYApnSlBp0_vs6v7LwXaff-fH3Ozh6Mz2KlMIFixOoiPqqqQG8ZD9Fn8nGElfxitp52XBi81ACLok-xRYYgd4pWQMtWtIZCMH0VkKJsE6IiICIX2h1MIoRN_3mw.jpg" alt="" loading="lazy"> <p>Dec</p> <p><a href="http://Mistral.ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://Mistral.ai/</a><br> 113M out of the gate. Think they gave away half the company.</p> <p><a href="https://t.me/chillhousetech/317?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Post #316 https://darkolorin.com/posts/316.html https://darkolorin.com/posts/316.html Tue, 03 Oct 2023 20:05:03 GMT <p><a href="https://greylock.com/edge/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://greylock.com/edge/</a></p> <p><a href="https://t.me/chillhousetech/316?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Should I pin, like, especially useful articles and posts in the channel? https://darkolorin.com/posts/315.html https://darkolorin.com/posts/315.html Tue, 03 Oct 2023 15:36:21 GMT <p>Should I pin, like, especially useful articles and posts in the channel?</p> <ul> <li>Yes</li> <li>No</li> <li>Naaah, just one more notification only</li> </ul> <p><a href="https://t.me/chillhousetech/315?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A top-tier hook. https://darkolorin.com/posts/313.html https://darkolorin.com/posts/313.html Tue, 03 Oct 2023 15:35:19 GMT <img src="https://cdn4.telesco.pe/file/YwRx3CzJuWLHTDHFX4aOg9s2ppjKnUqigd7AMnn8_HEWVtGiF1_XB1UrxKfxu4rMbU8gH40Scze1x1ge6Ryuq5Pua6V9c2McR8RW4dPfZ2mA6t62XQWkW9wljJOf2CL9YWzlrAcOwSZCDYgk6v1_yAeFEc6Gyjc_jYdKCYHS-FFXhVItKM4ElLjIR23V-1KPyyFG45Pkftb4C7nK_r597ZWwF4oNWkWGl86myTGzVNyF-T_dBO2wD60gQW7Gq4Evz0E0Sbw3J5Ibmr6FpKvzuQyhYFiMXv5v2pcHM-jcOElUYsrmpxdV6Rreli8p6D7HCwuUBCRbQi_BXMRtqSJYYg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/WXYznh_zVICTZyZG-KqwyoJIHhloR6EFyyd_PP6UOXSpIKA9gc0BKxrkoa3clPFsRR5wWJp3P_bNy4F2QwatX4yC8x_tEgytOeQR7hyXyQ8pFU44Yj29NwK2plr2I-kJoLrbJi4AscRUkHsnmrs9FmWFnhVcdwI1hvf3bBDSseM-C7-ZrkyPlJRiSlpgva29BTpO8eqYenkcLkXZwo8HPOEuuCX9SwhcxkeEVBYgQ-UmWyIT8oKx-w20mMZ9etA47tM_FjUCy5Y79zo2I4s-0FGlfYc8gudr2KjJ9JCbQRmc3kuosHmeSs5Qd1x3AKnsLsrJkj4QYBxpQT38mtItBA.jpg" alt="" loading="lazy"> <p>A top-tier hook.</p> <p><a href="https://t.me/chillhousetech/313?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, not an ad, it’s actually a good channel It’s just that I don’t… https://darkolorin.com/posts/312.html https://darkolorin.com/posts/312.html Tue, 03 Oct 2023 15:29:51 GMT <p>By the way, not an ad, it’s actually a good channel <br> <a href="https://t.me/exp_fest?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/exp_fest</a> <br> It’s just that I don’t like that there’s a lot of ads, events, podcasts, and some kind of crap. But if you, like, filter it out, there are solid articles and cool people in there. In general, barely anyone knows how to do real A/B tests. <br> As for filtering, btw, some kind of Discover with a smart feed would help, although it’ll have its own ads. Whatever, I’ll think about it.</p> <p><a href="https://t.me/chillhousetech/312?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Just an evergreen post: people go to a specific place and there it already… https://darkolorin.com/posts/311.html https://darkolorin.com/posts/311.html Tue, 03 Oct 2023 15:18:03 GMT <p>Just an evergreen post: people go to a specific place and there it already tells them what to watch. People pick the place first, and then the content, not the other way around. And yeah, the post is cool. <br> <a href="https://medium.com/s/story/why-we-fail-what-i-learned-from-5-years-with-friends-netflixs-social-strategy-9bbb9cb98608?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://medium.com/s/story/why-we-fail-what-i-learned-from-5-years-with-friends-netflixs-social-strategy-9bbb9cb98608</a></p> <p><a href="https://t.me/chillhousetech/311?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> OFFTOPIC — Couldn’t pass up the result of a poll in France (I didn’t check… https://darkolorin.com/posts/310.html https://darkolorin.com/posts/310.html Tue, 03 Oct 2023 14:27:55 GMT <p>OFFTOPIC — Couldn’t pass up the result of a poll in France (I didn’t check statistical significance), but the conclusions are fire, just like in Russia when they ask: 41% of the population of France is in favor of the proposal to limit people to four flights for their entire life.<br> <a href="https://csa.eu/news/etude-here-les-francais-et-la-proposition-dun-quota-de-4-vols-par-personne-dans-une-vie/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://csa.eu/news/etude-here-les-francais-et-la-proposition-dun-quota-de-4-vols-par-personne-dans-une-vie/</a></p> <p><a href="https://t.me/chillhousetech/310?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A funny minute: Batteries where you need like ~100 tons to power a Tesla for a… https://darkolorin.com/posts/309.html https://darkolorin.com/posts/309.html Tue, 03 Oct 2023 01:13:45 GMT <p>A funny minute:<br> <a href="https://x.com/andercot/status/1708980183771058422?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/andercot/status/1708980183771058422?s=46</a><br> Batteries where you need like ~100 tons to power a Tesla for a launch. But the topic’s curious, since they hold a charge for like 5000 years :)</p> <p><a href="https://t.me/chillhousetech/309?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For like, the people who are really big on theory, this paper is about how… https://darkolorin.com/posts/308.html https://darkolorin.com/posts/308.html Mon, 02 Oct 2023 18:33:25 GMT <p>For like, the people who are really big on theory, this paper is about how important network is in a VC’s work. And there are even those who raise funds on just that, and sometimes they’re even successful. And then they invest by the follow-the-top-tier model. <br> <a href="https://www.cis.upenn.edu/~mkearns/teaching/NetworkedLife/VC_networks.pdf" rel="noopener">https://www.cis.upenn.edu/~mkearns/teaching/NetworkedLife/VC_networks.pdf</a></p> <p><a href="https://t.me/chillhousetech/308?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> This whole thing with ChatGPT and prompts reminds me exactly of how people… https://darkolorin.com/posts/307.html https://darkolorin.com/posts/307.html Mon, 02 Oct 2023 17:15:45 GMT <p>This whole thing with ChatGPT and prompts reminds me exactly of how people were learning to search in Google in the late 90s and early 2000s. Even now, the problem that Yandex co-founder Ilya Segalovich used to say out loud still feels unresolved to me. Search shouldn’t be a skill, search is a tool that everyone should be able to use. There shouldn’t be some kind of query language.</p> <p>The first thing he wanted to do was islands — yeah yeah, those very answers inside search. And actionable, too.</p> <p>Another still-unsolved thing: contextual search. My favorite query: Messi’s goal was so beautiful a week ago. Well, roughly. Search shows really crappy results.</p> <p><a href="https://t.me/chillhousetech/307?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/306.html https://darkolorin.com/posts/306.html Mon, 02 Oct 2023 16:22:48 GMT <img src="https://cdn4.telesco.pe/file/FPKFURp8PHzKaAoRfYTWmemMIUuep1pZTu-MDRqywrSVGqpHe--1mn3DL6RCwXXtLQX5xc7WZBY46anLAHAy5HkC8yMiD5nSBkrf5XxL2aH3bEdo8ePTxJqaAurEtz3uzg2HyZmQlhSAp7FahkaDcLhi4RgpD8fRXt16LCqWMJJQWmtEm0I58LuYOxNLt4lnK4c2TARd7n5_JFVd6Kt5Svq2QRHSQdjvM82Se_fnqM_5o_Wf1y6H7i6SJzFNc1jTg6RHgBU_kdM0JlbAtGPgbnWmyn7RlsIHGJetpI2vIY65pyb28MKO9BHebIitb-pIcOUDRaPl2F9MPmHdGDpVeQ.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/306?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well why is everyone trying to do everything all at once, looking for where… https://darkolorin.com/posts/305.html https://darkolorin.com/posts/305.html Mon, 02 Oct 2023 14:08:23 GMT <p>Well why is everyone trying to do everything all at once, looking for where users will find meaning? Looks like it, yeah. The tactic of hunting for added value for a big user base is of course fine, but game streaming feels unnecessary to me.</p> <p><a href="https://twitter.com/elonmusk/status/1708776354597003329?s=61&amp;t=hAd206XAdd6EpZySOPE5fg&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/elonmusk/status/1708776354597003329?s=61&amp;t=hAd206XAdd6EpZySOPE5fg</a></p> <p><a href="https://t.me/chillhousetech/305?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> This is not unlike the deal YC represents, except you get more equity, less… https://darkolorin.com/posts/304.html https://darkolorin.com/posts/304.html Sun, 01 Oct 2023 17:47:04 GMT <img src="https://cdn4.telesco.pe/file/Ngo_N3OAQxYjjT8pk-PQ0zOnDBm64cBPw9p9MchQ0b1-H78PpYHJSbCT5N4EuiU2goL7_HVQ0Z0pwhLnz_lI8c9je3zXnX9ixDjBDzEyCkOxVsHUtpGoy0zu4DryNxFtSpEarDHYnzq_zq2lM-GkDd7Gr0PpAk9r42vcE3rY4VFi60PQfV_DkfquKfRbpTLZeQ_hxwZV-XYTfhznv59PeJReqLP8rqlEUyHYn5s7TqijOB9Z6zsCkTKoT0-66A53WYuoybcmfO4WqAfvAejkBp3V5bItO5SyowB3GOZbERE9uPcwmNWUTBxA6E7I1VkzNDA08TfPid_loKA5_F-8HQ.jpg" alt="" loading="lazy"> <p>This is not unlike the deal YC represents, except you get more equity, less food, and you should probably not give up after only 4 months. (C) Paul Graham</p> <p><a href="https://t.me/chillhousetech/304?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> On the importance of distribution: Ben Silbermann (founder of Pinterest) got… https://darkolorin.com/posts/303.html https://darkolorin.com/posts/303.html Sun, 01 Oct 2023 14:14:53 GMT <p>On the importance of distribution:<br> Ben Silbermann (founder of Pinterest) got his first initial users by going to the Palo Alto Apple Store and changing the Safari homepage to Pinterest on as many iMacs as he could until they kicked him out.<br> He also shoved paper flyers under doors, by the way.</p> <p><a href="https://t.me/chillhousetech/303?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Is the future behind new devices like humane? https://darkolorin.com/posts/302.html https://darkolorin.com/posts/302.html Sun, 01 Oct 2023 11:57:31 GMT <p>Is the future behind new devices like humane?</p> <ul> <li>Of course!</li> <li>I don’t even know</li> <li>No</li> <li>Bullshit from the start</li> </ul> <p><a href="https://t.me/chillhousetech/302?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There’s this startup it was founded by dudes who worked at Apple for 10–20… https://darkolorin.com/posts/301.html https://darkolorin.com/posts/301.html Sun, 01 Oct 2023 11:56:41 GMT <p>There’s this startup <br> <a href="https://hu.ma.ne/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://hu.ma.ne/</a> <br> it was founded by dudes who worked at Apple for 10–20 years and consider themselves Jobs’ heirs. So of course they went to make a new “iPhone” and they’ve already even popped up with the device at Paris Fashion Week. A big-name event too, right. <br> <a href="https://www.theverge.com/2023/9/30/23897065/humane-ai-pin-coperni-paris-fashion-week?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/2023/9/30/23897065/humane-ai-pin-coperni-paris-fashion-week</a> <br> Personally I don’t really get where all this finger tracking and “moving the device into a holographic” is going. It’s still just dumbly easier to pull out your phone than to hang something on your jacket that someone can take off you :) <br> I’m not a hardware pro, but over the weekend you can ballpark it. What do you think? (Poll separately)</p> <p><a href="https://t.me/chillhousetech/301?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, I mean, why not, a pretty decent pre-seed, you can drop a ten right away… https://darkolorin.com/posts/300.html https://darkolorin.com/posts/300.html Fri, 29 Sep 2023 18:28:56 GMT <img src="https://cdn4.telesco.pe/file/pBDHU3VnzoKqsIqOgPkLUCZ-l78Lv9VzBBy8cRKlaK49AK8GzJ2aUBM49tx3py_-NzvgPkXVbFSTJ8m9WjAqKf1JurnhWoWJNndPmeJu8C3Zad9bHZb6xt9o2Lift6Upn5mBsDOWHOCag-KR3j2lSJtM0G_3pdDBTgpg7_5zcGT1a98zYaXcU5ILxF-Nv3flKEa9Nt1RMA_L8t_zW769wO5cg8SCXDqEEC1aG3MRqgneSWjl10OJvhgRUKpM6eHywVE3HOWsEpMqOfDX0D79am0i6lNiHpw7W0qQ-7QFmj9iY6ilGTIGeXx4RnSbH5zATicWW9Y7NhbNZ3Gj3mlxUQ.jpg" alt="" loading="lazy"> <p>Well, I mean, why not, a pretty decent pre-seed, you can drop a ten right away too</p> <p><a href="https://t.me/chillhousetech/300?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Game dev king has spoken https://darkolorin.com/posts/299.html https://darkolorin.com/posts/299.html Fri, 29 Sep 2023 12:16:40 GMT <img src="https://cdn4.telesco.pe/file/MfTPjFMgsQsLLIgq9cCu7S02eA24QH5HlWTSrHQjdmFEwoQV3dqzWPAHE7QxVm4vr9AC2AXnIwB1qa9EdXLxlfd578NiFTnocyGNb6YXt_ex2qVxWUQq16AUkFVwBcGOkEHesGixxRz4Bt4agiZB0vq7N6hokHKD0_0B2uBD-w-RgLkrHDUWBVIX0WIbC0Acxz2hpdezer5JksDPXBUxu0rxEQGzVzGJOWaW8mH5cqt1DwU-_t25h9FVK2q_Sh3vJbrbiiKLhmQdfjlrkIMtZFKZz2mWEgLmYYXWaLNZQsKrZAfcdbt3JfycbhZkaF843Pu53i3b6wc6_m5J5psqSw.jpg" alt="" loading="lazy"> <p>Game dev king has spoken</p> <p><a href="https://t.me/chillhousetech/299?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Pretty often I hear that, like, everyone’s waiting for a new iPhone-style… https://darkolorin.com/posts/298.html https://darkolorin.com/posts/298.html Thu, 28 Sep 2023 23:22:54 GMT <p>Pretty often I hear that, like, everyone’s waiting for a new iPhone-style platform, i.e. a smartphone. First it was chatbots, then web3, then VR, now it’s already XR, and now LLMs are being painted as a platform (where the hell are those plugins?).</p> <p>So yeah, it seems like right now is a divergence moment, when we’re watching several big platforms pop up at once, just not as all-consuming as the smartphone.</p> <p>That’s why, among other things, we won’t see in the next few years any new products for literally everyone (social networks, search, and all that). I’ll package the second reason into a mini-post sometime too.</p> <p>Sure, there’s still a chance we’ll see something again on smartphones. But unlikely.</p> <p><a href="https://t.me/chillhousetech/298?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/297.html https://darkolorin.com/posts/297.html Thu, 28 Sep 2023 18:51:10 GMT <img src="https://cdn4.telesco.pe/file/JYM2tqz91j3280Le0imDPZAfi9GcbqNHVpP9bWcHsvXQQDdW3Jn0G_HS4SAwW697v57f--j7zJn36Z0QulZiQDFOVYFZNyg4EZzhCgiPwlPtL1va64By60r_W-C9dECrafKefC_e6mjqPU4O7_48GGinOXSX4FPUyiyNlBw5xUee0q2x0wFbSH-laWvzk3myRm5O81Nix0FlEAetdDTOLWcjdKtdozltc5FHDDxiWN-NGNyOkegZ2jeWqdEe0GgEzk7DuuuZ7k2nGE6ZRMmGfSmV9OutmXdh0fOM6cJLhPW_azdk6RKlYi8H5CV4mMTU29ZIFFgwaHcIlzs9UebR0A.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/297?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> VCGPT https://darkolorin.com/posts/296.html https://darkolorin.com/posts/296.html Thu, 28 Sep 2023 15:46:30 GMT <p>VCGPT <br> <a href="https://nfx.com/chat?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://nfx.com/chat</a></p> <p><a href="https://t.me/chillhousetech/296?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Not even kidding, it’s a necessary feature https://darkolorin.com/posts/295.html https://darkolorin.com/posts/295.html Thu, 28 Sep 2023 14:24:50 GMT <img src="https://cdn4.telesco.pe/file/RJGdoXuNmRNz-7q3SREj4bNpbG8J7ZCOHNBCyloqsBZ9jmDM0HXlXZ3Gi_NBVvmnWqqFu_rXs9SnIpR-nC9o3pJoiWVy2YEs9_zTWFD2kk0WnNB0x414p5rI8xh1I4nFdNnh9ZxRkYmHfb-UUXmbBQmC_cQi3poMzBtIHZkT8ZDISdvopsB3EwZOOjXtVY7w_diWUYokNX55tLPRMQo36hNOkIan_h8hvJPjtKjx03RkF5vXCsvef_mV1rzhrwqrp3ph_W5O9h7_HOwHIChYQ1KpYrAJ9xocG7UDTduPex52-7nJW81YtkiDN19mORypsMsX6l6ZHhCrvfiVhwe_fA.jpg" alt="" loading="lazy"> <p>Not even kidding, it’s a necessary feature</p> <p><a href="https://t.me/chillhousetech/295?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Finally https://darkolorin.com/posts/294.html https://darkolorin.com/posts/294.html Thu, 28 Sep 2023 14:02:15 GMT <img src="https://cdn4.telesco.pe/file/uDvdRGHax5PpBhLYppubKjylY4ElpOEKbNCYiEr_zqPqOplNrKwoUfp_ZJDv7x3fHxb7HBZWmyPZYBRBLTHjB73l-bl_bA0jAOOipvfRpk5d6rZ3u0gJicPSy-oRWiJfhFMLjRcOwBt9dmXchjnrcxSGzohKReJ6-Ogwfzwe4nV1uqmzpZRmhwubnPtc4Rk1qA8_Qf2rmAw8REtYzc5LKYWfLsXB4oRiqYbKX4hvT4MVmshXsmqNj4gZKRYoGSznMa5CfHXyLOln_TKY1hFtVQhoIQT7oZgzH9LUaGg7ug023KpMGDJdA15APiYUQ31BZCwK12JV-4VHKNTV9awNUw.jpg" alt="" loading="lazy"> <p>Finally</p> <p><a href="https://t.me/chillhousetech/294?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> When will we get a Jarvis from Iron Man? https://darkolorin.com/posts/293.html https://darkolorin.com/posts/293.html Thu, 28 Sep 2023 10:13:20 GMT <p>When will we get a Jarvis from Iron Man?</p> <ul> <li>1 year</li> <li>2–3 years</li> <li>5 years</li> <li>10 at least</li> <li>Never :)</li> </ul> <p><a href="https://t.me/chillhousetech/293?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And you guys are talking refocus or whoever else burned money. https://darkolorin.com/posts/292.html https://darkolorin.com/posts/292.html Wed, 27 Sep 2023 21:59:11 GMT <p>And you guys are talking refocus or whoever else burned money. Here’s what burning money looks like. <br> <a href="https://www.wsj.com/articles/amazon-aggregator-thrasio-engages-restructuring-advisers-ed1f0450?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.wsj.com/articles/amazon-aggregator-thrasio-engages-restructuring-advisers-ed1f0450</a></p> <p><a href="https://t.me/chillhousetech/292?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meta Quest 3 with no batteries and no cables. https://darkolorin.com/posts/291.html https://darkolorin.com/posts/291.html Wed, 27 Sep 2023 18:07:56 GMT <p>Meta Quest 3 with no batteries and no cables. And cloud gaming included. Works for me.<br> <a href="https://www.metaconnect.com/en/home?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.metaconnect.com/en/home</a></p> <p><a href="https://t.me/chillhousetech/291?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile in Indonesia they’ve banned paying for goods (e-commerce) in social… https://darkolorin.com/posts/290.html https://darkolorin.com/posts/290.html Wed, 27 Sep 2023 13:21:58 GMT <p>Meanwhile in Indonesia they’ve banned paying for goods (e-commerce) in social networks, supposedly to protect small business, now you can only advertise. Oof and this is just the beginning of a decade of regulations.</p> <p><a href="https://t.me/chillhousetech/290?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/289.html https://darkolorin.com/posts/289.html Wed, 27 Sep 2023 09:53:52 GMT <img src="https://cdn4.telesco.pe/file/lQTcZ2RYnbKZqwARogvNvSAO0FJvF0SN4bb1CpCyBI7zJwTys63Afv12XECgS2hUBVzmtrIs0zcWraqXVyJW9HIdbkMX1JIuPTXkwV1czLWdEa11Q9ekNz5h2Nv2NhN9o9-_O8-XTfD5bHWBfEQfZilKSHAoFdwd6EC1eSOH2jPYPKy1AX-THtW4cG2QdGuvZkWJwt05ugK7Xf7dcHrUSwmdlwobddH1dMzEUT7tqzuMIAg1iK-9H_BeRxPQqWmfnqGRMf70KlDMahkHv7lsWBb3r9Tbp7iNHKBdny0GWiqE2NuJW-W51nA2gtG0AdCbU2m_s3P_G7rGCZfh8A2Gzw.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/289?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> More and more Series B/C startups, i.e. https://darkolorin.com/posts/288.html https://darkolorin.com/posts/288.html Tue, 26 Sep 2023 21:11:39 GMT <p>More and more Series B/C startups, i.e. already pretty big (200+ people), just don’t hire PMs at all and don’t do A/B tests. Which is definitely for the better. <br> <a href="https://x.com/dsiroker/status/1706766014778650925?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/dsiroker/status/1706766014778650925?s=46</a></p> <p><a href="https://t.me/chillhousetech/288?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Recently the #1 app in the US store was Lapse. https://darkolorin.com/posts/287.html https://darkolorin.com/posts/287.html Tue, 26 Sep 2023 16:08:57 GMT <p>Recently the #1 app in the US store was Lapse. A little story about this product and the company.</p> <p>In reality the app’s been around for a couple years. And back in 2021 it was already in the top of the UK and US stores.<br> <a href="https://x.com/soundboy/status/1443653493563273219?s=46&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/soundboy/status/1443653493563273219?s=46</a></p> <p>The idea was basically the same as now, except for a few growth hacks. It’s a feed of photos with your friends that get deleted and then show up there after a certain amount of time. Remember Dispo, same thing. Only here you’ve got a shared “pack” for your friends. You collect 36 and then reveal in a day (well at least d1/2 retention ok).</p> <p>What’s interesting is that this is one of the most difficult and complex apps, with a number of hacks that just don’t stop surprising. Here we go:<br> - pyramid onboarding where for the app to even start you have to send 5 SMS invites, brutal<br> - it works starting from 8 friends overall (seems like the requirement changed over time)<br> - random pushes (when some critical mass of photos built up in the feed, like 36) and all of it with a personal caption just for you<br> - a like limit per photo</p> <p>You can tell the onboarding is really sharpened. They managed to drag in a huge mass right away using waitlists like “wait until there are 8 friends here.” That’s how they bump downloads. Plus in the beta there were 20k+ users who, right out of the gate, obviously sent their 5 SMS.</p> <p><a href="https://t.me/chillhousetech/287?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Another approach to diffusion, well, let’s say it’s not diffusion there, but… https://darkolorin.com/posts/286.html https://darkolorin.com/posts/286.html Tue, 26 Sep 2023 15:21:17 GMT <p>Another approach to diffusion, well, let’s say it’s not diffusion there, but the point’s the same. Physicists are trying to apply other techniques that give more variability. <br> <a href="https://www.quantamagazine.org/new-physics-inspired-generative-ai-exceeds-expectations-20230919/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.quantamagazine.org/new-physics-inspired-generative-ai-exceeds-expectations-20230919/</a></p> <p><a href="https://t.me/chillhousetech/286?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The most top-tier feature that needs to be added right now for wallet and… https://darkolorin.com/posts/285.html https://darkolorin.com/posts/285.html Tue, 26 Sep 2023 12:05:14 GMT <p>The most top-tier feature that needs to be added right now for wallet and Telegram is letting you promote posts in your own channels with ads прямо on mobile and pay from wallet. That’s exactly the move Instagram used back in the day. <br> I’m ready to tell you how to do it right :)</p> <p><a href="https://t.me/chillhousetech/285?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> so, who got how many users? https://darkolorin.com/posts/284.html https://darkolorin.com/posts/284.html Tue, 26 Sep 2023 11:55:26 GMT <p><a href="http://sillycovalley.com?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://sillycovalley.com</a><br> so, who got how many users? Screenshots in the comments</p> <p><a href="https://t.me/chillhousetech/284?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If you’re a startup, you shouldn’t have a growth team. https://darkolorin.com/posts/283.html https://darkolorin.com/posts/283.html Tue, 26 Sep 2023 08:32:17 GMT <p>&quot;&quot;&quot;<br> If you’re a startup, you shouldn’t have a growth team. The whole company should be the growth team, and the CEO should be the head of growth.<br> (C) Mark Zuckerberg<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/283?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Today a whole lot of useful stuff keeps catching my eye for some reason https://darkolorin.com/posts/282.html https://darkolorin.com/posts/282.html Mon, 25 Sep 2023 22:04:02 GMT <img src="https://cdn4.telesco.pe/file/Sas5eg7WCwiDPaRWcH799_0SCZ8y1MAj5PgUMu8sq_DfhsfX_U_o9_OPaXbnsUc3P9TUKaTFHACoufyMIFptHWbLxvQcsZ-sMb_IpwlRskNTuh-HOqhcoAr4J8lRF6COu_X9WPyrgF2Al3h9rCSHxPhWOOhN-sOAthG-4Mzo2Yvv66WU964r1q-d-bR6_1xuxTNpZKyXhvfAG1-MCTQO4pA9LEI0SGxcqyorI8D70cxG0yDVebqxgRm0Up2yt9CB_ccVJ5NFwQ7MSH3aEoUAJ0vVx3h5AJEPmOoRIJ7OCla8IuE1aSZUf94fsgMw66zXWqLGvzWLaJyL_DMxYBjENw.jpg" alt="" loading="lazy"> <p>Today a whole lot of useful stuff keeps catching my eye for some reason</p> <p><a href="https://t.me/chillhousetech/282?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> AnimateDiff by stable diffusion https://darkolorin.com/posts/281.html https://darkolorin.com/posts/281.html <p>AnimateDiff by stable diffusion</p> <p>Video available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/281?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Curious how it’ll play out on Reddit compared to the creators programs on Snap… https://darkolorin.com/posts/280.html https://darkolorin.com/posts/280.html Mon, 25 Sep 2023 20:53:28 GMT <p><a href="https://www.theverge.com/2023/9/25/23887226/reddit-gold-system-contributor-program?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/2023/9/25/23887226/reddit-gold-system-contributor-program</a> <br> Curious how it’ll play out on Reddit compared to the creators programs on Snap and the rest. Different approach, same point.</p> <p><a href="https://t.me/chillhousetech/280?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> It’s extremely interesting to read stories like this. https://darkolorin.com/posts/279.html https://darkolorin.com/posts/279.html Mon, 25 Sep 2023 20:46:50 GMT <p>It’s extremely interesting to read stories like this. Great investors, wise sages of the market. <br> <a href="https://www.reuters.com/article/idUS258723239520101112?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reuters.com/article/idUS258723239520101112</a></p> <p><a href="https://t.me/chillhousetech/279?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Off we go with applications of generative nets https://darkolorin.com/posts/278.html https://darkolorin.com/posts/278.html Mon, 25 Sep 2023 16:09:35 GMT <p>Off we go with applications of generative nets <br> <a href="https://techcrunch.com/2023/09/21/youtube-shorts-to-gain-a-generative-ai-feature-called-dream-screen/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://techcrunch.com/2023/09/21/youtube-shorts-to-gain-a-generative-ai-feature-called-dream-screen/</a></p> <p><a href="https://t.me/chillhousetech/278?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And I’ve always liked movie localizations because sometimes the dubbing was… https://darkolorin.com/posts/277.html https://darkolorin.com/posts/277.html Mon, 25 Sep 2023 15:44:33 GMT <p>And I’ve always liked movie localizations because sometimes the dubbing was actually cooler than the original. But overall it’s one of the few directions that’s actually worth anything. <br> <a href="https://www.theverge.com/2023/9/25/23888009/spotify-podcast-translation-voice-replication-open-ai?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/2023/9/25/23888009/spotify-podcast-translation-voice-replication-open-ai</a></p> <p><a href="https://t.me/chillhousetech/277?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Need more gold https://darkolorin.com/posts/276.html https://darkolorin.com/posts/276.html Mon, 25 Sep 2023 12:52:05 GMT <p>Need more gold <br> <a href="https://www.anthropic.com/index/anthropic-amazon?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.anthropic.com/index/anthropic-amazon</a></p> <p><a href="https://t.me/chillhousetech/276?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m watching closely how Discord’s monetization is getting rolled out in TG. https://darkolorin.com/posts/275.html https://darkolorin.com/posts/275.html Sun, 24 Sep 2023 12:31:59 GMT <img src="https://cdn4.telesco.pe/file/rt642q4MXUu5acQEiwVZ0nG8PsFETiP8UWmy9KKjB7_MkBV9YpaYDlrnCLeWG-QGwfInqgdEWB0EpcTwKj9p15M0Q0ofJNO41tblO4kS1jKrNqdziwckC-3sUp0SgMQDjoiH4XGnZAo_7kIR-P1FhLE58zdcJhniI0QdjrsNtT-pUaOOIbwBn-WF1uwBM9tFWBK7y_oVI5AFEqi5nH0-FBIhKDmLlmSUfhQN2qZyQAd_lk3_SfHL4S64n4_IZtZbwdmUCssE5DsMiBQfjwfBLojiY8WmYYAFr9FkaKw0uRY-LhTIInl7E1PN4X16UvH_xexevu2VfLvTqHMjfEg18Q.jpg" alt="" loading="lazy"> <p>I’m watching closely how Discord’s monetization is getting rolled out in TG. <br> 1) Nitro for users with extra features <br> 2) boosting servers/channels for various features (though right now it’s a weird implementation)</p> <p>TBD: <br> 3) channels buying boosts themselves <br> 4) channels releasing their own stickers so people join and buy a subscription</p> <p>Also Discord ran experiments with its own game store (see tonapps and all that) and it didn’t go all that well. <br> Overall everyone wants their own WeChat.</p> <p><a href="https://t.me/chillhousetech/275?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/274.html https://darkolorin.com/posts/274.html Sat, 23 Sep 2023 15:49:26 GMT <img src="https://cdn4.telesco.pe/file/A0eAnYoOlcivWKdD3qAHi5LeYQZZUhEZ8x3jphrRW4kjwSi9zALJZ7wK0P4FrxAJ254AHZDlcnoR0eDmV7AIH3qOmS0F2MoFFXOis0PNGrhfnBVnliSzk9pz_1F26qV2jOmxMmGRDFosAcFNYeilPoQZz6Vt3Ggr3xBEGXX080cSL3SPMwmDTkliZs5tWLD2cDyGeit2QSjsVe68cjfR5L_pi0OQyg8VAl4C9kfuvpITZFlh6gv74sy-mlL07sNsxPYfXpfoe1bPpuieU6pCHpqhtIzx8-zmo6u8z2ZymcIbwZ8-fM4u4L6EGXP5ZEsqaRS0mjrIynSiv9RfmNUuLQ.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/274?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Remember those things called SPACs? https://darkolorin.com/posts/273.html https://darkolorin.com/posts/273.html Sat, 23 Sep 2023 13:40:04 GMT <p>Remember those things called SPACs? <br> A small analysis came out <br> <a href="https://www.linkedin.com/posts/yuechenzhao_chamath-spac-investing-activity-7111059130328522752-vRs_?utm_source=share&amp;utm_medium=member_ios" rel="noopener">https://www.linkedin.com/posts/yuechenzhao_chamath-spac-investing-activity-7111059130328522752-vRs_?utm_source=share&amp;utm_medium=member_ios</a></p> <p><a href="https://t.me/chillhousetech/273?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Let’s talk about nostalgia. https://darkolorin.com/posts/272.html https://darkolorin.com/posts/272.html Sat, 23 Sep 2023 12:53:14 GMT <p>Let’s talk about nostalgia. I have a hypothesis that a lot of consumer social services in the future will be built around the feeling of nostalgia.</p> <p>What’s the trick here? We’re stockpiling a huge, massive amount of memories in our phones; for gen z basically every day of their lives is documented either in messengers or in photos and videos. And when it hits you, you want to go back, feel the same things. So yeah, that’s where the entry point for a new product will be. And as we know, the life moment matters when new companies show up.</p> <p>Facebook — you’re going to uni. <br> Discord — you’re going on a raid. <br> LinkedIn — you’re looking for a job.</p> <p>Basically I think nostalgia and working with it is one of the few strong use cases for VR. And it’s massively underestimated.</p> <p><a href="https://t.me/chillhousetech/272?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/271.html https://darkolorin.com/posts/271.html Fri, 22 Sep 2023 20:20:47 GMT <img src="https://cdn4.telesco.pe/file/HOSz8OT3iXr9ma4Yzq2A4NK4CVV7hmZIi1lLXbBTw7jchb13NutiZXVgIlGBmx-TMvn_l7tD2789VFen81JM_EaAiUSMNWro6VxwO1kxn2AywYBpoSTHCENUA1_LGLgm9cEelxFGxQ_S1lFM-L0clmJOTKI_OheheANCsVP6J-6V1oGZnecfF_3rT0WByze20j1HUbLOIxbMjuWBDoAD9FyWf9Tymw_kurFZksi_FYpOv14tqzde-XHjr8yfplj0BSvQB9-4vz6oDMBEDCsim0wBmHwTYAdJwqWvV0_5JknXfToZj-0584UEqCRk2FFTAPhXXV3INOWDo9aXWhgyTg.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/271?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Expected. https://darkolorin.com/posts/270.html https://darkolorin.com/posts/270.html Fri, 22 Sep 2023 18:12:12 GMT <p>Expected.</p> <p><a href="https://blog.unity.com/news/open-letter-on-runtime-fee?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.unity.com/news/open-letter-on-runtime-fee</a></p> <p><a href="https://t.me/chillhousetech/270?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sequoia’s little report on AI TLDR: tech is tech, data is data, but they still… https://darkolorin.com/posts/269.html https://darkolorin.com/posts/269.html Thu, 21 Sep 2023 17:44:50 GMT <p>Sequoia’s little report on AI <br> <a href="https://www.sequoiacap.com/article/generative-ai-act-two/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.sequoiacap.com/article/generative-ai-act-two/</a> <br> TLDR: tech is tech, data is data, but they still think the user is in the center and the winner will be whoever grabs users’ and customers’ attention. Well yeah, basically nothing changed.</p> <p><a href="https://t.me/chillhousetech/269?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Google will win https://darkolorin.com/posts/268.html https://darkolorin.com/posts/268.html Thu, 21 Sep 2023 17:18:58 GMT <p><a href="https://blog.google/products/bard/google-bard-new-features-update-sept-2023/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.google/products/bard/google-bard-new-features-update-sept-2023/</a> Google will win</p> <p><a href="https://t.me/chillhousetech/268?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Next version of dalle but I still think midjourney is cooler https://darkolorin.com/posts/267.html https://darkolorin.com/posts/267.html Wed, 20 Sep 2023 18:33:56 GMT <p>Next version of dalle <br> <a href="https://openai.com/dall-e-3?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/dall-e-3</a> <br> but I still think midjourney is cooler</p> <p><a href="https://t.me/chillhousetech/267?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Little by little everyone wants their own WeChat again, they just never get… https://darkolorin.com/posts/266.html https://darkolorin.com/posts/266.html Wed, 20 Sep 2023 08:48:42 GMT <p><a href="https://www.reuters.com/technology/whatsapp-adds-rival-in-app-payment-options-india-commerce-push-2023-09-20/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reuters.com/technology/whatsapp-adds-rival-in-app-payment-options-india-commerce-push-2023-09-20/</a><br> Little by little everyone wants their own WeChat again, they just never get tired of it</p> <p><a href="https://t.me/chillhousetech/266?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/265.html https://darkolorin.com/posts/265.html Tue, 19 Sep 2023 22:36:46 GMT <img src="https://cdn4.telesco.pe/file/uSidzoXHWeB89e4XenSYH30ngQSI7CPwToSUSMOStjX4y-G1BFWktLKKJu3YGJb6fd1EnEQAwKcWzMSNnMUmibArVobyQSoBtK4CpEjMJczSUuzfA-Ey4TacyZDGW59FuOf2QLDQ05p1yff4xc3GlHqfw8wdlGTjbeXSlmuyJv8As2S7l_l97BXUtsYNQFt5e6vcEKCJck7HDHyFevjcl8YBbYne-z7CYOVvyomRovI2zzGkcXWZu-tpI8KP-R3Ob0cgr18C5z_InOzEYPJjjd42n87PVoMVompNdQhMZQT171aM6nQdGY5VPsrQguOUUL4z2DSakBXftATB5Xfxwg.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/265?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Google will win https://darkolorin.com/posts/264.html https://darkolorin.com/posts/264.html Tue, 19 Sep 2023 14:38:49 GMT <p><a href="https://blog.google/products/bard/google-bard-new-features-update-sept-2023/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.google/products/bard/google-bard-new-features-update-sept-2023/</a><br> Google will win</p> <p><a href="https://t.me/chillhousetech/264?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Heh https://darkolorin.com/posts/263.html https://darkolorin.com/posts/263.html Mon, 18 Sep 2023 16:01:12 GMT <p>Heh <br> <a href="https://unitedgamedevs.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://unitedgamedevs.com/</a></p> <p><a href="https://t.me/chillhousetech/263?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, looks legit, quote next: https://darkolorin.com/posts/262.html https://darkolorin.com/posts/262.html Mon, 18 Sep 2023 01:55:59 GMT <p>Well, looks legit, quote next:</p> <p>Here&#x27;s an example of a typical successful startup journey (US) recently:<br> - 2–3 cofounders, idea, customer interviews: raising $500k–$1M pre-seed on a $5–6M SAFE cap.<br> - 6–12 months later, with a team of 5–8, and early revenue ($50–300k ARR): raising $2M seed on a $9–12M valuation (likely SAFE, could be priced). Some companies call this &quot;pre-seed&quot;, however I find &quot;seed&quot; more accurate for a company with traction. That $2M cannot be enough to get to $3M ARR which will take at least 2–4 years, so..<br> - 12–18 months later, with $1M+ ARR, the company raises their second seed: $3–6M on a $15–30M valuation, most likely priced, with a lead and a new board member. However, we also see this sometimes as an extension in 2023, with no new board members, and could be on the same terms as the previous seed (if previous seed was overpriced).<br> - Another 12–24 months later, the company claims that they reached PMF with $3M+ ARR, and goes and raises a $10M+ Series A at a $40–100M valuation.</p> <p><a href="https://t.me/chillhousetech/262?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/261.html https://darkolorin.com/posts/261.html Sun, 17 Sep 2023 12:10:07 GMT <img src="https://cdn4.telesco.pe/file/fOIa19M80sIZw_sN_tdMfWtbT8nK2bIXRRXlScoJ-YAEb5yfmCaYttGI3ZkR3bMLCTb1Q5fTAVTYMrQCNK-SMgoDkux41PC-rP2l1gubXCuPMkpZIOiOd1X94LDDUSpbRSDV0fg8JRVT9eZJhAdCjAirB2G8DltWkTusyp-_lgEQfC4hghScTlTvsW27DHaXCRI0rzsBqnEdLHTUsx8SBM33Jfi0Pb-Yxa8ZxZ9wvaVXfRDbJ62NSkQMDqc6ah_TasA3tS5GUs4Ko8x9RQHurwOdvRNtdNUHUmXfAHqngpmJgeLYjV8ubhYm0GWz2dGMZ8F514Ctz0nCLaNKUhPIEg.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/261?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> That very article where the term “hallucination” popped up in conversations… https://darkolorin.com/posts/260.html https://darkolorin.com/posts/260.html Sat, 16 Sep 2023 20:19:00 GMT <p>That very article where the term “hallucination” popped up in conversations about neural nets. And in general, already in 2015 you’re hearing words like attention, reasoning, memory, and so on. This is just to say: the ideas are hella old, and the guys were in the topic a long time ago. <br> <a href="http://karpathy.github.io/2015/05/21/rnn-effectiveness/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://karpathy.github.io/2015/05/21/rnn-effectiveness/</a></p> <p><a href="https://t.me/chillhousetech/260?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> “Typical startup” column. https://darkolorin.com/posts/259.html https://darkolorin.com/posts/259.html Sat, 16 Sep 2023 19:06:58 GMT <img src="https://cdn4.telesco.pe/file/igl3uhg04uyMDLWm66-rAxZbbbT_5sr-dBo3x2_Irq8kSDhQNk-G1yjE8eypub4jHSsZxUDK5XRADY4pF8R7PDixxZL4qI_5KHCkCBpCWzdvxfi3KZ75Hzd7JSrCmAVupbeYLCbJoScYCJZ7uDdN9XgjGfqUNPk1r_WM2gf1hyyDNllN2niMP7Gjb3mj8EcyNo2Y0NUSHYPr83mVw-93ETD54b83_50SfQSGwXTXfGvKxZDpFIQdGVkFv0xed_x9wHKWdNW-Wvwl27hg2vi0z4NVU6WlVXDUg2r1T0bGIGrWgBAEWmI6neWdK_NphyVSGqnqmL_2gWOpPm6sR3-wJg.jpg" alt="" loading="lazy"> <p>“Typical startup” column. Still saying: you should be building, not selling courses. <br> You do a startup — that’s like hiring a team, figuring out what to do, coding, making designs, writing tasks, all that. But in the modern world I keep seeing more and more that “building” a startup == raising money or getting into some accelerator. Doesn’t even matter whether a product will come out at all, the main thing is the achievement.</p> <p><a href="https://t.me/chillhousetech/259?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Now *that’s* what I call guys getting ready for a showdown in advance https://darkolorin.com/posts/258.html https://darkolorin.com/posts/258.html Sat, 16 Sep 2023 18:52:56 GMT <p>Now *that’s* what I call guys getting ready for a showdown in advance <br> <a href="https://www.bloomberg.com/news/newsletters/2023-09-14/googlers-told-to-avoid-words-like-share-and-bundle-us-says-lmj27bhl?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bloomberg.com/news/newsletters/2023-09-14/googlers-told-to-avoid-words-like-share-and-bundle-us-says-lmj27bhl</a></p> <p><a href="https://t.me/chillhousetech/258?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Advisory shares https://darkolorin.com/posts/257.html https://darkolorin.com/posts/257.html Fri, 15 Sep 2023 19:23:44 GMT <img src="https://cdn4.telesco.pe/file/ewIp60d-Gr9BLcqVt8Fjbb8vN6JJEgsbxLm5Ds0nHQf6YNsxcLB6XQfj4ZVt8UJCKtryqTG3Tn6eJAyhhyMDbonqiCsrr6rOAVQqkJLcEvyB2iLHsjJ-TRX5JUYB-xHibAwI880T9oWPP3QBRE_43ChsSGriQSzxcqzQpPBswNVlrkCVclj8O5kHt493k0na9bm95Cj45VM9QIvuyQ0FisRIT44GXbchhvW6EcgA2UPfzJQG_4REMxXXhLfn_beA13KsXDlyRMX5quHLe9H4NK7vQHimVaX023UIwyEnyMdO541s0kLr3rVqWX8sevs8SRaEXmWloJl454GQV_h5pw.jpg" alt="" loading="lazy"> <p>Advisory shares</p> <p><a href="https://t.me/chillhousetech/257?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And here we go, the heads of ad networks and platforms have started posting… https://darkolorin.com/posts/256.html https://darkolorin.com/posts/256.html Fri, 15 Sep 2023 17:55:21 GMT <p>And here we go, the heads of ad networks and platforms have started posting too — just don’t take a cut from the install, okay, or they’ll slash all the budgets <br> <a href="https://www.applovin.com/blog/my-views-on-the-unity-price-change/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.applovin.com/blog/my-views-on-the-unity-price-change/</a></p> <p><a href="https://t.me/chillhousetech/256?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, Apple pulled a great lifehack for boosting revenue: - they removed… https://darkolorin.com/posts/255.html https://darkolorin.com/posts/255.html Wed, 13 Sep 2023 19:27:56 GMT <p>By the way, Apple pulled a great lifehack for boosting revenue:<br> - they removed the cheapest option in the most expensive model. That automatically bumps people who were just dumbly dumping stuff into iCloud into the next tier. And there you go, plus a hundred bucks off a couple million units.</p> <p><a href="https://t.me/chillhousetech/255?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Refocus fell apart over there, Refocus. https://darkolorin.com/posts/254.html https://darkolorin.com/posts/254.html Wed, 13 Sep 2023 18:52:58 GMT <p>Refocus fell apart over there, Refocus. Dunno what went down and how, I don’t know. But I can say what red flag I saw. <br> Founders of a successful early-stage startup are posting courses, giving interviews, telling everyone what’s the right way and what to do, basically running a heavy PR campaign for themselves. <br> To me the hardest moment for a company is to keep the traction you got, and that scaling/growth moment is brutal as hell. And it’s not compatible with what they were doing publicly. IMHO. The rest is just a preface and it doesn’t really matter what the problems actually were.</p> <p><a href="https://t.me/chillhousetech/254?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Ffs, traffic games are screwed https://darkolorin.com/posts/253.html https://darkolorin.com/posts/253.html Tue, 12 Sep 2023 20:22:59 GMT <p>Ffs, traffic games are screwed <br> <a href="https://blog.unity.com/news/plan-pricing-and-packaging-updates?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.unity.com/news/plan-pricing-and-packaging-updates</a></p> <p><a href="https://t.me/chillhousetech/253?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Dunno. https://darkolorin.com/posts/252.html https://darkolorin.com/posts/252.html Tue, 12 Sep 2023 18:22:27 GMT <p>Dunno. Apple keynote — nothing to say. Been bland for ages. They’d be better off releasing some software and showing that.</p> <p><a href="https://t.me/chillhousetech/252?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A healthy LinkedIn for a healthy person. https://darkolorin.com/posts/251.html https://darkolorin.com/posts/251.html Tue, 12 Sep 2023 17:48:23 GMT <img src="https://cdn4.telesco.pe/file/DpqTDtjIxi2V6VNBkCqXN2i31z8CTfW2PuEgJIfe3OrNsp3gaxw9Ax_W4EqnhkF0iopZFre67iWAGiemkDUt-YBmyAoNtZD3csPfR-1cr0xLAyHWgaD5bMe78DDmrJVk4jaAtHLRd_podbTqwapzq304YrIFC_wrjyS1wjR8ineKIi-w_wuXLaL904z15bRiWGvu-NLrjGnuDKAIYwwuoVGtINDEEoHuXoltVfkAd4cQ1UFCIy75PEx7G0Lt0apPYEvWWaG5engt6RybEShxztfHe_KZllnCYQLANN3krJQYIhIIy5QChKr2p3TdkK-iZYlapEfYLjXstau5ER3ZPA.jpg" alt="" loading="lazy"> <p>A healthy LinkedIn for a healthy person.</p> <p><a href="https://t.me/chillhousetech/251?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For the record. https://darkolorin.com/posts/250.html https://darkolorin.com/posts/250.html Sat, 09 Sep 2023 22:24:06 GMT <p>For the record. How Dropbox was discussed on Hacker News. Arguments, criticism. <br> <a href="https://news.ycombinator.com/item?id=8863&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://news.ycombinator.com/item?id=8863</a></p> <p><a href="https://t.me/chillhousetech/250?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The guys scraped together a measly 126m on seed and immediately moved to France. https://darkolorin.com/posts/249.html https://darkolorin.com/posts/249.html Sat, 09 Sep 2023 20:37:12 GMT <p><a href="https://sifted.eu/articles/poolside-raises-126m-relocated-france-news?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://sifted.eu/articles/poolside-raises-126m-relocated-france-news</a> <br> The guys scraped together a measly 126m on seed and immediately moved to France. That’s how the state, in the form of billionaires, pulls companies in.</p> <p><a href="https://t.me/chillhousetech/249?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile they’ve finally started optimizing for LLM. https://darkolorin.com/posts/248.html https://darkolorin.com/posts/248.html Sat, 09 Sep 2023 19:05:05 GMT <img src="https://cdn4.telesco.pe/file/B12jxNsNp3T6yq7l1P2EFdRAPEnjbWI6wq4XPcL7amKvEXx4LqnfixAdFZnQpreEE5ElACcVrNM9zdkttUgQ6-F6-Rpn01DPXAECaVTQdtJiHGBqevEvyzYf6opoplMwZiiBMFqwbcRfsEbBc_n6VFENqvMvuEGKS5IXKjSxYjNw6gIj-WkrQ0M4SzHeEcbGi1JDQfEruJIxAFsiJMmXkqf9JNzOdlBEOByxrMe3aFpj7iQuHNgxBZLleN5er0yMC8Q1cMDNqP51A7Ikd4Bs2DtFcS64vNFZqAogm0VMBqDiTCLzUJc31VPa3T_VPrC0-T08ti93ZZLjC6J_hamiyQ.jpg" alt="" loading="lazy"> <p>Meanwhile they’ve finally started optimizing for LLM. Overall I’m expecting lower costs and, accordingly, a speedup of this whole thing, including pretrain, by like 2–4x by year-end. <br> P.S. I think we should short Nvidia in Q1 2024.</p> <p><a href="https://t.me/chillhousetech/248?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, already 299, churn won:) https://darkolorin.com/posts/247.html https://darkolorin.com/posts/247.html Sat, 09 Sep 2023 18:56:26 GMT <p>Well, already 299, churn won :)</p> <p><a href="https://t.me/chillhousetech/247?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Let’s mark the nice round number of 300 subscribers! https://darkolorin.com/posts/246.html https://darkolorin.com/posts/246.html Sat, 09 Sep 2023 18:35:58 GMT <p>Let’s mark the nice round number of 300 subscribers! Thanks, everyone! It’s cozy here :)</p> <p><a href="https://t.me/chillhousetech/246?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A bit of AI trends, but specific ones https://darkolorin.com/posts/245.html https://darkolorin.com/posts/245.html Sat, 09 Sep 2023 18:13:08 GMT <p>A bit of AI trends, but specific ones <br> <a href="https://epochai.org/trends?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://epochai.org/trends</a></p> <p><a href="https://t.me/chillhousetech/245?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> It’s kinda funny writing scripts like this in general. https://darkolorin.com/posts/244.html https://darkolorin.com/posts/244.html Sat, 09 Sep 2023 17:27:05 GMT <p><a href="https://github.com/nat/thingies/blob/main/slackpaste.py?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/nat/thingies/blob/main/slackpaste.py</a> <br> It’s kinda funny writing scripts like this in general. Wonder if you could ship some little product like that, where you basically tune it for yourself and it runs some typical actions through GPT.</p> <p><a href="https://t.me/chillhousetech/244?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Gonna drop here a post by the founder of Loom. https://darkolorin.com/posts/243.html https://darkolorin.com/posts/243.html Sat, 09 Sep 2023 15:49:58 GMT <p>Gonna drop here a post by the founder of Loom. Feels like he gets across a super simple and correct point: don’t chase some “genius idea.”<br> ————————————————————<br> How 9 months of building without success, 2 pivots, and 100s of rejections resulted in Loom:<br> In reflecting back on the last 8 years, I&#x27;ve become convinced that most founders place too much emphasis on finding the &#x27;perfect&#x27; startup idea.<br> This thread covers the story of Loom—and why you&#x27;ll probably have to start with an imperfect idea before landing on your winner:<br> Loom didn’t begin as Loom.<br> It started around a whiteboard with Vinay, Joe, and myself. We wrote down six ideas, and went with the solution that had the lowest risk &amp; barrier to entry:<br> A user testing marketplace. We began building. <br> 7 months in, we only made $600 and learned an important lesson:<br> Companies generally cared less about advice from experts (what we were selling). Instead, they wanted to hear directly from their own users.<br> We pivoted. The new idea?<br> Build a SaaS offering that’d let companies get direct feedback from *their own users*.<br> As Vinay started to rebuild the product, Joe and I spent weeks cold emailing 300 different companies offering a free trial...<br> Soon after—we heard back from a research team at Harvard.<br> They used our product (called OpenTest at the time) to gather insights from 7 students for a campaign.<br> As a result—they received 7 different videos with various insights.<br> Instead of sharing these videos back with the entire research team...<br> Someone at Harvard used the same Chrome extension (built originally for the user testers themselves) <br> to record a 1-minute summary of the 7 videos.<br> That 1-minute video is what they shared back with their team.<br> This wasn’t how our product was supposed to be used.<br> Yet it was a valuable insight: We realized that people were more interested in using Loom as a communication tool.<br> When we were able to realize that...<br> We took the screen and video recording extension, launched it on Product Hunt in June 2016, and had 3k people sign up in the same day.<br> Which was 2,990 more users than we’d had in the 9 months prior.<br> A couple things one can take away from our experience:<br> First, de-stress yourself from “finding the perfect idea”.<br> It’s impossible to predict how things will go once you start—and it’s better to start and adapt than never launch because you’re worried something won’t work.<br> Just like cooking, playing guitar, or learning a sport, coming up with the perfect startup idea is something you likely won&#x27;t get right the first time you try it.<br> Second, the three traits that got us to Loom were:<br> 1) Intense focus<br> 2) Timing of the market<br> 3) Pure luck<br> You might build a beautiful product—and it could be good in theory—but without those three things, you&#x27;ll need to pivot to something that works.<br> That’s the story of how we came up with Loom: An idea that worked, after a couple that didn’t.<br> As you build your ideas, you&#x27;ll hear a lot of advice from outsiders giving you good and bad advice.</p> <p><a href="https://t.me/chillhousetech/243?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> On the topic of mentoring, coaches, and all that. https://darkolorin.com/posts/242.html https://darkolorin.com/posts/242.html Fri, 08 Sep 2023 20:08:05 GMT <p>On the topic of mentoring, coaches, and all that. <br> The main question, as it seems to me, is this: who’s responsible? And since, as I was told, a mentor in a project or at a manager’s should be for the long haul. And like, he helps with strategy from the side, from his experience and whatever. Meaning he kinda helps, roughly, to build the company? Right? Then let them get 1k bucks a month and a tiny 0.5% like an advisor. Because that’s fair, isn’t it? <br> The person doesn’t do anything with their hands. And if he’s helping a startup, then here’s its future that you signed up to support. Otherwise these guys want 10k bucks a month for their experience, or like 5k. <br> The other side of the coin is, like, what’s the motivation of these people in general? If you’re that cool and experienced, then clearly you didn’t come to chop up cash here and now off an early-stage startup. Orrrr? Or I don’t know something. But cool dudes are usually ready to help if they’re simply interested and money isn’t a big deal at all.</p> <p><a href="https://t.me/chillhousetech/242?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh, Shorts in the Google store, guess they got tired of the whole Discover… https://darkolorin.com/posts/241.html https://darkolorin.com/posts/241.html Fri, 08 Sep 2023 19:27:48 GMT <p>Oh, Shorts in the Google store, guess they got tired of the whole Discover being somewhere else <br> <a href="https://techcrunch.com/2023/09/06/google-is-putting-short-videos-on-play-store-to-aid-app-discovery/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://techcrunch.com/2023/09/06/google-is-putting-short-videos-on-play-store-to-aid-app-discovery/</a></p> <p><a href="https://t.me/chillhousetech/241?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Friday stuff. https://darkolorin.com/posts/240.html https://darkolorin.com/posts/240.html Fri, 08 Sep 2023 10:20:47 GMT <img src="https://cdn4.telesco.pe/file/W2piYdvK-XiFLE4tpepWPfja9BaZRFujCKUUc9Lsf8oP3Fzj3Bfci5YisznKhSOn0ZVZmgTWycIjUYfh4oUNcUPHzeV0JlSuPSRGIl5Z4_dAhWcpPEbW75zaibPROLBWciK3dWS-07Osc34H-p4xovXiqu0lu6MT9YXNULwOfIcE8JdDSII8PtiA_6h6yD13eQtXybGiPRQVTQZX83x5BqlL-fJmm1bdhg6h45FcND_e6McCCtWKJ91NHLNQngsAqFIOlQaoTnx4wTBBgkB_n3U4x93JSMyHdKWosqIvVSUbvY6HzfEiD3PtGuitoo2er24F7Kob2u7xabkq1pMPHg.jpg" alt="" loading="lazy"> <p>Friday stuff. Well, honestly it’s a standard situation. More than that: the function of product managers is optimizing, in most cases, what already exists. Doing something new is rare, and a big one. So dreams often crash.</p> <p><a href="https://t.me/chillhousetech/240?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> To understand how top funds think. https://darkolorin.com/posts/239.html https://darkolorin.com/posts/239.html Thu, 07 Sep 2023 23:41:45 GMT <p>To understand how top funds think. And of course god forbid they go back to the terminal interface. <br> <a href="https://greylock.com/greymatter/seth-rosenberg-product-led-ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://greylock.com/greymatter/seth-rosenberg-product-led-ai/</a></p> <p><a href="https://t.me/chillhousetech/239?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> This is one of those investors you can just DM on Twitter and he’ll actually… https://darkolorin.com/posts/237.html https://darkolorin.com/posts/237.html Thu, 07 Sep 2023 23:37:19 GMT <img src="https://cdn4.telesco.pe/file/YkB7cDK4ele1EWGib98Qo1JgGOperejlgIpAsB_A4NRCto84cstRa66jgAeal3tps7893tt8kD8vet6L-Cut9LuZUXV9gvf3xprY_GRp468nPO2KRs8-2dxwx4tcrk_pKzDDe-shFSflYRoWydNQ-cr_joiZlT0ngQ_IdYW-d4yCORVNyVpTNx3LlMl8hYkcj7K96m14ES2pcfnZ32To1HCssb-sJV3J96_pIayjldQKazlh13XB2eJSHc0MLbjW9PZs-HLwIfX3ZmdRztaB-vntwNNyerWtp_s1og5TNb1oxm3ng4u9vEC00XSImsNh4r8VVIHH19MbfG4asnzoVQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/jpsiajahh9jwpbQpykO5Aa0ICiowN2ivBJTvEGcYwSCVLjDvr3uqbOsSgN7Nb9Cg1T7TzgeafAsCDZ4OGJnMSmYQFm5puoKx_83l959fph2-PoE753p4yWEYzcCAqiCn-mTdjPto_QGZYydQsyNo5RLkmO3wtYW_muz7qINEU-9h4AC7xCIxdE1jlLT8m30MycS-LqgJ4Sk0252CsOke08Guo8pPf5PMTBaPUWqmd_frHI6QovfJ8Ur1tJQEGxx2GJn5D7hqvfKpfEk-3AXK6m6UC1ofZcFJ804LSUqEctZZ3r8KmvD6OOuYZFYjr01r9si8dTRi6C2yWJTDP8gt3g.jpg" alt="" loading="lazy"> <p>This is one of those investors you can just DM on Twitter and he’ll actually take a look.</p> <p><a href="https://t.me/chillhousetech/237?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And here’s Apple’s accelerator for devs https://darkolorin.com/posts/236.html https://darkolorin.com/posts/236.html Thu, 07 Sep 2023 23:17:00 GMT <p>And here’s Apple’s accelerator for devs <br> <a href="https://developer.apple.com/entrepreneur-camp/?cid=c00086-M00164-EN&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://developer.apple.com/entrepreneur-camp/?cid=c00086-M00164-EN</a></p> <p><a href="https://t.me/chillhousetech/236?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Every time I click into the profile of some Russian-speaking investor, like… https://darkolorin.com/posts/235.html https://darkolorin.com/posts/235.html Thu, 07 Sep 2023 22:58:11 GMT <p>Every time I click into the profile of some Russian-speaking investor, like half of them have Brex, SpaceX and all that stuff in their portfolio. As I understand it, to stand in the same line with them all I need is to buy a couple secondary shares tomorrow and also put in my experience — investor in Palantir Technologies.</p> <p><a href="https://t.me/chillhousetech/235?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Friends are making a cool game project where the levels are UGC, and the… https://darkolorin.com/posts/234.html https://darkolorin.com/posts/234.html Thu, 07 Sep 2023 00:51:55 GMT <p>Friends are making a cool game project where the levels are UGC, and the levels come in a feed kinda like TikTok. Nice. Just need to polish up the actual game mechanics a tiny bit more. But conceptually I really like it. <br> <a href="https://apps.apple.com/app/id1659966642?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/app/id1659966642</a></p> <p><a href="https://t.me/chillhousetech/234?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh, Clubhouse is still alive, and it’s getting redone too https://darkolorin.com/posts/233.html https://darkolorin.com/posts/233.html Thu, 07 Sep 2023 00:48:16 GMT <p>Oh, Clubhouse is still alive, and it’s getting redone too <br> <a href="https://www.theverge.com/2023/9/6/23861862/clubhouse-reinvents-itself-audio-messaging-app?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/2023/9/6/23861862/clubhouse-reinvents-itself-audio-messaging-app</a></p> <p><a href="https://t.me/chillhousetech/233?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Big businesses require either (a) a product people want to use pretty often… https://darkolorin.com/posts/232.html https://darkolorin.com/posts/232.html Tue, 05 Sep 2023 22:48:30 GMT <p>Big businesses require either (a) a product people want to use pretty often, or (b) very high value created when they use it. <br> Truly great businesses, of course, combine both. <br> If you have neither, it’s time to change your product until you’ve got at least one. <br> And yeah, it’s easier to turn usage frequency into value than the other way around. <br> Quickly translated this thought from Twitter from Twitch’s cofounder. Agree 100%.</p> <p><a href="https://t.me/chillhousetech/232?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/231.html https://darkolorin.com/posts/231.html Tue, 05 Sep 2023 22:04:28 GMT <img src="https://cdn4.telesco.pe/file/U93HBU68DKKs0CCRlqAjN2yGQb4QqGpqUGfgEzU4v4pat-P1p5VULELOWiAGaZKxI1UOorwM_zDXWMTd79BfnWncrRonWG29rhLYwlixd_d0RU-KnVBaZlQnB7lOmQ2Mjwc9ZSrcpy8bzEw3aQW474r9_xrAg4P1alFm32sqqKuOwnJu8S4LIMcpwAeDV833kc3re_hRkp3dViF_WCb3Or8_375GfW_AGMCqDrmz2Vd119OeUCPsb0wwZphCNAL98v-dUaA9dmKWrrME6p8nzG0ifYTEnzrnfuxFbU07wS_tuSNKPM_OEJT0yX5GlGv1pyZeAcJS-NFWCQXPoxrseQ.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/231?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> That’s roughly how I think about a lot of those showcase stories from the… https://darkolorin.com/posts/230.html https://darkolorin.com/posts/230.html Tue, 05 Sep 2023 01:24:24 GMT <p>That’s roughly how I think about a lot of those showcase stories from the market. You can’t learn anything from OpenAI or, like, BeReal. Basically a lot of companies that, explosively or diligently, still took off can’t bring you anything. Their stories are unique, and sometimes even deadly, because in your conditions it won’t repeat. <br> Garry Tan: Unless you are somehow deeply supported by a Thiel (Palantir) or PG/JL (OpenAl) via deep investment and time, I would not take any lessons from them (he meant OpenAI and etc)</p> <p><a href="https://t.me/chillhousetech/230?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sarah is one of the smartest investors. https://darkolorin.com/posts/229.html https://darkolorin.com/posts/229.html Tue, 05 Sep 2023 01:15:59 GMT <p>Sarah is one of the smartest investors. Really glad I got to know her. Her essays are always extremely useful. <br> <a href="https://www.sarahtavel.com/p/ai-startups-sell-work-not-software?utm_campaign=post" rel="noopener">https://www.sarahtavel.com/p/ai-startups-sell-work-not-software?utm_campaign=post</a></p> <p><a href="https://t.me/chillhousetech/229?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The unicorn story is like that, the main thing is there’s a line June 2021 https://darkolorin.com/posts/228.html https://darkolorin.com/posts/228.html Mon, 04 Sep 2023 18:56:38 GMT <img src="https://cdn4.telesco.pe/file/jxKakjnwY6TjUFtL5i3PQk_VIUXAHcivD0-bi3QOweYFRpm5UBtTiYvCid5L0fVKVPvgdMCUQ_ALh_Dztd26QzB7MtXhET_nHkkr5kEZAAN1Sg-BwqYueq_9-LNhS8BDwvQ2DHl7w1x6YZKe_MJ4rYLBvKY0yBNFqwiApyaJYPSucDT14Vo7xWxBNPkUsx2sfwAz48lYtyfrk3GEHSvrbAdpjGbB7vnWsk22UR6jas9mc14Xkov0gA9ptQMtleFwUAm8RC9IElg47fG8UYTZuaXQXGbFU9P7HsQqcIcLnTv_090pBnttG4jwAoFxxbOlqfGPxa38VpnCvqgJi-ybwg.jpg" alt="" loading="lazy"> <p>The unicorn story is like that, the main thing is there’s a line June 2021</p> <p><a href="https://t.me/chillhousetech/228?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/227.html https://darkolorin.com/posts/227.html Mon, 04 Sep 2023 10:58:01 GMT <img src="https://cdn4.telesco.pe/file/rrMXhY2iJufttvuuh2U05C1j5CdKxwSy15RJNCBra2XO84jgwoVL2JoPIFw7c_zrwqtpSLkgr9gs4kDqAhBDOovt3ufzvCp_zM7z2UcVxLBRfd55fcvqvRtCAgqwxN2h-p35d7gpLog09zd4D8xH9jWjTys2MIn94i9FvgRwA9m11oM6XYh7EyguM9OKrUIU_KEqmWCoPwjcDXj2DBTfqgebhA5ie9a0032XCNijqCenZlBrBhXNkwEYIgAYKGAdITAGsZr57VfeSAZheWhQ5bpM10PIVhBRAc3acTLrVS75OmTTW4Up0aZUgDfOy3VW9lljwlmh8gSkt1sOy31bnQ.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/227?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m one of those people who believe in quantum computing and quantum computers. https://darkolorin.com/posts/226.html https://darkolorin.com/posts/226.html Sat, 02 Sep 2023 15:49:24 GMT <p>I’m one of those people who believe in quantum computing and quantum computers. One of the very important algorithms is Shor’s algorithm. Which will slightly break modern cryptography and RSA. Although I think they’ll switch to quantum-resistant ciphers, but it’s gonna hurt. So there’s an ancient translation of an English article on LJ about the basics of this algorithm. There’s not much math there, but there is a little. <br> <a href="https://nabbla1.livejournal.com/21789.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://nabbla1.livejournal.com/21789.html</a> <br> Why do I think this is important. This is an absolutely different way of thinking and computing. And the future is behind it.</p> <p><a href="https://t.me/chillhousetech/226?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Let’s pour one out for the superconductors “experts” https://darkolorin.com/posts/225.html https://darkolorin.com/posts/225.html Sat, 02 Sep 2023 15:06:42 GMT <p>Let’s pour one out for the superconductors “experts” <br> <a href="https://nplus1.ru/material/2023/08/14/lk-99-resistance?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://nplus1.ru/material/2023/08/14/lk-99-resistance</a></p> <p><a href="https://t.me/chillhousetech/225?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, seriously, why don’t they make landlord reviews on rental sites? https://darkolorin.com/posts/224.html https://darkolorin.com/posts/224.html Fri, 01 Sep 2023 10:35:45 GMT <img src="https://cdn4.telesco.pe/file/nG3PAFJAncEe22kulgb5HdQiekyBpUHUIlktvtlcxuNZN2CB0hkhjSD0AAL5PWhLUhMqt0FVAeCc3lEc4a41AvISd-J6vJZwWd1v89dySOaJQzJFiXK19d1qk3w9P7nE5sGOMdD34yxs71WPdDln4hR655qjFTq3dFnXO3vBYjJIdNRZk0p4sWYQ2Jtj3ui3YUJUkMUNK80X4HHMR0-guaHZcEyMAlaZuMZVptc-freXglvo7T2W58aSgHKqoiDtlevBhfDyCXsB-c5MJd2793dXnh7Pv_YpLHaK8JSv8-c3ysvD7CZ7dLuOAEAMvJyR--IMRocT7F_DOaaCEaEnoQ.jpg" alt="" loading="lazy"> <p>By the way, seriously, why don’t they make landlord reviews on rental sites?</p> <p><a href="https://t.me/chillhousetech/224?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So true. https://darkolorin.com/posts/223.html https://darkolorin.com/posts/223.html Thu, 31 Aug 2023 14:33:05 GMT <p>So true. Really nails it about experts, experts judging experts, and all that kind of stuff. <br> <a href="https://amasad.me/meta?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://amasad.me/meta</a></p> <p><a href="https://t.me/chillhousetech/223?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/222.html https://darkolorin.com/posts/222.html Wed, 30 Aug 2023 23:42:55 GMT <img src="https://cdn4.telesco.pe/file/WL_YuwsPZEuORMVl2OCKW_sX3BbFUWgE6v424tpfOaSEnnYMXHfdk0INKXLrXejISo4SZGTwW-bqKR-FzRgz9YHWGBx9CQ-K8kPIXKJF7Y0DnIo5jjREHBuXWdgSremKNoMlTfUHpi9m-5GI9ItQjQQsUYdm8fQ7SOeUldrC1D9kJmEhz41q704REsE1WBitH7akbFV1d-QQjGmLjRF9ISiA_YeuUBvJqZpgGJm9y3rsrdtuGApYqkS6pJsfYjNpwb_PNd_wIpQ5JPy287RA83-sZ5wwrBmEkPES7Binkch4ElF_cS_CKhjEYAjPVI5ilxx8LPXQ4oBco1J1OtOy8w.jpg" alt="" loading="lazy"> <p><a href="https://journals.sagepub.com/doi/pdf/10.1177/0146167218783195" rel="noopener">https://journals.sagepub.com/doi/pdf/10.1177/0146167218783195</a></p> <p><a href="https://t.me/chillhousetech/222?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Deep learning is overrated. https://darkolorin.com/posts/221.html https://darkolorin.com/posts/221.html Wed, 30 Aug 2023 22:07:10 GMT <p>Deep learning is overrated. Mark my words.</p> <p><a href="https://t.me/chillhousetech/221?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Of the noteworthy bits, looks like Instacart’s main business as a grocery… https://darkolorin.com/posts/220.html https://darkolorin.com/posts/220.html Wed, 30 Aug 2023 13:47:14 GMT <p>Of the noteworthy bits, looks like Instacart’s main business as a grocery delivery company is gonna be ads :)</p> <p><a href="https://t.me/chillhousetech/220?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Uuuu, looks like they really need money https://darkolorin.com/posts/219.html https://darkolorin.com/posts/219.html Wed, 30 Aug 2023 12:10:49 GMT <p>Uuuu, looks like they really need money <br> <a href="https://www.theverge.com/2023/8/29/23849611/snapchat-releases-ai-selfie-feature-dreams?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/2023/8/29/23849611/snapchat-releases-ai-selfie-feature-dreams</a></p> <p><a href="https://t.me/chillhousetech/219?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So how do you like stories in Telegram, want ’em? https://darkolorin.com/posts/218.html https://darkolorin.com/posts/218.html Wed, 30 Aug 2023 11:59:29 GMT <p>So how do you like stories in Telegram, want ’em?</p> <ul> <li>Yessss</li> <li>No</li> <li>No and there’s no way to make them work</li> </ul> <p><a href="https://t.me/chillhousetech/218?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/217.html https://darkolorin.com/posts/217.html Sun, 27 Aug 2023 15:19:23 GMT <img src="https://cdn4.telesco.pe/file/HiFimjr0tKZH7kGADMAypAFe-w8JwLMaLk0fXth3JwErxUiK7vw5AwFEb2bKIwZ4AQtIkD6J6dkuFIgsyIr7nW3Vv21XNA8aeg8KGutTiI3QlCM9M4Lxv7XT9EKT6vhkJ54h-idnXxmJ0Fzk9yW6RVof8SqOwCJ9YUc7FuCZLyYWtxyDND0noCQgUEOThLOH_GZJpVJQCfpk0kmsGYAoGWRgo1tARjDWZzn1hF0BwLYzpGLQuIvrE_hj0RT4WcmpGt1_ID4rGvCtS_pdDFrRjqBVUo52nG37n7Ept9XKPedjwK9_JmfBCJ4Z0196j1lQCIggNomQEYVAqb1U15JrOg.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/217?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> OnlyFans’s annual revenue = $5.6 billion, up from $4.8 billion the year before. https://darkolorin.com/posts/216.html https://darkolorin.com/posts/216.html Sat, 26 Aug 2023 21:10:10 GMT <p>OnlyFans’s annual revenue = $5.6 billion, up from $4.8 billion the year before. The number of “creators” shot up 50% to 3.2 million, while the number of fans grew 27% to 240 million.</p> <p><a href="https://t.me/chillhousetech/216?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A playful industry bigger than the music and movie industries combined, times… https://darkolorin.com/posts/215.html https://darkolorin.com/posts/215.html Thu, 24 Aug 2023 11:03:43 GMT <img src="https://cdn4.telesco.pe/file/vMW1JYrO1wy7lDj5gCHnIGA7Asx0tzXakPjpJhwgZjI8cnXWJb4bzfnadS9ul1tfABTrh8c8i7TFGeHVA-VksTkCx7m5dsPympQjdVu_r6KiTkFLmfDijHMTtI5iQdYI1hcrDaMn30yqqrT2oIJgRtwWOFueKfBkMGls5TRWSFVXdrTx8YuGP4a7a9zv9BNe4i7JjaECkV0gPQLPr_VGuec28g3qSVBZjX10OvaUooRPyauUb5z1vw380xXZ1RU1zCNHr7xecR0TObMLbKo1t2KIFmqXrcnG2bmrfhtuIgG7tBtj-U1u8ZYgok4fM9-WULGSWDb0tDcxlN5wIZTenw.jpg" alt="" loading="lazy"> <p>A playful industry bigger than the music and movie industries combined, times two. Wonder why they rate the influence of movies and music so highly and don’t look at games at all?</p> <p><a href="https://vgsales.fandom.com/wiki/List_of_highest-grossing_video_games?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://vgsales.fandom.com/wiki/List_of_highest-grossing_video_games</a></p> <p><a href="https://t.me/chillhousetech/215?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I once posted a deck for a consumer draft along the lines of my thought… https://darkolorin.com/posts/214.html https://darkolorin.com/posts/214.html Sat, 19 Aug 2023 15:44:21 GMT <p>I once posted a deck for a consumer draft along the lines of my thought process, here you go</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/214?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I kinda think the same https://darkolorin.com/posts/213.html https://darkolorin.com/posts/213.html Fri, 18 Aug 2023 21:29:01 GMT <p>I kinda think the same <br> <a href="https://x.com/olivercameron/status/1692320417841856706?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/olivercameron/status/1692320417841856706?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/213?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I recommend this bank to everyone They’ve now automated SAFEs https://darkolorin.com/posts/212.html https://darkolorin.com/posts/212.html Thu, 17 Aug 2023 15:06:06 GMT <p>I recommend this bank to everyone <br> <a href="https://x.com/mercury/status/1691877085839524332?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/mercury/status/1691877085839524332?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a> <br> They’ve now automated SAFEs</p> <p><a href="https://t.me/chillhousetech/212?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/211.html https://darkolorin.com/posts/211.html Wed, 16 Aug 2023 18:03:01 GMT <img src="https://cdn4.telesco.pe/file/kGj8OCyAF9avRWcrlAF4zPLC05zojOccJiMnQpow3XtsRUtVOOxMggTjG4qMwzWukeiQlvH6Kv0G7W8u2hqyPE3gYBtron9kWTJMzfAFEgIcFU0GWMA7YRqJByoOzghIBwvQomvO_uUipQz9hGGFvpqIdIOwnL-NjQID0eF87t6e9drTNzcb2CMivuUXvUvhXW4FcsSuwaBVL6KamVp_38AQPLNiPrTelUnx5Dl0FHIGyXBNIOxZrMi56iLMjAx-869t1Qc4YTc66c6AANxq_6ozZOtiS4JpTHQLyCCk1y9mhK6mQ1DtbdMJ2PXeCbGIw-CTmxdCkd8VHkr4lCUYTw.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/211?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/210.html https://darkolorin.com/posts/210.html Tue, 15 Aug 2023 14:41:26 GMT <img src="https://cdn4.telesco.pe/file/PQ6MOV4mt4Muixa_D-7Yle1tlYYQ3Ore1zjp8yyVKM3BWjBFvlM2HNVwgqGKc3tVflParfNLHAwf98LJH6qQ-YR7YUfav4Bw5WpmIcwld_njYaFHlQ7lZ2KrmeyaS_f0kBoHUp8OUDdSKVNYEFkFGXQzriTxs9tqOHORiGqxqRkijRtmA_dRexZLlk0XTkHxyxeu2hbY5kKRUYgiJI5qqTXuF7ls9Ago6wb2vFrc73unUP0kLUIe6r9Se0VS7RlMok8uTr-TESzobl2CitFHGDddhCjZVEwASRVe2HKQny62GNy9i_agOd5aAOR_aEGm2Jud1K4kzOJ4cyUDnqCg0w.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/210?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Welcome https://darkolorin.com/posts/209.html https://darkolorin.com/posts/209.html Mon, 14 Aug 2023 19:35:46 GMT <img src="https://cdn4.telesco.pe/file/Xewde4SoW6ZdDNJ2ttacssxzrM-SczCc9OCg0kArZJMwnyA2eojI3e5V0TUd5VDvpWygDkLZqDXeXop5zv1VM2jY1LR3ZbWx93QG2fFVNXUzj56ED2XYbjCdSWHHu8qA115Dyv2b8Fb1yusp3nonwtltMI33SWxZsljs6wvtgTprtUFvLF2mE_xoVb7911OaQsY9-Tuyxno2HyeQBHSYotWm-tF-k_WJP0_NHP2g-PVnfKQJX8s6MsMl2sz0xp66_OY9RzLdmOeGOeUWklp44_zgfZNP2DkH5Re7CaruPNRJZhPX3Op1RiqEAxQSFR1QHOBXd3bie26p1qA4QYZp3Q.jpg" alt="" loading="lazy"> <p>Welcome</p> <p><a href="https://t.me/chillhousetech/209?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If you have your own channel, is it hard for you to pull in new subscribers? https://darkolorin.com/posts/208.html https://darkolorin.com/posts/208.html Mon, 14 Aug 2023 17:33:49 GMT <p>If you have your own channel, is it hard for you to pull in new subscribers? Need discovery?</p> <ul> <li>Yeah yeah yeah, it’s been time for a while now</li> <li>Yeah, but feels like discovery isn’t really needed</li> <li>Nooope, everything’s fine as is</li> </ul> <p><a href="https://t.me/chillhousetech/208?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/207.html https://darkolorin.com/posts/207.html Mon, 14 Aug 2023 16:18:08 GMT <img src="https://cdn4.telesco.pe/file/v7yNASCCTzYkFb03uN6FNbCHi1Qi5KRbxY_Kv6LU0okGrNP6oBAe5-23PRQj-huHonHFr6DgwB28w8oL89-bkZhWKX9aFZN--A0RUDCAIgeFWHNaWxS4RSknDMbIkjImRHdPnP1bAdI81LlHxN7-nGRYsb5PASmXzglebnBCSNyZmyn_W-XWi-qef6ph51kV9MgEKccwE-hv9R86Ho62fZ_SGWxRIZpH_1vaueAbkmGaBwMzfUHqOEf3VHf0_zGLzg7sfPNxKeGKV8BTHFzYwShQ1yCpSbObTeLisl96WgeUxFhqNMXRsositPZ6EfEYxgPVSbwMCpkkHa6-6MIiGw.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/207?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Observations (off-topic). https://darkolorin.com/posts/206.html https://darkolorin.com/posts/206.html Mon, 14 Aug 2023 13:09:25 GMT <p>Observations (off-topic). <br> My YouTube clip, which I just uploaded there in like 3–4 minutes, basically pulled 160k views in 9 months. Just a transition from that Amazon Tolkien series. So for the whole history of the channel I’ve finally got 0.5M views. What for, why, eh, just because. <br> Gotta throw up some more edits, maybe I’ll even get to monetization. Though they need subscribers there, not views. Watch hours are easy to rack up.</p> <p><a href="https://t.me/chillhousetech/206?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The digest of fun interfaces is with you again. https://darkolorin.com/posts/199.html https://darkolorin.com/posts/199.html Mon, 14 Aug 2023 00:15:58 GMT <img src="https://cdn4.telesco.pe/file/JsLN1E1NXGkZfPDWAB_C-KtAP1kyMHZ8sb4kxLFweIYIYYybKIIDt4sUVsVnSvj8HAwkwPqMJHlM1YUY2RTePtWjK0ZJZJAPmyN9KNCK8bAZUff3VcpgVfiYiifCUbtV6aGMRS6Y_pzETLzwWBPqTI0UBe_2xNR54xFL8-rRRz6qzaSi1m6rhnUOXonqs9dhZPt_3fCKrI7EBH-kN-ep7DBcKxnKPhwlAf9aAVgamvX2eGVYLbK3own2H6ToO4L7LwvPjvmZvga3-8JMSZbXBKKeMX6qI9fJW4RyvXath5RNJpOyuWQngrMRcc77WURrEXgZibx-dA3Z0mKjMv1Rgw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/Bzvlt_3qDUeknz6JHZxnaFlQoWpOWPV_qIPAwz9sQqharYccN2GPeZLg8X4BLnpmMPX9cc-ro9k0ph1856m2kyHUNBjQYKwmd9Qu35OzyalQt7tZu-hpwSFTHK3EnVJceXI3ycnm4WoDClS_b45kIbdJSVU28ge0HkzlCSSH8qqYwxfz1BSobHPxzOgG_1yhPobv-6WGJUs6ZibDL-ghQ8KQtBHHmNAm1s7iRSKCH8-pOFCZpbFnX5RXt0_IV2Lzvrva6dDswG-HTfFzbAzaVOh3mzsMEkSVxIZTwRpojnG32QZ3pVXciFFEcAVt2jlAKUUhBVY7hlZm2Kiq5oLWyA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/pFpz5RiSEC15sX7HGlRjV5jR6XXiGmRHEhxzApuofcSGOLC2NseBSNhSIbpNlinTm2u7TJ2gDShbeXpfBHG5PYDyinpkWxQphEAIRv58qZdYaNWfwWDKr9FZg8rM2a-YZOH-8oacTPWpKWIO1Sq4dWVKOM_Ks7TgE3TPm5HRMA1TFS7x50WuO-jADrrvrjaxsynq3M2Kgi09tXJTrNZUppgzDCoEZcoaPjlbRVtoj2kuXvLzqrI0L8qsKl6verEAyyRz2VyJBEgu5T-ouYaD7qgNFu063_VKKoVqDGjmeuYI4hxnwi5w12IcT2gHNr53pdj-fBfSZa5eDYkegr1vrw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/gc9g6e2dL7-Pa6f82nvNHdpMuKXsd6uOB-S6oIEmZn6avGwVtIzaCN2PbEJrlq-3UAOL4zTqr4PrYC3iI1YnoHBX7IAjiifeWmj2oG3lxZ9EMhk9q-wIh5tvzJisOop-z0zMqsJXEQp78tNq02boAnhuSxg0jnPy-5NZEd1z2tUH17u-wlcgbH_tVrTthkBFhwKT2Tasp-TPSkASiA_U3hwVDFPAsDPBws8YqghnBibWTNqt48X8eqh59koDXrngdbjz3DGmWbgmkX-sju7IhqEtGby1o_-6bcjngButwaT6G2z9MqfzGP3PT3LO5GMKXs83D04mSUpL6cim8Uo_xQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/lW8EAC195Yk8iD3mw3DcK8QbXl_kwnIh7bqLM89u-Oa01oISGsI0M9v_t4usI9ZfyjTmu_SIEWxuD26TStJO05bTfObdFNvVmP3vESuhY-yictEOi8lYV8bxROtB09foNaeDMzw7Wp3dO4lL08jv7h_fw7wCuK1HII9AMTRoTX47NjkKu1b-bY43qDuOj163r2OBMaXENrV219mgivpVJlw_dbIUWXSc_fvOFSs09ns5Lit3Vx-y-1Qmor-RlG_l3tUC77rt37z0EkFsi0XL278igCisUxLXDevJp8S6qJPs2BkHSRd9L3k4RViWHL5DIGUKiFbr6PVpk8RhDAHqaw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/Vnx2myH4ke0kfxksI5AmWc1XLx8i77lKsZQWoFxjtgmo_dCWRazfV_512nxQt8dRNGZ9LcjFWWWtMIFLETocD2gqy24dT6LLbvubsq38gMrnG6Z9mBpsJq4vCTp-IK-FXjdgWzucGc2gegCFBFUzsLikM1ZpOhU93mvVA87BhHnr5xT2zKzviR2rgFv3HDmij2Eq6A7jKTQXKpW4Er_vkFwTRXLu16tfWbo5RQevy_4EXXl48ApeoNUPDgN_nOP1mWbNktPLE-2vrmRpMMV6Amt5XuIFWQRViJORFsGm11AqhGOEjEqDoZK1tRPhh5Uj3W7afWei673m5Ab5qMGXOg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/XjLErtQfS6a0tgLrMZ3EEAngFVCDt2vR5QM1dv0BD7lrcayR3zUDsMZWcJtcnHjmK6ySiMJy-cn8uFtNKtsYTewMJMYlUsw8u8-guCOxZeZ1IZ7uOu0qmmvAGQUnIEygcnWo6auBfbIR0wBwCA8hFguAPVRWD9r1x8Ftwy9dd8iNoPVpi7Q7ZjVYGnjdoI6THXRfp9NogaiNFQNOqChCjByISGjBHJP6FQ5UXblbvjLC9LyzALZ30_MtztwgzwSKZeu5jeNE-rS9PTmssEO062qUOqe9aWAWsaoT5bAJrhxxRMJ9x5azlwoQcUE6Njwu5VgyixSJrZ39HJ1rUTNO6A.jpg" alt="" loading="lazy"> <p>The digest of fun interfaces is with you again.</p> <p><a href="https://t.me/chillhousetech/199?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Maybe Elon decided to buy it? https://darkolorin.com/posts/197.html https://darkolorin.com/posts/197.html Mon, 14 Aug 2023 00:11:50 GMT <img src="https://cdn4.telesco.pe/file/JKTJM4AtxJ_AeusTE0zpV-OTuDNPpjWAZACvX5xY6UlS2CKb2PHrhRRmidrT0aJlNwSWHQpQ1m_Ko2OtyGJhuoX8pWI0g2CvME73zLVnLf-nL-_nvquoPxYEkUFjlgHXWcGsRIFSkZ53ZCkqulC-KQWPiaRLTVRY4oayaFMTfF--Tx8UpRRDJc6OXCKvmNGFz4FjBVcSdmNmGq0_Ouo-SFAB1IaEGkk4LTmqEUau_SM4ilL3yj-uSpb7bYMOju0pb5NQpM8KM9_XpSx03nYex7a88s9OhsbTY6DRm3WpkUXDiDYq5wWiJKNo5ANcgaDCyXCgo3222xDpBdjaDPe_qg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/UHzs8vupWyjln2MWKGdRifrmPTt0aRkRqyPN6dX4grsfYZO-OZ6SVqpfyTHnZ3jbVmCOd6AXK5Wj7PP5Pu3-C40UCBjuv3OFTtuRphMi9gxh9zZBUerXGTd0cvzWACflJ2GY2S99XOq1cPIkQl1J6KpfLPqslMzASauyK-Ioo3h3AI3mR42_uHJDto6KCK5QgkNMBbmOCEFTTwCu4OFD_7VxW98ToSPJqm22h_nxoxyjSRk6QRbI8qZkeNvHI-tBVeYVe1EYw7OM4AdrFpx23UW0B2bdoavZmxhEdxXYbIF5YminWhrzWc8rzn_6xn6ApEenOjFB7RZHUxio5n6FrQ.jpg" alt="" loading="lazy"> <p>Maybe Elon decided to buy it? Just for the ticker, you know.</p> <p><a href="https://t.me/chillhousetech/197?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Food for thought https://darkolorin.com/posts/196.html https://darkolorin.com/posts/196.html Mon, 14 Aug 2023 00:09:58 GMT <img src="https://cdn4.telesco.pe/file/JvbF5WbnDMvDT61E0Xgiiam38gZr1tt0vYxGNpoeqb-7rjNApBGh2BdnenMxfJZQxaX1kTNyY4TvYqMW5ft0HBBwIhr65KSwOPINRNty-y-hl4myxEiRQk34K97wDlnd8-wznRY7Q0GMz9Hu5mH-gfqMj9QvE57kp0xp9I-_KaqvMmJ-HkTPCIqzlu90tvgCW1gUrl8YkODaUoPrcy68rUpr9hIgug5vQAUj_RauPyM7hcD3SApyhfv6fyh8TlFK4V2OX8OEkVGsg0zrDPS19WmdKw5VykbCOuAy431wp9F8_xwrlldhKJc6ubTApOCTUtqF86s-JDWi8daz6ualiA.jpg" alt="" loading="lazy"> <p>Food for thought</p> <p><a href="https://t.me/chillhousetech/196?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Elizabeth Holmes's daily schedule Circa 2005–2009 In case anyone’s not in the… https://darkolorin.com/posts/194.html https://darkolorin.com/posts/194.html Sun, 13 Aug 2023 17:58:07 GMT <img src="https://cdn4.telesco.pe/file/MuGx2YtEHESYjm7e_EmLruJ2dszu49bhyOBC0u7YSz4zjbBdetNa5cQqXgWmAfpq6Z5KTFV6FW3ouA594mOk-HSrWV0OEPRe0weZGTB-EkKXvTvxnIt5S7-AGkecrKJRXcyFisVsx3fIqjkRgUKo45W1FGEcaSLuCcCrkd7ucvhqgiUVShOrJW8GNTipg7oVJeIqhSyUdM3QjmfVE_QNUBXYL3zWXEIXOIJw5QqeNDY_tfxFojrazYyznwJFGv3Qq-tPYI38MLbAOaU7cliSbEXB-0lGf6JMovmIunatsQ1cxWLiYNh16XRJRmfsc3dJX7EW2MPGS09fWcFLm7Chqw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/i95VS0LvGiiEW9YkdRx-b0BbuYreArqJWuDVBEnEFXBO4rBlKtZ-IABgDY8B3AqBktAsc0OFOuhi2DYlZofj7uRX92qo2cPbkyXmOx8Z1t2JJibpe8kXqGAahPSDXfFuuqubTRCgWeDPZNcvNlb_rUlJ5heKbTBWOqccIQ8N2uy1OsI1UuRS8nW-vhHPN_E9WgLw182nubQ3Iqe_ag3g8bte4Q-fT7Ta1kfkO5ZhWUHkxDLRxe_safWsD80Rc0vYomWrFGdUzSJWSFNpU1hplQc_L7p6AxXA7u7PlXXINfXg-1rRdluSoZ5lp1A6dikFzNPrqVyXlWXwsGyy_QwiIA.jpg" alt="" loading="lazy"> <p>&quot;&quot;&quot;<br> Elizabeth Holmes&#x27;s daily schedule<br> Circa 2005–2009<br> In case anyone’s not in the loop, this is Theranos and all that.<br> &quot;&quot;&quot;</p> <p><a href="https://t.me/chillhousetech/194?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A new longread by Andrew Chen basically boils down to: - if you’re not growing… https://darkolorin.com/posts/193.html https://darkolorin.com/posts/193.html Sun, 13 Aug 2023 17:50:54 GMT <p>A new longread by Andrew Chen basically boils down to:<br> - if you’re not growing then go peek at what users are doing there<br> - think maybe how you can bump retention<br> - what if users don’t like it or there’s something cool you can hack together there</p> <p>Yet again I’m convinced all this advice isn’t relevant. In big companies where product work has been going on for years, you need specific advice, not abstract stuff. So yeah, all these product courses are nice and all, but you don’t really need to know that many of them. Better to have real experience in a specific vertical and a head on your shoulders.<br> <a href="https://andrewchen.com/growth-stalls/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://andrewchen.com/growth-stalls/</a></p> <p><a href="https://t.me/chillhousetech/193?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Verizon is shutting down its video calling service BlueJeans, just three years… https://darkolorin.com/posts/191.html https://darkolorin.com/posts/191.html Thu, 10 Aug 2023 09:46:30 GMT <p>Verizon is shutting down its video calling service BlueJeans, just three years after acquiring it for $400 million <br> Efficient.</p> <p><a href="https://t.me/chillhousetech/191?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> That would be interesting https://darkolorin.com/posts/190.html https://darkolorin.com/posts/190.html Wed, 09 Aug 2023 11:59:12 GMT <p>That would be interesting <br> <a href="https://www.hollywoodreporter.com/business/business-news/disney-apple-deal-1235559416/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.hollywoodreporter.com/business/business-news/disney-apple-deal-1235559416/</a></p> <p><a href="https://t.me/chillhousetech/190?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Cash on hand: Apple: $167 billion Berkshire: $150 billion Microsoft: $118… https://darkolorin.com/posts/189.html https://darkolorin.com/posts/189.html Wed, 09 Aug 2023 02:16:30 GMT <p>Cash on hand: <br> Apple: $167 billion <br> Berkshire: $150 billion <br> Microsoft: $118 billion <br> Amazon: $65 billion <br> Meta: $53 billion <br> Tesla: $23 billion</p> <p><a href="https://t.me/chillhousetech/189?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Based on weekend thoughts and watching the last few years of the… https://darkolorin.com/posts/188.html https://darkolorin.com/posts/188.html Mon, 07 Aug 2023 16:08:36 GMT <p>Based on weekend thoughts and watching the last few years of the investments-and-startups world. <br> I’m betting that, in today’s sense, the early-stage startup industry will collapse within 10 years. The economy just doesn’t run on the “garage revolution” commandments in tech anymore.</p> <p>The modern ideology and approach to building tech companies is “a garage and a late-70s team.” And its continuation as the 2000s software industry. Unfortunately, time’s up. The market is different and grown-up. My analogy is the auto industry. At the start of the 20th century it was easy to just build an engine in a garage and even put a car together. Not now. Funds stopped a long time ago looking at “let’s build a car in a garage” as a potential big story. Systemically.</p> <p>Now we’re entering the same cycle in tech. Making something truly big—the kind the modern venture industry is built for—becomes basically impossible (early-stage startup stats). SaaS is essentially turning into car washes that locally (on the internet) service big businesses. Consumer product has, for the most part, collapsed into content, which is, well, very expensive. And traffic, and how it keeps getting more expensive, is proof of that. The “put in a million, get a return” economics just doesn’t work anymore.</p> <p>And then the conclusion is: you need to reduce the number of investments and increase their size. So there’s a chance to build a business. Only a limited number of funds can play that game. Everyone else will be left with the market of pumping teams up to an acquihire. Yeah yeah. All m&amp;a already right now is really just buying teams and people. And big buys and x’s will be a rarity. Early-stage funds will die by the hundreds. Because they can’t deliver the required return.</p> <p>It was fun. But the industry grew up. Time to move on.</p> <p><a href="https://t.me/chillhousetech/188?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Hm https://darkolorin.com/posts/187.html https://darkolorin.com/posts/187.html Mon, 07 Aug 2023 00:10:59 GMT <img src="https://cdn4.telesco.pe/file/YxqVl-icnv53WjMpfvFaYIKJwfbz4i5-x5qA5DoIh91VE7emnzUuRprPH3Dxuebo_Qn9ZXzquCjwR9Ju63pHGXXLbCTSsSqvWKMYxVRDU-HXFMf_X4j5MXteTTaExcwW5DHOMpL-CMD85q8zSgT4J7SiVSSea_YLm3Yo51YkNpVXivuJtzyHNBEu3ijnWul-tD693U0apVVRs3cUuo783bMh8CbfB6ZjPFYLR4o_SCNcKYQjcZ1n1u3AEJ2LaHejO8rGen4z9mCHMynKrt1QNEMkxc--Mv1hq7gxv9Nzn63R5N78QNX73BLxm5ha_lXFn5yuWEuqUQzeBvs9m1E2hw.jpg" alt="" loading="lazy"> <p>Hm</p> <p><a href="https://t.me/chillhousetech/187?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Great presentation by Yann LeCun. https://darkolorin.com/posts/185.html https://darkolorin.com/posts/185.html Sun, 06 Aug 2023 22:49:41 GMT <p>Great presentation by Yann LeCun. <br> <a href="https://drive.google.com/file/d/1wzHohvoSgKGZvzOWqZybjm4M4veKR6t3/view?fbclid=IwAR0QP41738YShCNfwit3Jf7yevf0VvJhnRDErrxUADCzqRs1domSK0crv9A_aem_AWm322uYUAPo8UJJ2n-k3-NQsKpieuvDkmmB0LhVXC--5w_9EBxSjeU3lNJ4oP04idQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://drive.google.com/file/d/1wzHohvoSgKGZvzOWqZybjm4M4veKR6t3/view?fbclid=IwAR0QP41738YShCNfwit3Jf7yevf0VvJhnRDErrxUADCzqRs1domSK0crv9A_aem_AWm322uYUAPo8UJJ2n-k3-NQsKpieuvDkmmB0LhVXC--5w_9EBxSjeU3lNJ4oP04idQ</a></p> <p><a href="https://t.me/chillhousetech/185?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> or maybe it’s bullshit https://darkolorin.com/posts/184.html https://darkolorin.com/posts/184.html Sat, 05 Aug 2023 00:53:01 GMT <p><a href="https://www.nature.com/articles/d41586-023-02481-0?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.nature.com/articles/d41586-023-02481-0</a><br> or maybe it’s bullshit</p> <p><a href="https://t.me/chillhousetech/184?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meanwhile in India: India Is Moving Forward With a 28% Tax Levy on Online… https://darkolorin.com/posts/183.html https://darkolorin.com/posts/183.html Thu, 03 Aug 2023 12:27:18 GMT <p>Meanwhile in India:<br> India Is Moving Forward With a 28% Tax Levy on Online Gaming, Casinos<br> A 28% tax to be slapped on online gaming starting Oct. 1<br> The GST panel to take another look at the levy six months after it goes live</p> <p><a href="https://t.me/chillhousetech/183?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, looks like the majority voted for the post. https://darkolorin.com/posts/182.html https://darkolorin.com/posts/182.html Wed, 02 Aug 2023 22:45:46 GMT <p>Well, looks like the majority voted for the post. Let’s go. <br> Part one — theory. <br> First off, I want to note that lawsuits like this are rare at early stages of a company’s development. There are known cases where there was stuff like this <br> <a href="https://www.businessinsider.com/secret-founder-sells-ferrari-2015-4?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.businessinsider.com/secret-founder-sells-ferrari-2015-4</a> <br> and it’s fine, they sorted it out and the guy is still working and doing pretty damn well <br> <a href="https://www.linkedin.com/in/davidbyttow?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.linkedin.com/in/davidbyttow</a> <br> I’ll be happy if someone even finds a non-crypto case where they sued a founder of a company that raised like 10 million. <br> Usually this whole fiduciary duty practice is applied very rarely. But they threaten it a lot. Also founders and company directors often have indemnity. That’s when you get sued and the company pays for you. Think about that when you read news like “LegionFarm’s CEO got sued by their own investors, who want him to return money to the company.” It’ll be cringe if the directors have indemnity there. But we don’t know that. <br> A lawsuit in the US is a whole separate song. I know it not by hearsay, since my company got sued. And it was very expensive. I could’ve done it cheaper if I’d hired cheaper lawyers. But either way it’s hundreds of thousands of USD. For what? To get back hundreds of thousands of USD? Even that you still have to prove. And it won’t be that simple. It’s important to understand: if the CEO is under a lawsuit, you have to replace him. Because nobody from other companies will want to work with someone like that. They even write this into contracts. Like, guarantee there are no lawsuits. Disney once pointed that out to us.</p> <p><a href="https://t.me/chillhousetech/182?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Have you heard of LegionFarm? https://darkolorin.com/posts/181.html https://darkolorin.com/posts/181.html Wed, 02 Aug 2023 21:44:50 GMT <p>Have you heard of LegionFarm? Curious about my opinion?</p> <ul> <li>Yes </li> <li>No </li> <li>What even is this LegionFarm?</li> </ul> <p><a href="https://t.me/chillhousetech/181?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Oh yeah? https://darkolorin.com/posts/180.html https://darkolorin.com/posts/180.html Wed, 02 Aug 2023 17:33:04 GMT <img src="https://cdn4.telesco.pe/file/JyIkwZsILm4z2pf5h0Kd32BLyDgXvoIwFSHSPgG24URq9Q0Uwhlo6U3yp0ukRzP_oEG87dCnhvSoPttJpJ5ls6ZQ6IYLC513TZ-gE5_BxFYf4fBQLgIn7xFxAUGnBNbHG84JXJUEREIvS3l_SuPHaAElV3_XrUO5ZRy-C8eE1OIvJXRNmeQh13O1l7Wr3pwVAZatR0ArAlPmbJePmZ1-93Bsq2oSICB56-fF-kWfTI62CNe7_G2rZ7V7jrP1mCLr36LCqwFGqcuoGHeigkakOefjTtGkhOX7fs9sz4BZBccwzZeoGuMhTdd3lYdGCDoYt62Nbc8SyVLYBvxpg5MgEg.jpg" alt="" loading="lazy"> <p>Oh yeah?</p> <p><a href="https://t.me/chillhousetech/180?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Looks like superconductivity at room temperature and pressure is real. https://darkolorin.com/posts/179.html https://darkolorin.com/posts/179.html Tue, 01 Aug 2023 15:37:29 GMT <p>Looks like superconductivity at room temperature and pressure is real. That’ll be quite the plot twist over the next ten years.</p> <p><a href="https://t.me/chillhousetech/179?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I can’t open that Twitter link from here. https://darkolorin.com/posts/178.html https://darkolorin.com/posts/178.html Tue, 01 Aug 2023 01:47:53 GMT <img src="https://cdn4.telesco.pe/file/X1GeP82bWaZlT2OJ0vZvx_BVX3zyPZsJcuDlIqYGxQY41fbsSw11Ay29tinj3dICB36wCVqGx9jN7RGzmPyAv9ztLeplpqBYVHxdkzbOCFcc-9ZM6x4YhnZsNSosmMZK-x_CrsqmJ-K_f3sPpRWoU6t85JY9xqcxy_Id5RORUl0fodj1CDiadyzvMP1n8WBJ0cT_HZtvcje4fWbghXIVDIgcE1TD5dW76uW_dFNa9UvtXy94jF5uQ_G7NV1jlgIBNcORKQs3feMfsjzjFmoql1TOSrd6IG7hFU51x71COtgXwdv3Wnd8Hv54kuDvrVk34LbdDQcCYgUN4jA44wQEyg.jpg" alt="" loading="lazy"> <p>I can’t open that Twitter link from here. Paste the Russian text of the post (or upload a screenshot), and I’ll translate it exactly in the same tone.</p> <p>Links from the original post:</p> <ul> <li><a href="https://twitter.com/ttunguz/status/1686133256058486785?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/ttunguz/status/1686133256058486785?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></li> </ul> <p><a href="https://t.me/chillhousetech/178?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/177.html https://darkolorin.com/posts/177.html Tue, 01 Aug 2023 01:39:11 GMT <img src="https://cdn4.telesco.pe/file/vVFQ9RGLdumgEJ-Cb2aPlXshdVOplLL3aqVBtgsmZizPCwamD2ROvwDotD2B4asAwD-Tm9Y5Av2CVlhPWg1Jujy361hwlCUV3KC9wvDV6QZGaYICo0fT_FVK86msohnpx4ITCRrLPBzyBBLYeITOCz4rjdnNU0TJDzeoE6TeRKTUffCx_kRcSzzNqfTfBpi7xc4xTSERtRRnDGA1CB2Zjc1vfXe-evudPKQ4p8Wukz5bj-xT8IL4IBMXdahoSMs7hEKoQppH1vKjrZ5lX_strWlMmc6fKcuMAEHf_i7C6zwm3qcKiAheLPS2U9xpxN6IRtNSfErP4_KoKjr__iSlZg.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/177?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> YC S23 Breakdown https://darkolorin.com/posts/176.html https://darkolorin.com/posts/176.html Mon, 31 Jul 2023 20:22:54 GMT <p>YC S23 Breakdown</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/176?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Mute trap This is a problem of modern big products. https://darkolorin.com/posts/174.html https://darkolorin.com/posts/174.html Mon, 31 Jul 2023 14:04:40 GMT <p>Mute trap <br> This is a problem of modern big products. When different types of notifications start coming from one product, you start doing a cleanup. <br> At the core of the cleanup are people’s internal motivations. For work apps and chats you’re thinking how to make it so you do nothing and answer everyone as fast as possible — the Zero Inbox concept. And you turn on the silencer for half the chats, channels, and even after 6 pm. <br> In the info field, as soon as you don’t understand what exactly to read today or content starts coming out so often (lots of subscriptions) and distracting you throughout the day, you start filtering (the Bell concept on YouTube). The scariest thing happens in Discord. When your main goal is to hop on a server, play a match in a voice channel, and then they dump pushes on you from the whole hundred text channels on the server, you instantly get used to muting absolutely everything. And also people don’t like bad news :) <br> Mute trap seriously cuts ad revenue and can kill a potentially viable ad model for a product. It also turns the service into Intentional use only. Which, in general, heavily limits growth models.</p> <p><a href="https://t.me/chillhousetech/174?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The notifications. https://darkolorin.com/posts/172.html https://darkolorin.com/posts/172.html Mon, 31 Jul 2023 14:02:35 GMT <img src="https://cdn4.telesco.pe/file/DN8_8Fx_lZ4UiXnXMwSm7pKOSYs8d2R1KDRvoko-G746XvfViJU_hb166Lrv_e33rtsUsOnEKJ0s2DNDdmexs6tFC-2DFhv4mmap3HM0hACT4g1je0rNU57_hbgoK-bknWjyRV_GsC5UhaU5qEwQMeyxoKfW6HfYIisrc9b1FyO4LnELDOSEOuTl9w81GOmXXXGx-Z71g-_oEeO6dobFr183vk6hgN_Ex6yY6UAdTbswChN9dlN-gtwKnWcnvNMUwjgV9vt5aoAkLGuszC_zJ4Je5jSrax76_B9K_wZ44SwEUfTjZXsNFd5kGmY5FatBKWWKmt4GxYLz188VN3Qnag.jpg" alt="" loading="lazy"> <p>The notifications<br> .<br> This post was prompted by me pitching my community messenger Capture to Sequoia Capital partners, and specifically Mike Vernal.<br> While pitching the messenger we landed on the fact that it’s super important for a person to come back to the app as often as possible, and the only tool to control that is notifications. But what types of notifications are there, and what do you need to do to not fall into the Mute trap:<br> - personal notifications are when it’s straight up “to you”, like your friend Alexey wrote you<br> - service ones, like it’s time to service your car or refuel your Tesla,<br> - and finally informational ones — news, weather, YouTube videos, and so on<br> The strongest and highest retention is with the first type of notifications and then down the slope. Usually everyone wants to build a product that operates the first type, and it’s based on the social graph, because it’s people for people. And often 1–1. The bigger the 1 to N multiplier, the higher the chance that the notification will flow into the informational type, and that’s where the Mute game starts.<br> Pretty quickly everyone acknowledged the power of mobile push, and then on other platforms too. And they even started shoving ads in there — I used to do that myself once, in my first app<br> <a href="https://habr.com/ru/articles/122738/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://habr.com/ru/articles/122738/</a></p> <p><a href="https://t.me/chillhousetech/172?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I made this channel for myself. https://darkolorin.com/posts/171.html https://darkolorin.com/posts/171.html Sun, 30 Jul 2023 21:04:41 GMT <p>I made this channel for myself. To try to share what seems actually important and interesting to me. I’ve got a draft of a book about consumer. I’ll post bits little by little. <br> It’s really important to me that the people reading are interested, that there’s new info or a point of view. Write if you want to hear my opinion on some question.</p> <p><a href="https://t.me/chillhousetech/171?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well overall, it’s really hard to make sense of it https://darkolorin.com/posts/170.html https://darkolorin.com/posts/170.html Sat, 29 Jul 2023 17:13:29 GMT <img src="https://cdn4.telesco.pe/file/UJgiXcMc62BOZNT1z2AH_5OP4jWFZrW5pZnL1y-7WkkKYS7nNX25QHW0FTQXtOXchFM-PylGwEu5kmkW2eE4Xxd57A0S1nSSEKxCNcGxdsBYu8hvnaT9KqSjWqnSuLIa_j0RXGAs4j4Aeu3KNxWyaJEDWPUpfd8SX14AGhxub82I6-QH_7NdGzgma4qQjlXXEXMzxRAVNrDsOqWrAjT0KzGtkksv5_DCwGPIVgoF0v9UAa9dOdsxTLFQEFmTr47hB875BY2fSmMWhdSlE_LqSUy5Cgw0FjpHjiSe_cbq7oxvU-GZsJTZKb6vUC-dTwEv_kCAlprAVJLO0z2Nkt32Rw.jpg" alt="" loading="lazy"> <p>Well overall, it’s really hard to make sense of it <br> <a href="https://www.ft.com/content/0a5a4404-7c8e-11e7-ab01-a13271d1ee9c?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.ft.com/content/0a5a4404-7c8e-11e7-ab01-a13271d1ee9c</a></p> <p><a href="https://t.me/chillhousetech/170?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Yeah yeah Peter Thiel paid extra to people who live near the office. https://darkolorin.com/posts/169.html https://darkolorin.com/posts/169.html Fri, 28 Jul 2023 20:47:16 GMT <img src="https://cdn4.telesco.pe/file/chEzy187Eg-1cK2PCo75ERSCm5lPJaqCxzyaR-s-JVLpKi48Y24nXC6l6jZI8jJXaJPiHaf9Ey0N2_QlqwT7l8DpbFThFdZpBYo6RQBpudWYbUH0zKmcKFd4oHApYWeDACVOKdWPhZ0cJ2LA50ijdLVSKSXC7GFnFXQGYmA9w1h9MDLhYSAhp07trmEVXe421wAMZNwJHkJyONfxGX4uCtsev15PsL7PV-SIPta2yal1ZsuUcv3MRcBEzFqvi8Tfa6zoCv8ucoHpszMmB9eMXll-WBvSZIIyq_E-UoJOFeIsGGR7UWZ6bTlbVmQd4wkM8acajD3Yi6IaSCGaimaVXA.jpg" alt="" loading="lazy"> <p>Yeah yeah Peter Thiel paid extra to people who live near the office. <br> <a href="https://www.amazon.com/Paper-Belt-Fire-Fight-Progress/dp/164177245X/ref=asc_df_164177245X/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.amazon.com/Paper-Belt-Fire-Fight-Progress/dp/164177245X/ref=asc_df_164177245X/</a></p> <p><a href="https://t.me/chillhousetech/169?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Meta announced that Reels (across their apps) is now on a $10B revenue… https://darkolorin.com/posts/168.html https://darkolorin.com/posts/168.html Fri, 28 Jul 2023 13:45:13 GMT <p>Meta announced that Reels (across their apps) is now on a $10B revenue run-rate, up 3.3x since last year. <br> For comparison, TikTok made ~$9.5B last year. <br> Btw TikTok is doing 3x yoy growth. And for reels we don’t know.</p> <p><a href="https://t.me/chillhousetech/168?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Friday stuff https://darkolorin.com/posts/167.html https://darkolorin.com/posts/167.html Fri, 28 Jul 2023 12:30:54 GMT <img src="https://cdn4.telesco.pe/file/c0-xTbPzk2WVDIqsteMSwjzYum-j1ODZJJkKGpwvbu0VeNMfs9xO7WaJ_v-katM-UxUU3MKhdoGmvvwkT1oKrEl2-k5J260PtG6BJlOiun86XQuGa_l4QhnJW2iwyeUXXmxD5z5x2WGaMZlM2u6Hles0lSwK_zXDRQBkzgWw8zKWjVzdO3LBiRrLGYonrUkngfxSDtvb1R-s-pvxpKarYg0HeQxfkJ6GqDNbu6w9pQHZ_NYYaecrqfx29_RVC5FuCQXJ4Dc4FqRwlMFispVL2P0zLRMAyLuLcKKEEV_z-sckt3k3v_mTufMimY-u2Tam9xsbuUTnO-D6b_0xdQKXpg.jpg" alt="" loading="lazy"> <p>Friday stuff</p> <p><a href="https://t.me/chillhousetech/167?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> From a recent Harvard CS grad working “full-time” at Google: “I deliberately… https://darkolorin.com/posts/166.html https://darkolorin.com/posts/166.html Thu, 27 Jul 2023 14:30:16 GMT <p>From a recent Harvard CS grad working “full-time” at Google:<br> “I deliberately chose to work at Google because the work-life balance is really good.<br> I spend two hours a day working at Google and the rest working on my startup.”</p> <p><a href="https://t.me/chillhousetech/166?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, about censorship and control. https://darkolorin.com/posts/165.html https://darkolorin.com/posts/165.html Thu, 27 Jul 2023 14:20:35 GMT <p>By the way, about censorship and control. <br> <a href="https://twitter.com/nearcyan/status/1684242509847822336?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/nearcyan/status/1684242509847822336?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a> <br> And here’s another little link with the full description <br> <a href="https://github.com/RupertBenWiser/Web-Environment-Integrity/blob/main/explainer.md?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site#participate" rel="noopener">https://github.com/RupertBenWiser/Web-Environment-Integrity/blob/main/explainer.md#participate</a> <br> And, for example, the use case the guys are solving — Users playing a game on a website want to know whether other players are using software that enforces the game&#x27;s rules. <br> I think that within ten years the global internet as an entity will be a wildly unstable construction.</p> <p><a href="https://t.me/chillhousetech/165?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> This VR of yours is some wild shit, of course https://darkolorin.com/posts/164.html https://darkolorin.com/posts/164.html Thu, 27 Jul 2023 13:34:07 GMT <img src="https://cdn4.telesco.pe/file/ryv6IyTw4XDF92Hq7I87qaXW1jfbfvorhiwJXuRHILRn1kbCEH_rHgge_onsBvAjhwyhscTG-pYugpKLv2TOgfdluIpbmbkuKQRSYJeAE6j2MhzoVDNnMfs8t_OCJf1qjmVFD0WLvnCGXMgpKShMftsDiOKGuR3j3l5aKqZs543Gp7zPl4Tru0nQrxgEyovNZWVF3s1Bf310pXjHLhXUv-uMVxAnUQg0q9u9AS4GvNVWBNLVJh2bu4Nej3lGsGGiVOeBZKBIHTKXgTpk03BccXuGoJ9AJ6tK2IpJYbq_ZI9O4J-bwBTvkkqO0QNQCAIn3NvaatCMA11AKM7pHUfyvg.jpg" alt="" loading="lazy"> <p>This VR of yours is some wild shit, of course</p> <p><a href="https://t.me/chillhousetech/164?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Document https://darkolorin.com/posts/163.html https://darkolorin.com/posts/163.html Wed, 26 Jul 2023 17:45:24 GMT <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/163?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Post #162 https://darkolorin.com/posts/162.html https://darkolorin.com/posts/162.html Wed, 26 Jul 2023 17:44:47 GMT <p>Links from the original post:</p> <ul> <li><a href="https://t.me/chillhousetech?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech</a></li> </ul> <p>A message was pinned on Telegram.</p> <p><a href="https://t.me/chillhousetech/162?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> About seed investing by Sam Lessin. https://darkolorin.com/posts/161.html https://darkolorin.com/posts/161.html Wed, 26 Jul 2023 17:44:39 GMT <img src="https://cdn4.telesco.pe/file/SAYTGWA9_j9MQUSqwR2SZJH-WHOIoWN0BW8DD--ke1vpsdg3WILAV_39Kqetd3FfzAcR4VhaMOFHrCgydzt4jr-ZyZ9N43Q2jP0zxIjVw1o3daRmzh0xWvk2-BDIwSh5CQBA1KAIJmkjMRUV9-qtVlhnGtoET0_rsv5g9BSFYWaMxONcdw-HupB1JPr2DiSe18W-8toExNcdZ-aUpMgm2jN5zvPUm-hy7mZOVoZDQ9nPdKOH6f9DP-kbVtUkYYzH9ev7cXqBYtc7YVr6obxNf1rFOqaHhvRZtYr3V3LOlHs-_qfMiuY9n0Aue11LkQ2kSq1ELYFfFVdNQds85TI3Pg.jpg" alt="" loading="lazy"> <p>About seed investing by Sam Lessin. <br> There’ll be a tilt toward more transparent valuation models instead of pink castles.</p> <p><a href="https://t.me/chillhousetech/161?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And as an add-on, try to extend the principles above to this picture of… https://darkolorin.com/posts/160.html https://darkolorin.com/posts/160.html Wed, 26 Jul 2023 13:05:26 GMT <img src="https://cdn4.telesco.pe/file/MP-T--e9b72CwKUXLjgqz2bwyQZpiRsRdncULGBz3d_lhwToo7-QLb1bUgpSK5EPGtJglLrfZB6RNI1vqsfXXSr23l4aKeYUZVglQondcaEofXyPPLsJwV2S_-DdqyPe5enKECo96N0TvdvT02f-IcgF1GKKYhfiFyATcL-XbjyKCrSSfjfKaD84Py1ksOM5aucq6RHIvvmlg3yYGo5WI8A21EKWpVgxk1a_l_96yTqF0nSjJ9EKDE0QrSp0QyZPycLbzLJqZU2ChLAXrNpKcvrnPLqobvp2uZ1hfyoOhhbicQUA_xmKhg9v-4G5_kJlVJes3enEN0XEONB-KxZJeA.jpg" alt="" loading="lazy"> <p>And as an add-on, try to extend the principles above to this picture of feature and product copying in big consumer.</p> <p><a href="https://t.me/chillhousetech/160?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Something to read up on strategy and take a look at what Epic Games are doing https://darkolorin.com/posts/159.html https://darkolorin.com/posts/159.html Wed, 26 Jul 2023 13:00:19 GMT <p>Something to read up on strategy and take a look at what Epic Games are doing <br> <a href="https://matt-rickard.com/embrace-extend-extinguish?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://matt-rickard.com/embrace-extend-extinguish</a></p> <p><a href="https://t.me/chillhousetech/159?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The man who backed Christopher Nolan https://darkolorin.com/posts/158.html https://darkolorin.com/posts/158.html Wed, 26 Jul 2023 12:42:26 GMT <p>The man who backed Christopher Nolan <br> <a href="https://twitter.com/michaelsikand/status/1683923270595141639?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/michaelsikand/status/1683923270595141639?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/158?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Hm. https://darkolorin.com/posts/157.html https://darkolorin.com/posts/157.html Wed, 26 Jul 2023 11:05:47 GMT <p>Hm. Stumbled upon this <br> <a href="https://news.ycombinator.com/item?id=36864624&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://news.ycombinator.com/item?id=36864624</a> <br> And also a bit here <br> <a href="https://www.reddit.com/r/worldnews/comments/159g2k4/roomtemperature_superconductor_discovered/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reddit.com/r/worldnews/comments/159g2k4/roomtemperature_superconductor_discovered/</a> <br> Either way, worth a read. Either our world will actually start changing fast because energy efficiency will jump by multiples, or it’s the same shit work again.</p> <p><a href="https://t.me/chillhousetech/157?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The long-awaited post about Threads. https://darkolorin.com/posts/156.html https://darkolorin.com/posts/156.html Tue, 25 Jul 2023 22:10:42 GMT <p>The long-awaited post about Threads. Or maybe not. <br> I watched the launch of a new stand-alone product from Facebook with interest (to hell with this Meta, not everyone’s “whatever” about it anymore). This time you could actually see Zuck was involved.</p> <p>I’ll list what I think are the weak spots of this launch in exactly the format it happened in (a separate app):<br> - a reused graph, the same people as on Instagram, except the content is kinda text. That’s strength in distribution because the integration was very strong, but it’s a weakness as a Twitter competitor, which is basically built on an interest graph.<br> - in general, trying to roll into another product on the FB graph feels like a mistake to me. Right now the trend is what’s called the interest graph. Telegram channels, Twitter, TikTok, basically Discords, and Reddit’s rise.<br> - growth that’s too fast is always bad. When the audience pours in really fast, it’s really hard to create a mass of content you can actually consume. Meaning the inflow of people outpaces the speed of content creation.<br> - all of this was done in a rush, watching what Musk was doing (I’ll write about him separately, there’s this product respect for him, but that’s a separate topic)</p> <p>And just like it took off, it fell. And the attempt to build another pile of inventory for their ad machine will most likely end up as an internal feature in Insta.</p> <p><a href="https://t.me/chillhousetech/156?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> 🤔 https://darkolorin.com/posts/154.html https://darkolorin.com/posts/154.html Mon, 24 Jul 2023 23:53:40 GMT <img src="https://cdn4.telesco.pe/file/IT_FkrNNPf83NbbpbV8Ig1FF3L6En9JHdAx0ftVA0wrqa73zqcVW4mGLy_no_sCALGOmhPoS6dwhXeOEGmhOqq5zBGpZF25YnSiPNrqfUSXier9Y0q08qc4yDYAZsbj0dD4OCLKYSWLwSGwuIPCPIhyrlpLX3T_ieGRxdI4rPTFSiGFXQnKJ3vPuQUKdHy5cJqLZnUlOdlcA4k48DoqCyN9sr3GXHvYWmsAV8Z9rt_Iq_33rdj2o5Ik4os1pe7-kbLDSBslgC96dlQYN9_ItT5mvBI4KT3Kjtot5sCid9Q-1-Pz7bhdu_4_9r-S6uBKEfmvl-3Nilxf6axujTslHzA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/mQltxPh4R1aNpV5ItfXWycXgfXi2sdienEh1PGjdAJl3-0fC8dhXTI7pjNBcw2GJ-g6wcrBxCvQ4mmCdhMEm8ms0lb63lf18ZztBIslOK9MjpVVqLiUMPGJxtochf_pENqvOMlIVTHQTQtis4Hwsmmrk2a9_MmMRpWQSe1RmldQnhWvtZKa6kbyvVjnGwa9Ud4N95dVI9hRx7IJtHkTMveruXRca9n1KQtmxILsN7zlb4HTRbc6nPOlvgfStyj0qdtibyiEzIR2BjZyIN33fgzCy8gUoDQeInH544gHWbwcHWrcVUsCdy8CDoDy01Qi7vVX40c_gtiz3SZ914U85Hg.jpg" alt="" loading="lazy"> <p>🤔</p> <p><a href="https://t.me/chillhousetech/154?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> That’s pretty often how founders make their money. https://darkolorin.com/posts/153.html https://darkolorin.com/posts/153.html Fri, 21 Jul 2023 16:43:30 GMT <img src="https://cdn4.telesco.pe/file/H3BlEh62zAAm0HwWKweN8WApA2NzPQIG3zyrz55L2V2a1aTHuEJsHpe4Agf6ma9yDzs8cQ9vVPZSvb6cYhQDZ25eYIm8WZZGQNAKOHAFVsRnSNEkWWE2jXqTbG1C53o2Rrhq4mz4wjlM9Y54o9QJJWSSkGB9DxACJEsVrsaBPoa0bztjwP2w6UalffSlyKnQlb7JBx_4g0pu3iHJD6fsrtXGituAjv_RBCnmIP7rYsTrjnoMPUbB-tW1pdzXnG3JBKLR-D9-D_cc9Q5pjLXra3lp4Z0dZ7GYajohhey6_jzsurG-q6JG--BO6dDosBIsFnFmTN-WuBTcHXFgM94C-g.jpg" alt="" loading="lazy"> <p>That’s pretty often how founders make their money.</p> <p><a href="https://t.me/chillhousetech/153?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/152.html https://darkolorin.com/posts/152.html Fri, 21 Jul 2023 16:34:10 GMT <img src="https://cdn4.telesco.pe/file/Iw5e3feyyxngosNQ9ZkAlKWQRCmNy3wdHVa_faLvy5RjmiU4BX88ecpuWnvMuefub6GoDRtv43G-haqtVvm4PLuqNEwe7fxT-imEdIPKmpgqwTvJywv1Y3pmEKZVLpCqrm2Pg3UpPvosCk23Y7GBX1hpIVzX2VAxBCODTF9ntRdSW-PWU4ktZFUJ3hHmaFneitzeeaHT71NIaVTTrnYFMqVd_HwEMHc3frnlOCe5QSsNeFi1HvjYQ_HWKgfTP5gSXvuiu2Y2vwG5jraVwrbXunE040vErV_WT4suvCL-3_BMdW9e0OhnUUVaOpYeHm48DtOOQxznA6jUwlZaOlY3ug.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/152?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Hard life of an investor in Paris https://darkolorin.com/posts/151.html https://darkolorin.com/posts/151.html Thu, 20 Jul 2023 14:15:11 GMT <img src="https://cdn4.telesco.pe/file/oUD6vrfyZbeacCuq9bL5XnwsNgp9N94CsHr7bUnWOvwZma_jum4nKi2mWli2eXQEn3k0XeOzmUp4EsRaQ9aWmaKmpoIPqiTUHJOBTJFrx7MsIuRqEcA2z6a4BXrO58frJsQDQlO-ftrBCkYHb5BAr4PKLBiKY_NyQ0wLPrX20_aWxUaru7TfOE6G1gRxAcZnIjtHu9fm_T2db4bVBrLOKm-BqCGuuwyh8P3mGDNqyOUfFfmm_MCXM0DuL74fpdmGNIrvNy34QrQSCxgKoK206xjMouL3gTy9k85IAun8xshbbI7tt1cyrLerChG4z4FMkPx6tNA-N6IRo0hZDOa5Ow.jpg" alt="" loading="lazy"> <p>Hard life of an investor in Paris</p> <p><a href="https://t.me/chillhousetech/151?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Let me remind you about the hypes https://darkolorin.com/posts/150.html https://darkolorin.com/posts/150.html Thu, 20 Jul 2023 10:07:18 GMT <img src="https://cdn4.telesco.pe/file/rI64zChQhQkF4cfAI7u9nQg637tO-w-jH1tOHepZusFP3t515YJ_6jGdSzF9wNDAQbg-b8yPYkqx_O3Rehk47bYPyA_mv23NV5tak4LGJ4f-9bLOPr7wO39kvIZxj7h8fPfdOGGxt387Ya2eRtycWMq_wH4d5kyFlby_s4qCNIAgLM0YpvkoggslYopbC_2bjrNPoDlVsE5q01h4wQQl64q8Ule5L2XqW93XyaByiSjlM2R3zz8o3gP6t_F8aHoqm6INS1RmDz21HQtmNqQ7Ig55-a_FDMWK0Eak1e_XLH6JX6ZCvXAX4D6bQVrg8Jlkn9vnuKLcdBY-FLvY9cQahw.jpg" alt="" loading="lazy"> <p>Let me remind you about the hypes</p> <p><a href="https://t.me/chillhousetech/150?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So here’s this lovely link for you to read https://darkolorin.com/posts/149.html https://darkolorin.com/posts/149.html Wed, 19 Jul 2023 21:48:30 GMT <p>So here’s this lovely link for you to read <br> <a href="https://book.stevejobsarchive.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://book.stevejobsarchive.com/</a></p> <p><a href="https://t.me/chillhousetech/149?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I was thinking here and it turns out the channel has a whole 2^7 subscribers… https://darkolorin.com/posts/148.html https://darkolorin.com/posts/148.html Wed, 19 Jul 2023 21:48:02 GMT <p>I was thinking here and it turns out the channel has a whole 2^7 subscribers, both guys and girls</p> <p><a href="https://t.me/chillhousetech/148?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> and soon we’ll be putting up posters in the metro to pull in traffic, might as… https://darkolorin.com/posts/147.html https://darkolorin.com/posts/147.html Mon, 17 Jul 2023 11:50:12 GMT <p><a href="https://www.politico.eu/article/facebook-instagram-norway-ban-track-users-ads/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.politico.eu/article/facebook-instagram-norway-ban-track-users-ads/</a><br> and soon we’ll be putting up posters in the metro to pull in traffic, might as well at this point</p> <p><a href="https://t.me/chillhousetech/147?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> To the product collection And overall Sarah is really Awesome, one of the best… https://darkolorin.com/posts/146.html https://darkolorin.com/posts/146.html Fri, 14 Jul 2023 16:08:10 GMT <p>To the product collection <br> <a href="https://www.sarahtavel.com/p/dont-fall-into-the-trap-of-taking?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.sarahtavel.com/p/dont-fall-into-the-trap-of-taking</a> <br> And overall Sarah is really Awesome, one of the best VCs in the world</p> <p><a href="https://t.me/chillhousetech/146?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Negativity in the media keeps growing every year https://darkolorin.com/posts/145.html https://darkolorin.com/posts/145.html Fri, 14 Jul 2023 14:14:58 GMT <img src="https://cdn4.telesco.pe/file/lWM8ANz8qrziYFEwjqddzwD30jYxqdV0CBMXrwd6RkmDijhI_4tc2p_t8G0ILSWM4njhNUOxbPnH1AErDQJ22wHX0-hngRrOy-D62tKjsj8U8iOu6AxgkOhmWQTKldtzygXdQaGBBA1BvEPm-yc9Hii197ib8oM_cChWJ29vlQ9cBMlpfrY_Ng4bs7j520CPgQC536ucW1qA0mpqf8ZyewQIhUryMLUWZEtEWs2mJCMFv1WGAyYUvExFPEV66eZp_VeuZmkyOeiolnXtRTf4M6ltN9HeETUzYNciSw-AnyM1sQK78CSrCaD-t-5gWbIEs6Y_PQTIwztKCztVN697_g.jpg" alt="" loading="lazy"> <p>Negativity in the media keeps growing every year</p> <p><a href="https://t.me/chillhousetech/145?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And here’s a classifier that beats deep learning transformers. https://darkolorin.com/posts/144.html https://darkolorin.com/posts/144.html Fri, 14 Jul 2023 10:00:25 GMT <p>And here’s a classifier that beats deep learning transformers. The trick is it’s made of 14 lines of code. Still, the mass obsession with deep learning imo isn’t the most correct direction. <br> <a href="https://aclanthology.org/2023.findings-acl.426?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://aclanthology.org/2023.findings-acl.426</a></p> <p><a href="https://t.me/chillhousetech/144?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Typical premium features https://darkolorin.com/posts/143.html https://darkolorin.com/posts/143.html Fri, 14 Jul 2023 09:32:13 GMT <img src="https://cdn4.telesco.pe/file/GeuDScd0dArAd2NotzeUkdFLjtCJrgyY5O1ERtl9ht_DR3Wu7d_a4YzCwh3CbT8rKyTzm6KaP6xeGa4KPneq8IcOLMBt2iGW3-Axn5FbKQwV2YI8O5vVKtdnNrGHLca1BjZmKRBPb7qzYFNV99lAPx_Qf-Mjaf3LXByVxFEH0msBsUbzjCUsOpr16y95mylabzTfjhgnS4ls8jt3W5l-ZygPq79R0txyBkhdlRJMyl4v6upW_bDwZx8GoS5hSwrQecg6SFoC0GDJiaP8PlJqpvSIZI4kMxDMtizEobKWC4fnBflu-oG9_wsQ_JvvLNoWtsGre-lccSX6GApvPFdeHA.jpg" alt="" loading="lazy"> <p>Typical premium features</p> <p><a href="https://t.me/chillhousetech/143?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> In one of the channels I stumbled on a post saying that VCs started by… https://darkolorin.com/posts/142.html https://darkolorin.com/posts/142.html Fri, 14 Jul 2023 09:19:24 GMT <p>In one of the channels I stumbled on a post saying that VCs started by journalists are so wonderful. <br> And I think it’s the worst thing that could’ve happened to the industry. Think about it: people who’ve never built anything with their own hands, who don’t understand what retention is and why you even need it, who can’t verify a founder’s words (let’s remember all those interviews with company founders where the users are bots), start making calls on which company is good and which isn’t. <br> Feels like the LPs’ bet is that luck is basically your whole investing strategy and you don’t need to understand anything.</p> <p><a href="https://t.me/chillhousetech/142?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/141.html https://darkolorin.com/posts/141.html Thu, 13 Jul 2023 14:14:47 GMT <img src="https://cdn4.telesco.pe/file/Ll52Ndiv4vlSPFgcLri-XtCMZGUG6jb_VDQu-JgmRVTO6bI1-N-m5VpxQF_KPbC_eo1dDDRIK0PrGcHhLEbF2XbdM9WwqzWe5YgDPM7BAf1thZTpkYo-Zp9rioZbyj2zM86saVGuN1YPs5HHKU1CUu_CBOwu1C5WggDdDqIuMeBBwy30NtCXlPzukaFVmX9d_NhzbO666RjAjBWWMUVm2MqIAdazIoP4f3eGo-HCjpYchQ5UxP7axCmwtA9k0TB67O8VRNa_dEH5x7MkiE8YDzY__9asGFEWvgLLlJ1iNet6nMTrt3gw_U3qPaCc6VNhNKGlfd2jT6NT_22a0DkR3g.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/141?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Lately I’ve been thinking a lot about different companies and how they’re… https://darkolorin.com/posts/140.html https://darkolorin.com/posts/140.html Wed, 12 Jul 2023 20:18:39 GMT <p>Lately I’ve been thinking a lot about different companies and how they’re built and what role investors play in all this. <br> Out of a pretty big pile of stuff I learned this year, I pulled out a few points for myself. I’ll post them little by little.</p> <p>First. Modern consumer software startups/successful companies, for the most part, are so-called one shot companies. That’s when basically getting to product-market fit boils down to one iteration around a specific use case. You came up with an idea, built it (preferably faster — you’ll see why in a sec), launched it, and then you’re already looking at whether it flies or doesn’t fly and all that (I’ll cover that separately too). And if it doesn’t, most companies just switch to another track and run the same cycle again. As long as you have money, you try. Again and again.</p> <p>But there are companies of a different type. Like OpenAI, for example. That’s when to become successful you need to produce several sequential results that depend on each other. First you do massive research, or wait until someone else does it, then you learn how to run it and optimize it, and only then at the end you find a use case and ship it in a way that this use case actually finds adoption with the target audience. That many steps creates an unreal level of complexity to succeeding. So I think the guys are kind of unique.</p> <p>I expect the same thing in 7–8 years from Musk’s new company. And I wouldn’t advise any venture investors to waste time on multi-iteration companies at all.</p> <p><a href="https://t.me/chillhousetech/140?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Build your character on an LLM https://darkolorin.com/posts/139.html https://darkolorin.com/posts/139.html Wed, 12 Jul 2023 20:07:48 GMT <p>Build your character on an LLM <br> <a href="https://github.com/a16z-infra/companion-app?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/a16z-infra/companion-app</a></p> <p><a href="https://t.me/chillhousetech/139?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Just a must-read for anyone who’s even remotely making any kind of consumer… https://darkolorin.com/posts/138.html https://darkolorin.com/posts/138.html Wed, 12 Jul 2023 18:19:32 GMT <p>Just a must-read for anyone who’s even remotely making any kind of consumer software <br> <a href="https://medium.com/positiveslope/crafting-the-first-mile-of-product-7ed25e8f1027?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site#.751hpkf8k" rel="noopener">https://medium.com/positiveslope/crafting-the-first-mile-of-product-7ed25e8f1027#.751hpkf8k</a></p> <p><a href="https://t.me/chillhousetech/138?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The second person on the list makes me happy https://darkolorin.com/posts/137.html https://darkolorin.com/posts/137.html Wed, 12 Jul 2023 16:59:06 GMT <p>The second person on the list makes me happy <br> <a href="https://x.ai/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.ai/</a></p> <p><a href="https://t.me/chillhousetech/137?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Interesting, is this Paul Graham’s official channel on Twitch https://darkolorin.com/posts/136.html https://darkolorin.com/posts/136.html Wed, 12 Jul 2023 09:51:26 GMT <p>Interesting, is this Paul Graham’s official channel on Twitch <br> <a href="https://m.twitch.tv/notpaulgraham/home?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://m.twitch.tv/notpaulgraham/home</a></p> <p><a href="https://t.me/chillhousetech/136?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Dropped Threads, predictably. https://darkolorin.com/posts/135.html https://darkolorin.com/posts/135.html Tue, 11 Jul 2023 23:18:17 GMT <p>Dropped Threads, predictably. Had a discussion about it. Do you agree that in the world of social networks, one wins. By “world” I mean the format. I’ll unpack it.</p> <p>Say there can be only one text-based social network, or one network for photos, or one network for vertical video. We’re talking specifically about the main use case. I think that’s a debatable claim. Especially if you take different regions into account.</p> <p><a href="https://t.me/chillhousetech/135?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Found an interesting project. https://darkolorin.com/posts/134.html https://darkolorin.com/posts/134.html Tue, 11 Jul 2023 23:14:23 GMT <p>Found an interesting project. The idea is as old as the world. <br> <a href="https://intro.co/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://intro.co/</a> <br> Even signed up there, we’ll see if they approve the application. Gonna be interesting to post it around on my socials and figure out if there’s any demand. <br> I’ve been wanting to find something like this for a while, to see if there’s demand for my consulting or my opinion. But to organize something like that myself you’ve gotta deal with it and all that. And a service is faster. We’ll see.</p> <p><a href="https://t.me/chillhousetech/134?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A good meme account about VC. https://darkolorin.com/posts/133.html https://darkolorin.com/posts/133.html Tue, 11 Jul 2023 09:43:26 GMT <p>A good meme account about VC. <br> <a href="https://twitter.com/vcbrags?s=21&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/vcbrags?s=21&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/133?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/132.html https://darkolorin.com/posts/132.html Sat, 08 Jul 2023 20:52:12 GMT <img src="https://cdn4.telesco.pe/file/iIgM0dtLk5pydczM-asiaT5FfBPdZcM_wHZySGSxIccN_l2hLuiJ79YoRESZoX5Lwsu5xl5ewuzjPtgg9JMT---QD-e0joXn2WpFGTfjHdtqYoGxRx-893RXz7kUTWMoM6RQK10KwXI1c1Ay_pY4JHCVcZMP5q3dLIXBox_mc9VErmdFQ134g-yPPbZlZSS798v8tzQ0CTlWJn91ffDTwMSAvNbetopfs1zQZzUeleuet5iBkb6yDHngpySujVi7tJgzCChjRmSSrt__0N0zxgi5cZvfk1A4t3nMXH1aZBDpZ3DoSBfKeBKjE6MBHs7xXaXHUk9iQfZ7jFncH5rT_Q.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/132?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Ok. https://darkolorin.com/posts/131.html https://darkolorin.com/posts/131.html Thu, 06 Jul 2023 00:12:15 GMT <p>Ok. Let’s try this out. <br> <a href="https://www.threads.net/@darkolorin?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.threads.net/@darkolorin</a></p> <p><a href="https://t.me/chillhousetech/131?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Brave new world. https://darkolorin.com/posts/130.html https://darkolorin.com/posts/130.html Wed, 05 Jul 2023 12:37:43 GMT <p>Brave new world. The EU regulator banned the launch of FB’s Twitter analogue because they collect too much data. Like, the app isn’t even being used by anyone yet, and there are already complaints. What’s more, for basically all data on modern OSes you have to give consent (well, almost—when it’s one ecosystem there are actually loopholes, and this is exactly that kind of case). But the very fact is pretty damn funny.</p> <p><a href="https://t.me/chillhousetech/130?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And just to sober up sometimes in terms of how you assess businesses and their… https://darkolorin.com/posts/129.html https://darkolorin.com/posts/129.html Mon, 03 Jul 2023 23:33:07 GMT <img src="https://cdn4.telesco.pe/file/VjVvBM5olAthARXhk-otAWCo-eZzzcGKatzafx4KhjFDSqlgC_jsPytG7pJSLMd8zq1se15AQc_f1uk0igUwlc2yD_CtbQC6wVKa5iq731AGZO7la8UetzIlMRi6RyZNLEDXz5dS2omL2gaZvEF5BZCKE7MLkFf8Pjxz1CpqE6rwvBAWx_PcffrrhhQQap9-BCkVuezR8BNHF-WOz6BzEz7qh38kVElqGfS-2Xr7NcWRKvP_6G7hDevUqSc8VZ1A0I4gucttsE_kAMFEwdyba74d3Q5bK7Z62ozy2dOfqVdF8RdTs5gGXP5edUMvw5e6Eq7hJAQQjzUNODlat5HtnA.jpg" alt="" loading="lazy"> <p>And just to sober up sometimes in terms of how you assess businesses and their scale. Everyone of course already knows, the info’s been going around for ages, but still.</p> <p><a href="https://t.me/chillhousetech/129?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Preorder for a Twitter competitor from Meta https://darkolorin.com/posts/128.html https://darkolorin.com/posts/128.html Mon, 03 Jul 2023 23:22:59 GMT <p>Preorder for a Twitter competitor from Meta <br> <a href="https://apps.apple.com/gb/app/threads-an-instagram-app/id6446901002?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://apps.apple.com/gb/app/threads-an-instagram-app/id6446901002</a></p> <p><a href="https://t.me/chillhousetech/128?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Raise 700m on scooters and be worth 25m now. https://darkolorin.com/posts/127.html https://darkolorin.com/posts/127.html Mon, 03 Jul 2023 16:03:51 GMT <img src="https://cdn4.telesco.pe/file/dV3E75zSdw6J3QGSwzM8s4QblAzkWQz0xQgE7PyjDzV0XxKcTHdi0Jj8uUJQqxJwntlCZsGmWsqxvMt31O1CAS9y8h0AQ5IS1GtyWCzp75SdUQewzohTlBAUYViu08GCS2Jz5hR_hyypqjdeTvtEpotRoK6W4xpoL2vY1ot31xLtcRF1qWm_di38igWqOdgESjnaaNz6EvMk0nHjFr4ssiygPJOMOaAnZ1wTnSSWU9M4tgYRwvpCpz8OmqY6MhVa8DCkV4xPQ0zzSJVvK87hm2F1tD-UPXtetr2bPy6NO7iVQMOxOIjoRE7m8peWt1I6_myeMx1YcrzGlbWtSf1MpQ.jpg" alt="" loading="lazy"> <p>Raise 700m on scooters and be worth 25m now.</p> <p><a href="https://t.me/chillhousetech/127?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Peter Thiel did an AMA 9 years ago. https://darkolorin.com/posts/118.html https://darkolorin.com/posts/118.html Sun, 02 Jul 2023 14:52:28 GMT <img src="https://cdn4.telesco.pe/file/iHL1Yl4sUodbNw-Gtb73DVStykarsggP3Tqf10gvSAMxOzi6ftWKCPrbPaONa0n2rtjRu8kBkZoO1tTs3KSO0TTA4lt3E4-KQ3GPGouH2FYthsAzc7xvmgbP2Aaz3SIKL0ldMLMEjmyz8VI4X-h2N1GUbon7HQN2SaxF5-2QdpHoS6fZX2U8VwqZw4req_nGuCrCJEFyKcjrTlBrJdpsL_24JlhtYMtprwn27jhTA6yiguW1tVDEwRAc-ggsjT83EzsgrufG6eyHJ3yXL-tbmd8ghNBgFh381TgQKZd7az3CNtJyd8YDaEL6HsGeUGGe--jUSeOGjHdR09j2_87m_Q.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/CiyNO4LHGcle9q5eDcHEH5LuVcwPYwMa4NyTTAK-KiYGe4fK7yVUXGKIYvfJYKkidML_491Iss9yHLU9Xys6lPoI4fDrisf4aXHaI6ebqfTRnalhcicD2h1pjLOs6ig-3qRgCmV5pyRXBdn2L3uR3mZLMB7u5GjKFjOXBiFFhwOfS87aXbSemjDIXsvK3cWJY4HeERGPkagoY6nbZ1IwEC7nsJjBqMQJpDyVQjpQxrlvmGz8i-dJsI97x2yUh7uAXSbOE4PspMS9XEaOXWuShFwHgOhJmorU6deH0VVF3udbjhS0wA_hGpk3cZu-1a8w2frxgf0rJPm0sQVmwRci7w.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/jzfCPEpKrE7VZ_NIrO4sabl3sQIs8HO6Dbh3m93zDhr3W20QFi6ip4IcNYOSRfy-zQRF0_fLsF26apQfxYv5BE1prNZTzeknulT7SetgnPgQ5pNWQ6Dg70nm83cT5lvFU7q5n-GxDn72GTW0FK-DQXUivn0XIWG3OjoNMDkllLAZi_Ybt5qbMU_aeBF8gjEuZ8jmjaiTFFJRrTERNBXgSf0YbltJBfTydUnyBNDeAku9gqSrwHzI2wIjeAAiGTlyr-u8NW-7mtsMv6gv6xb29VLAtLb-scNOYVwc59iUK4fVoBJ120USnUpu-sqSbaH1_mxNvM4ZrY9VIBxQGXQJ_A.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/BhvFQaCmhQxHc6dXb1fP1BCJMNgKI80vAzMvsM_dovdWXXYQNXaYBTWJ61spsDtPuLhxWPXvNIAuGxOOsZYQ5CAW2gJuyR3CfAK37-ZogyuDr_eI80_zhOSEJGuqUuIwEf7Ebh00k5RDUQ6hQBnYOqnyMxZXKqYBk4MkVMf42McaM8IxnHpLkQzAAz3a486eT79lsAtT03g32M7EK93fu1uKKWVt1lYNghYYt83WfnnBcsTIm9Un3zr0Yz1NbAuFInLGqHeN7e72X539-kO2H_D2AzSsERtT9-wV0NQtovGLxhNIeOwkTkeKmfnKMphbAxHUKnzjAANJREGYny6WQw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/ceVv_CtjEE57XLv8Ly2QcKlFV3fx79OrJf1Jyame9SdMaZUcTXCy5lWnjwBHpOdx7Qg1kIo2wjSHH742PGy9aU4x2PGMt011SMYQB1Q6bbb8oe7GjCaYq0LRsRyh54B-O2iKZgv5F_PN-1ditR86Gg-Tm9p_LK1KerkDMIeQQouhxQ8NXWA6flJCkNY-toNIvNhFHYHNe5WojK-i_c-fLgeu-hSLCWRC5aqzNXjbULKKEEJzgQivCVX6m2VPOWwS_wCvYmDn_jKgmIwcTZ8856CCyR-rTb8mdyFTsh9vQpPXy4Kr9uMwhpe-cFcmrGN-_lixV_5ptcfeYSfF7Wd2Bg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/sysNqMJp8yFHgHxRZavoxjn12wAGqYfc6LifibBLHpSo1sMBWyxt0ugJdSj5rRo_-6MzQrFJgBhP_LE-Qr-eGF2S8Y-kJEarlmKGT91F2dZCC9C4xGUN1KbdWkqa6bfJZBdlxYpnSyHT0byJl_spyhSMdKqIHmDiuLySp-EWlPzrhnuONPhguzMEtIwqtnbV5NfmUzQPbAREOe5qp_VjCtoMGD1bnJos_wltHWCk3vYzVG5P1hzpjZYnxmqqyZi6J4kt5KObbx5x_BXVWz6ZkhU5e_2DC1FV8Ej_xLQUXEtr20mt9eht-xMF5-CTzlIzhYzHeOzNYM0DduDRUEliSg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/ulT10Ij1CvPvjvpkGe8AgSOjlMHPDdGmW0R-1q8uiORZbR7ufslrIzL_XELMQv4UoDIOYe7F8Z6eunETsqP8nOOepANoAxjRqH3ZRaFn_dY--n7GhUl0IP2A0Vg39CAPqzDfj_wDdD6mVgBAl4oBk5lQUCK8KU8Lch_Heu1Spi5zid96Czo6ANpqXkMDF4vSWp7pKAfxUgs50P0yHk4bxn7Z_aZLrAcMvxPSuOPIqGzejDB9BK8ZiL5GTdBJ7fHHXrAyO0j8LTPJjXxY-GVGk62gOJ7QGf-gEDdX4NNMWSVIGsLd4Q9KFxj4Vu1haiCIYBpkkp8m0eeUSRDRAlrM3Q.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/AUmz6fiadD0puuH-W5tVzJkV3LvmhSTlcd1ps7auJyx37b8806Sn2JbGY1OYjg-gk80_x1k3l01-iZIuzIIuCIdNgqQjq_Q8J6P1VgT9ZSy47F5uz0ely6Dxnh1NrOS21c7f6_7RbEnVQY8-CibkFl-W7Z-xnHRaZXknvCa5eMRQB368ASi0NZkjt81zTI87eHktjROQqJQiwofh7t0VShofvgYprM6BpHbfHNfcOFXFAz_ZpTlxZ2NEs_PGLGS-aTAgsv742sCJwtz5eyOOEXdyE3jWWwtL3O6C3qLMKPtDHOH0GqquCUJYEFiXA6nX6dfnDYDpxqgtJCxWfiHAnQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/OZqZJ7X2Dkgh49Y0JJiDzlj1ntR9Od_bmtd4fmMiZXjMXmgKtE2otEA1BbNPS4jZL1kkbZUxwv829prEwBMVw6NNTW4_7SjBpPvsR0xYL5UqtoTMpHNcs4Be52yAY1C7IW0ooYAs1UR2puDA1zvR1wKtp_d7kmG3b42-f_hkUA8opB02PM3_Q10LDmJLKq9rYYepNhbmZNfIXF8j7Op9bLiNE5YRQ_URQLF6SXMlP-E97hNlr23JBFFekPYT6HsMWpTXSS0CRo3uTC0fi733Ses1_6FTDy8bgeIpRr8uWnKuz7fPdAkKGEKQX-iXZtahKJj2rwMtmCezX9NwRLjMGA.jpg" alt="" loading="lazy"> <p>Peter Thiel did an AMA 9 years ago. Found some cool answers. Sunday stuff.</p> <p><a href="https://t.me/chillhousetech/118?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Looks like they’ve started doubting those AI editors on text https://darkolorin.com/posts/117.html https://darkolorin.com/posts/117.html Fri, 30 Jun 2023 15:14:33 GMT <p>Looks like they’ve started doubting those AI editors on text <br> <a href="https://www.bloomberg.com/news/articles/2023-06-28/stability-ai-secures-convertible-note-after-fundraising-struggle?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.bloomberg.com/news/articles/2023-06-28/stability-ai-secures-convertible-note-after-fundraising-struggle</a></p> <p><a href="https://t.me/chillhousetech/117?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> It’s hot as hell, I’d fire every product manager who decided to ship an… https://darkolorin.com/posts/116.html https://darkolorin.com/posts/116.html Thu, 29 Jun 2023 21:07:57 GMT <p>It’s hot as hell, I’d fire every product manager who decided to ship an integration like this. <br> <a href="https://ouraring.com/blog/introducing-oura-for-slack/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ouraring.com/blog/introducing-oura-for-slack/</a></p> <p><a href="https://t.me/chillhousetech/116?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Miro has already become basically a pure Enterprise tool that people don’t… https://darkolorin.com/posts/115.html https://darkolorin.com/posts/115.html Thu, 29 Jun 2023 20:48:47 GMT <img src="https://cdn4.telesco.pe/file/f2UqglkhXKlIVGGNxhEUJ6ZbS53r1JWJMprqMchm4p8rKRFfXhLuAj8hYty4A0vG0iOklIwjIoAhUl6Gr__xurc4XhniLqwXK7CiWoHtoGbQhOb9mWkWm_s5Me4DEjXDM-6P5t6g0PI4FferYAp2ArWBHldZFwobvNSJuacqivZ14ba4afPvfSe4VUhLHubZfdTlzNuYj7xGitU0BcIVeB_TRqCH0Ra4sN9rXwLs9OkPgksJqmXLMZZNzJOOiVAgI9ZC4XUsjmMalLgxQqxGrk2wajgZQdD8_zbPeVITJGZiG2aNlMZmaHcHtmvbbBuRo3jed3LU4pUqMWpWQFwgng.jpg" alt="" loading="lazy"> <p>Miro has already become basically a pure Enterprise tool that people don’t really like.</p> <p><a href="https://t.me/chillhousetech/115?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One of the few fucking awesome deals. https://darkolorin.com/posts/114.html https://darkolorin.com/posts/114.html Wed, 28 Jun 2023 23:22:35 GMT <p>One of the few fucking awesome deals. In cash, yeah. <br> <a href="https://techcrunch.com/2023/06/28/visa-acquires-brazilian-fintech-startup-pismo-in-1b-blockbuster-deal/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://techcrunch.com/2023/06/28/visa-acquires-brazilian-fintech-startup-pismo-in-1b-blockbuster-deal/</a></p> <p><a href="https://t.me/chillhousetech/114?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Will land well. https://darkolorin.com/posts/113.html https://darkolorin.com/posts/113.html Wed, 28 Jun 2023 22:30:59 GMT <img src="https://cdn4.telesco.pe/file/L8fCYQYzFQJplfiGuWqrpuYnkicP4VoZfUIPMuYOPOb7O79RZbONebpI0zK-_C148UrD9tlGfaRRi6KKteVHqDz_e6KaVmaQ7eEU2JH1jIsCYq8f9n2l04zzOwJW5ynRd8_ADTTwoFTFxlYQbBrNaUWktfnRowxaaD83YS-JuCqxCxWWQIQyywy4nzZmMJJrWXJqVHCRGBir4liz-v8ESlStB8VTh0rdVgcZhZbbgLPj6losJXdaFEyZrlTqxjm5ls-PgIUx1Vmfjo5-SVCpE7Vio1iLMJ1nWlTAp2pTzSS1VKSilIJ1XYxrMyxNHbPs6fNBPQslKF9vbPXEVZZTZQ.jpg" alt="" loading="lazy"> <p>Will land well. An ecosystem feature. I’d also add it to work video calls and the meetings status.</p> <p><a href="https://t.me/chillhousetech/113?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> An mind-blowing number of tools like this is what’s waiting for us. https://darkolorin.com/posts/112.html https://darkolorin.com/posts/112.html Wed, 28 Jun 2023 21:43:03 GMT <p>An mind-blowing number of tools like this is what’s waiting for us. And now a question: do you have any use cases in your work similar to what’s shown in the promo video?<br> If not, can you come up with a daily use case for a tool like this, knowing the output quality will come with a small share of hallucinations.<br> <a href="https://twitter.com/suhail/status/1674124521543192578?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/suhail/status/1674124521543192578?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/112?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ll say it one more time. https://darkolorin.com/posts/111.html https://darkolorin.com/posts/111.html Tue, 27 Jun 2023 22:17:30 GMT <p>I’ll say it one more time. Telegram should be doing a video feed. In the meantime, you can enjoy the first episode of Silo, which Apple dumped straight on Twitter. <br> <a href="https://twitter.com/appletv/status/1673730750095712256?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/appletv/status/1673730750095712256?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/111?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Sometimes a meme account with made-up interfaces spits out perfectly usable… https://darkolorin.com/posts/110.html https://darkolorin.com/posts/110.html Tue, 27 Jun 2023 15:27:20 GMT <img src="https://cdn4.telesco.pe/file/TjJzqScO4aUV1i57tViO2isIjy55AquuxuFfXobIKf7GgIcIL9UYA9bB2Djx7jN3FQEcqPSIhQZ8iGw1OUeguMJDVKYb7ieExlaC7jpljIwsCOVLBY607JyVocqrVy9-_h1J6iYy1PaLrQXmQCi-Lo5mPdqr6e9jDxuAhFxS-sDirCEgKmVTVwCCHaMGWUfVjPl0xfI3SvfzVxGL6TYXG6ebSP5IRPxT7xxFCec-2JrcoZVdQNrEzOB7QftrcmdMq6f1k77SmgXAr9CLjvzrioyRwLWM4vkkqx1te9gXFPMRXKffaBuYSudwIjdpTJdbPcTiDCIJuBDXe6LDvMaatw.jpg" alt="" loading="lazy"> <p>Sometimes a meme account with made-up interfaces spits out perfectly usable features (the main note is don’t read the note)</p> <p><a href="https://t.me/chillhousetech/110?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> 😄 https://darkolorin.com/posts/108.html https://darkolorin.com/posts/108.html Tue, 27 Jun 2023 14:57:19 GMT <img src="https://cdn4.telesco.pe/file/P9JeHXkrVkh9uL4BH3Jz2tJwsGnnu4payqOVr4F-vjb1QxQkr4mayPdk1b3wfqF1FaRjk_kghDg1Xm6Zx3OxaHsWvkrlW1iMu29E0UzcHe7aDLpAJqzsgzgkl6dayZm-4tJHIoJpILxqiXmIu3Leu45iJDAcSNt7wNlFq8TrmWnaI8L7kRynrNePji56PFVz53Rg8BE-WIUpv8Di18DuqInXpOP9n9JIcatjB7wtGJ0K7RgBL4Iz1ntlGKmTDbhHgztnuc7stSr1zQZTPJ9wwjcGWunGkLMS6fXqSANUgUSKvUeeqFAJzADPo5tKtooP_kmTLzgED3cXEopoi1NrVQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/Nr8Jm8puRLt8iucHEMoezcD3X3l40xIDVMRPX85cpkW271_Bi_zZ9tIHvb6aaxiZZ4p3s6N7uOoDfkTqDPDFrJ8YtBX6mLdvQvhNb69rklgnnjxWsCkp9j4rB1THyrxMMG1_10ffdZmsRUubI8slrTI9oTbOzfMAj-muy_Eh8_J6gVNJMHoekmfAtYzy0c2Z4VH6CKZG_oggw82pEjBN31P9DOPkudsZQHGHylvNy2GjguV3vW6xLnU3zzZdUhcXDPGoFn6NuWjbElRaVWpzLtnUb-kvW7vKjtOmr2e9uG9rU2ethVkGvSYzkNsZxNtXZpkqqh2WNdFYaKgMQ0vshA.jpg" alt="" loading="lazy"> <p>😄</p> <p><a href="https://t.me/chillhousetech/108?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’ve believed for quite a while that young minds are a really important part… https://darkolorin.com/posts/107.html https://darkolorin.com/posts/107.html Mon, 26 Jun 2023 21:59:33 GMT <img src="https://cdn4.telesco.pe/file/o57ARiHXuT9eicQcDpC5SAxOIwzrMZoRN6RjhftDMGP6PPjI2V5TjgdgPVZRHu5MX_LBEPphdTH8YGsHR-Qs4CGyIBCFxL2CZhwILgURMR55398or2sucnzeH5mMlMi2827eb2-TeeQMPL1_3GTNxLy95MG79csM9xM0N2JiEM7ZLXGjfnqylN3UstbPcVcLDvpcH5udQkYt9iY7O99fmimTbavmL32mi39ooXCtarjZMxJ0jSqRlVW8PuQXmaQhGizIDoz-MAo9lFpQ2dGXLKNmHbq0_Pgnz9bpXgXwhBj_vMwBKJhLugb4yPjrwhr-892gL9wXItNlaPPRj8k1FA.jpg" alt="" loading="lazy"> <p>I’ve believed for quite a while that young minds are a really important part of any company or VC. Especially when paired with the right experience (you don’t necessarily need decades of it).<br> Below is an image with data on the ages of the Manhattan Project participants.</p> <p><a href="https://t.me/chillhousetech/107?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> well, now there are stories in Telegram, and what we need isn’t stories but a… https://darkolorin.com/posts/106.html https://darkolorin.com/posts/106.html Mon, 26 Jun 2023 19:57:45 GMT <p>well, now there are stories in Telegram, and what we need isn’t stories but a video feed, but whatever</p> <p><a href="https://t.me/chillhousetech/106?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And some musing https://darkolorin.com/posts/105.html https://darkolorin.com/posts/105.html Sun, 25 Jun 2023 22:50:17 GMT <img src="https://cdn4.telesco.pe/file/YGrFyUG-DOFVvpwHUjz9jk7lN6RiAkSbrdxY9IzfzPDUoIryeDhy8jcDwlgpZQHAQdNuAlWjXttChmWpz9E8jze_UKPWKRGRwMDRaSi5Eb2tE_Z1inIs5BTG1prx45HvyqIpfA8k3R9650Ua8QCTo17HfvUacT4W9uu81VTX-WBT1jRXZqF4KzpaSNisTkWaqk0LCiBsmRYf2sgwincFJI8bCbTxwZhw-V1mDnE54PivnnY1h3ZknB5LytfZ-vm4akPSBa6IeYCvXSiRbp3tXvwlTWfmF45QE9brd26ixkOqvKLoRnBexF28ehhEywOE9OLaAd8pQBdaoN_XJCBS2Q.jpg" alt="" loading="lazy"> <p>And some musing</p> <p><a href="https://t.me/chillhousetech/105?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Weekend essay https://darkolorin.com/posts/104.html https://darkolorin.com/posts/104.html Sun, 25 Jun 2023 22:45:57 GMT <p>Weekend essay <br> <a href="https://dilbertblog.typepad.com/the_dilbert_blog/2007/07/career-advice.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://dilbertblog.typepad.com/the_dilbert_blog/2007/07/career-advice.html</a></p> <p><a href="https://t.me/chillhousetech/104?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Just the headline and the gist that 95% of the users were bots is enough https://darkolorin.com/posts/103.html https://darkolorin.com/posts/103.html Sat, 24 Jun 2023 12:29:15 GMT <p>Just the headline and the gist that 95% of the users were bots is enough <br> <a href="https://www.theinformation.com/articles/social-app-irl-which-raised-200-million-shuts-down-after-ceo-misconduct-probe?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theinformation.com/articles/social-app-irl-which-raised-200-million-shuts-down-after-ceo-misconduct-probe</a></p> <p><a href="https://t.me/chillhousetech/103?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Here’s something to read: an opinion from a respected VC on AI startups https://darkolorin.com/posts/102.html https://darkolorin.com/posts/102.html Fri, 23 Jun 2023 20:55:37 GMT <p>Here’s something to read: an opinion from a respected VC on AI startups <br> <a href="https://greylock.com/greymatter/the-new-new-moats/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://greylock.com/greymatter/the-new-new-moats/</a></p> <p><a href="https://t.me/chillhousetech/102?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And in general, in Tony Fadell’s book (the Apple guy, who at a personal… https://darkolorin.com/posts/101.html https://darkolorin.com/posts/101.html Fri, 23 Jun 2023 18:22:59 GMT <img src="https://cdn4.telesco.pe/file/Ll7hCpfh5ru6q3GjcueXUHhjDHfqsHvJC7D2wpEUnEXIHeWWLGj8mHIOAOrEJS3LaRpwrjgntfY9_MloVlRe0WRN_FgpH1F4FNyMNJ8RaHWrf4yFKIq3xzIbpX_MOCSSyng823tMQjsPBNXHq-3__Xh9QNBykf3uBeol6uikW2Rvp1LHw2IvHOqF6XbGgrxFInT3pcSDh9pIc-oJTjHFsafurfhdDN3y3cUWEbZWXz_zAJ9xG5bVJ83MCdTJjYhE6huelD4V6XdzBzTPzCv9CoZsEokG_y0EyrDNY7Cejq3V0XvdgAnLk0YcXKDhbP4w7E8N-TkF1NsDVBnY81xFbQ.jpg" alt="" loading="lazy"> <p>And in general, in Tony Fadell’s book (the Apple guy, who at a personal meeting called me a goat — asshole) there’s, I think, an important point about roles in building a product. <br> Your messaging is your product. The story you&#x27;re telling shapes the thing you&#x27;re making. <br> (Screenshot stolen from twi)</p> <p><a href="https://t.me/chillhousetech/101?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> About product folks in the West https://darkolorin.com/posts/100.html https://darkolorin.com/posts/100.html Fri, 23 Jun 2023 09:42:37 GMT <p>About product folks in the West <br> <a href="https://twitter.com/spshulem/status/1672062580708323328?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/spshulem/status/1672062580708323328?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/100?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/99.html https://darkolorin.com/posts/99.html Wed, 21 Jun 2023 20:21:37 GMT <img src="https://cdn4.telesco.pe/file/NJMfRA3417SRiCio6IDdoAX6L65lQ4KXwwY7vx2vo7dp1MJNQvgtgnVWQijm3_9MVtB9ioFBhHXVMfuBPLvufbj4L_Pv87zvhRuYUONgByaLyjyxPT5E7rwAxchMVCCepzim9_wKnvXpme01hmLqA_HadhrcBpHAiyDG7B-fWDkXQ90pLPCqXyrgzBbKTBGyVGpUqGCE2QSeFuYDLn70sGZg3jug2RHRbMUTz8AqYjz9MNihJDIVMVSonU8b9aGs-z_pXlw9RxM2RyknM8T1W_XcEFDeWqsC7ZOw5LgrVa7AX6o6vUO9MGfq2JZIYYW72XE2GCUAz3YKV-19UYahtw.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/99?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Now *that’s* what I call a navigator that knows the traffic light timing https://darkolorin.com/posts/98.html https://darkolorin.com/posts/98.html Tue, 20 Jun 2023 21:16:34 GMT <p>Now *that’s* what I call a navigator that knows the traffic light timing <br> <a href="https://twitter.com/josephcohen/status/1670941332544671746?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/josephcohen/status/1670941332544671746?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/98?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Some light reading about Apple Vision Pro. https://darkolorin.com/posts/97.html https://darkolorin.com/posts/97.html Fri, 16 Jun 2023 10:51:41 GMT <p>Some light reading about Apple Vision Pro. No revelations. But it’s a pleasant read. <br> <a href="https://www.ben-evans.com/benedictevans/2023/6/15/vision-pro?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.ben-evans.com/benedictevans/2023/6/15/vision-pro</a></p> <p><a href="https://t.me/chillhousetech/97?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Been waiting on this release for a long time. https://darkolorin.com/posts/96.html https://darkolorin.com/posts/96.html Wed, 14 Jun 2023 22:51:15 GMT <p>Been waiting on this release for a long time. Welcome to the local world of generations <br> <a href="https://github.com/apple/ml-stable-diffusion?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/apple/ml-stable-diffusion</a></p> <p><a href="https://t.me/chillhousetech/96?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A bit of cloning history, or how it all went down. https://darkolorin.com/posts/95.html https://darkolorin.com/posts/95.html Wed, 14 Jun 2023 15:32:09 GMT <img src="https://cdn4.telesco.pe/file/IG2V3aXnntKXeRBvwUt8tMK_EHAex6PDURuLqAdY6XLIFUWcmY4XljQyOk8JhAu6ZTd79VqxkFGB7ZFuEafVGjM1iSq5bax7Q2O6Xmiis1zSe6qJnDRzszqjJ6D8XUYs7iVe1NkKNxyOlu2rHn--k9YXAhkeVgDhMoM3TNX6DLwUk9z8nEnEUz-LIqecW-XohC-UcdIYhdjZszZfiZIZwsqRP2Bs0VQNTNzUykPM6sfpyFSyRwpUEWnzzeOCsLMJoztHnkewq4HwQ9goxQ3OcqdKxsS7iEmQmBm9VVxsguhPX54alQFSLCyfPgzmvEc4wa9d6stKn4msVDBcsM913Q.jpg" alt="" loading="lazy"> <p>A bit of cloning history, or how it all went down. And just how much the ability to play distribution matters.</p> <p><a href="https://t.me/chillhousetech/95?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> About the agenda https://darkolorin.com/posts/94.html https://darkolorin.com/posts/94.html Wed, 14 Jun 2023 15:08:49 GMT <p>About the agenda <br> <a href="https://twitter.com/aaronsibarium/status/1668658866337218560?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/aaronsibarium/status/1668658866337218560?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/94?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Continuing the Reddit drama. https://darkolorin.com/posts/93.html https://darkolorin.com/posts/93.html Tue, 13 Jun 2023 20:23:30 GMT <p>Continuing the Reddit drama. <br> <a href="https://www.theverge.com/2023/6/13/23759559/reddit-internal-memo-api-pricing-changes-steve-huffman?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.theverge.com/2023/6/13/23759559/reddit-internal-memo-api-pricing-changes-steve-huffman</a> <br> A couple things are interesting to notice here: <br> - massive pressure on the company to crank out cash <br> - it’ll all blow over, but it might hurt, and the main thing is to stay within corporate culture <br> These are important points about the current state of things in the market, and a very clear analogy to politics in general.</p> <p><a href="https://t.me/chillhousetech/93?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> City. https://darkolorin.com/posts/92.html https://darkolorin.com/posts/92.html Sat, 10 Jun 2023 22:13:10 GMT <p>City.</p> <p><a href="https://t.me/chillhousetech/92?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Wait what https://darkolorin.com/posts/91.html https://darkolorin.com/posts/91.html Sat, 10 Jun 2023 15:02:20 GMT <p>Wait what <br> <a href="https://9to5mac.com/2023/06/08/ios-17-link-tracking-protection/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://9to5mac.com/2023/06/08/ios-17-link-tracking-protection/</a></p> <p><a href="https://t.me/chillhousetech/91?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Right now there’s some interesting drama going on. https://darkolorin.com/posts/90.html https://darkolorin.com/posts/90.html Sat, 10 Jun 2023 00:38:10 GMT <p>Right now there’s some interesting drama going on. Reddit’s CEO absolutely faceplanted in his AMA. And it’s so dead in there, the dude genuinely has no feel for what’s happening in the product and yeah yeah “surprisingly” in the community too. <br> Just read it: <br> <a href="https://www.reddit.com/r/reddit/comments/145bram/addressing_the_community_about_changes_to_our_api/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.reddit.com/r/reddit/comments/145bram/addressing_the_community_about_changes_to_our_api/</a> <br> And I’ll still go back to the evergreen topic about the importance of one fact: company management, first and foremost, has to actually give a damn about the nature of the business and the product. That exact trait that’s so hard to find in people firing off their product resumes to companies they’re not interested in at all. What they’re interested in is a career or salary iteration.</p> <p><a href="https://t.me/chillhousetech/90?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A bit of Zuckerberg on the headset https://darkolorin.com/posts/89.html https://darkolorin.com/posts/89.html Fri, 09 Jun 2023 16:59:46 GMT <img src="https://cdn4.telesco.pe/file/gmt1cPiPm-iiOIoX4zcFrv3qqFKYohy-YpKehW4dwqO3FR-Wvm7UNbqQdIDvhY68SIMjgTUfAMUjQ7UdgsTxscY13CwwSs1rDyaU8WydFXS1sH6g2eZG6uDRbUMFSFJ751YuTWpHv3hj2pXQRiyv3yAnimwsmjaXsS60dj7g_9yNyV4qpivzQ1Stgj7opwTLC_SVKS1IbWz6PxEmJdDrnb4gAHbeNHqwV753DH9AZci1b8YWCTIf1rTP8lGDSD1SjZRD7ir1ZiKvDC_9OPB9BsOiqscgBo2NnlfUXZqVdQg4vbUui6Hl9Ag46D0nlDZ7-SbzmVNVM6ulo-oORzBVtA.jpg" alt="" loading="lazy"> <p>A bit of Zuckerberg on the headset</p> <p><a href="https://t.me/chillhousetech/89?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A roundup of my favorite account interface from twt https://darkolorin.com/posts/81.html https://darkolorin.com/posts/81.html Wed, 07 Jun 2023 14:47:39 GMT <img src="https://cdn4.telesco.pe/file/lU-kujMYT6iFxKqHgrdVHKaL3GPaANXhUgQ5IZji9CQJ1olmLP1QTuzsSa96aKXj8bTX-JZuRlYGWEV1h_r0XqQz1wtgWJd9B81JDI5AC6LJF53bOwYcYWOv4mkduP87rsjL_nLi8jXi0AJv7xolsHhpf7IOEV-mQ0yVqxsXexZk5EOFR_rMfnHCEFfO4kBBUhCIQkk-eJg2Wa-T5Rw8h0oi7mcy7c36fr9Djn1ocuTYmgSh6OV4fo90yl67mLCkEwQfhK0KMlbEUW1G6pirzHLzfd_X-dGY0jjKTdr7M542PNztIXEr5d2KIr4hf2DyYTM-IGFfROWWNZXF_s2FbA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/e8Qea6OcGy8Zmv7YrKsWIfcziIEQx4UmXk91iSriGzLXvnDgJqXcIm_f3EJW49vJs7J3C5kKMTJdqp4tXqFu3j15dPIEL3gjOClDsQ8TeSpaFAVk1ykRo8NzHPQeqFwGfzPHOkvz8C1UbLucLfFSsFjmdejjkcmPKtNlpuZ3sm91SFKJkTvV7pzXMCTU48Qg_lgZ15KxEWfqAIcSxLQcR5JmpwVyAgbA_mEbrVHHt45HAW9pXWX5CXJ3BE_GifYEpqyuS9OSoGCTzfJIjT6GmDb4RUH7NxRzVV4BobVP_qxJVFGrBns5g0qLfQMvirFgQHnSV88maAUTvr6vFUWdgQ.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/MchdJlryAur7G3Gl-SfEXwUaUqoxTFIFgSEzviiILSId1Tcj4zYdY-F6kkfz63KFQg0zAqe5NEyejO7XhmiukXfahTOcxuPiZ5GDfJ8zfZjusI-_Dl-tiwyjrCzuc6UuVP8tvmSvAF4Aq0VYzN_DKw8yOoBWg5E-RT_gxmi9C9sWXkFVgQUAUulmCCyQgCSpHTh7g7i0BWqxKUPBsiYfAkGf75Jy-4A90zqJBxpSG1_C2Esjo1VK-86I969GTQScNAmf2DG2yzbIohbl2cVTDHp5e_f3qI3kBTzp4t1saQcZFE_zoE8hyQu0wDxzl6GDny_RZ6rcFPYFZJdlyjCrgw.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/Zzm93iYHBz87qjidyJRYnPNueWxBNiZo3NvhnyDxz6v9gACcdM177JqcGt5raA5r5BCk_pGsfT_T2fPaZzOSIOgwypFY3l5wAVjjvlvaRr3QC5ZWrA6mq0T2zj8hdu_mxDZ-TDYOpa59oqnsVHgwu3nnioFOcd398XyJTdlR-vBl355-79PP5mm2n6FEX7ZdaBR45ck4YMDNEBCHfT8gOuWeGgwv_N95zPOq61Ekfrc7snZCBT6ipjnZ46cDEmtXQEdpPsDTqHof24gXje-ZrifMqcYbReQW1OBYCQrhhfka04prd5waPfa-M7lNM3w8tkG9sVdkPZdIaxZgqQJc0w.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/Hx4OWZHLIP7-jUbiinJM-nb55DDnqNT53n95I0uVsvkwi1MuG_fa5Ep0wogR5KRPbtpqs8fSeh7WuvhXC-NgK0J7Np66EPS0k8VWPhvb8omgFXETcy9Z2vKtyVuKBL-CvPyw4F04xUx5lbB1rI0WNY-R-uR4juoI1OK5rMqk8MQ7_ozOktQDtJ1QOHaDFNtDXYmN1gUVRXP4tyZbWcmle6VexDLMANFutQUCcqOt65okyuyuGM0UuoGcYeUevYuzJDaDRclwBIf1aZ4KYLQu6yqtPQSGDI70G3zwD5e3l-uFnvnZFyhWG4xv-W5S5-sJp4I_sJrMQIYaeJv-GcX9Vg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/ds-N364O4hCNbQWsL4DCZzhKxsJrSRlZTqKsMLPzWNUFdSYo3TRCEMTEpf475sj17lwkZSxbUVLIrwvEoujwP-AYg4yIWnP-JLsr9EQpHH2FNhPJmf95psP7XOmHCyI8igZsgKhFqygH2-biO4EiUybFwGqk5ZXY6Ns69KZCz3ZFzYF7uYbVG5kaeT6gb-za_bnIqeOp3T_JF8T649aAA4tbM1HYPqhxKPefx52bEMgvnF2LFNvAVGap9QENw2-MTVmUTbffW7As5Sw4mlWlq68zSQpoFecaS9C_QgXa-Y8qxhd9TPO7CISDCsv8kVm3fYEhq2Yo_3ZXZu0FgeWxsg.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/uslKNpTdH50MIBKyqYdaBbdfn5cJuBBOFnIF8rjOQcNBqwZjvIdo8ncTU6OgwppnseokQUyXgmeyYpG5sVE4EfTmohrtp__BcZTOkp4X6qHEzE8wHW-0K2KgCv-EH0UO-DVGbZaJd_fvs96_P6IXwBZYIQxAzHX7aq8jfHpB7GOFz9tTWdrPWwqHM27XJ1mToJUNZ8KV1qzZ9azM99tyDBIjK2vMP0Sr_WzVSW8AUkphjgX-WJDRzP7CA7XWtwR2BxxqArEwjZkIA7wHk64W3XQpze5gsVK4vHSIrMn1mYfzyT_as6ULOlhogm5OXPhgV7jOqjaAV4lVW6XixReDrA.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/T0hO2Jq8sCmVEH9dMKGx8l4bU1yDVqBxwWaimUf-8SziQZuOwJmfHYdnoYT92Q1RYQDeqEY1eOgKBKN6C3DrI5u7QMUW2Xy9wcqvgm1zjdav2VwZF-xU8McbNb7ywgKL43GHnXrHJpRDDktkE9-WM0Kb5NY3fQTQRve7Nk3HSVERTsmQxPYG-bEC0FWHR-QQDC4Kl4yVgr0fnAMBC1puLmtfKa7cGf0gTBiCJQx6RIfiFCpYf51Rk_Ho24AAQuB3G3eCM6e364y9IIBxnfRF4shvom4BIqxMTjHrFAFtPq9m6PNXPjH-obD2NIIGzEwo2IMkbjARXYXYV_geeVrlIw.jpg" alt="" loading="lazy"> <p>A roundup of my favorite account interface from twt</p> <p><a href="https://t.me/chillhousetech/81?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Love this author, worth chewing on his thoughts https://darkolorin.com/posts/80.html https://darkolorin.com/posts/80.html Tue, 06 Jun 2023 13:50:41 GMT <p>Love this author, worth chewing on his thoughts <br> <a href="https://www.matthewball.vc/all/bigtechbiggestbets?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.matthewball.vc/all/bigtechbiggestbets</a></p> <p><a href="https://t.me/chillhousetech/80?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/79.html https://darkolorin.com/posts/79.html Tue, 06 Jun 2023 09:36:17 GMT <img src="https://cdn4.telesco.pe/file/E_EhB5fA0GgBbMjtswe879J_WUXngkCBKabhyR9r3UlfNUzGD63fpWxzh1b0z8GpbPjrhTFF-l2L9ivnRH4JXQ1yWekHPcvnlXE-XoIDE7AxMoUWRaS_2xKUcjGDpdc3dVR-5zkPQIMWFKBL9NCDIqWzuVZtUuNeTadtH0B2R_wpouEgYRL6hlMX1-MYyGR4vwWdzF_mS-o3gZ5Klg9pUSH3jquFH2YZq10rlV_Hg4yz8ofAd_yDJCcV3CRqzayCyAxbkaE0eXWkvun5ensnF9U4KP2-ZQ22sGgBX3v2E_8rP-JkdX_qZp37qy9TUoeebyInDfq-jQ9ccEjF3HOysg.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/79?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well, here’s a little exclusive for you. https://darkolorin.com/posts/76.html https://darkolorin.com/posts/76.html Mon, 05 Jun 2023 20:21:20 GMT <img src="https://cdn4.telesco.pe/file/fXFhmjDeZB0swwtdikeVjgE4CjU4SX9rUaBG8zpZC_3DWDBVgUGVUV5HqCB2btSFJaLzZ0Hx44UxP0zIPzwbDYCsDpPidxpHtLEloolZrAaDf-BCIqT3Vlbj9_e_6xi2URcgHvoPQV89IPH69HLC8GEbWXqPML5f85bJr-qp8RyIhjE6MClJHrL5ca8m66CWquZn0MOL5XniY5DdlqN6zMJ0VarTso2Rv1RrD6iG0lPPV2tFJoqvfSK_2kJCUCT6eCtj7TJqSGodYQ_I6A68BhDyjnVr4FtjF6ceMWL-Wq6hd9aMKPiV-w6fk5Ii5J56j4Vz-ZZx7XgD8soHcbfmig.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/neE9IUJd3EYg7TsNmkSNkm8FXAZZZpLaVgGtb4J8PGGXMzdZgqv-IHKzkc1e1l3oSORDHWqYyB7ZbSa52m9HnhxJidhWr0dxqJTCU_D1zy3EIRLYLiFDSsaRQ9UieVXmOPFzs8pj0NxcIK7-n-dGelVCwR4IwOdKAhyvaWWyw0ZaSo1FSkEzjk_WW5X1TqDXjC_utNt3EsDaZn8FoGh0-6pwV5Ekqy-_IpNtHN_D3J29lZ4J_pgFtpqgQnwirOkTGS8xcTn7jWk0pBQXfyhgtm9hAUhuAl9yDuwGZ05mD84vCgKfmUppL_gsEWJUienc1y_wOLs4FMoobqB3WEev3A.jpg" alt="" loading="lazy"> <img src="https://cdn4.telesco.pe/file/czVkMhcLye7Ol6e_SmZSN0Hg86fuS01zCnEpQbmsqT46d0EgukbPgimUtttoqdsf3iYboU5tchdUuUvr_SF1y0slEKlTzTl3x9Ra_W38dQdpyrOjvzekI66oIrOZOje0gHIyoVdyDY1ME4Z3WUGnAx-t2IezTEcVFWKWJrBCj9-tPodG66C5bO6Pcqzto3GynuO3WpL68mrpu8E10Wx4o2e4r85Jb2eyfQN43Tpu8wr74rHY9fUZtSFNiHeHvNMyuVXMtVLYMzztNLLlYi0oWyAqq5ckUdCCN07kdHgJ9Rsuv8BDSiSC6_wA7ZJgoCnHU4VErlFJTPM92Qp2HF4ucQ.jpg" alt="" loading="lazy"> <p>Well, here’s a little exclusive for you. Once upon a time back in 2018, or even a bit earlier — we actually started doing it at the end of 2017 — we at Prisma made this thing. Kicked off by my dumb itch to play a real MMORPG in the real world. I pictured how the experience would change if I could run around with friends throwing fireballs and all that. So that we’d all be seeing the same thing. We made a demo and I even showed it inside Apple. Right in Apple Park. <br> <a href="https://blog.prismalabs.ai/ar-synchronization-system-by-prisma-3c41ff55a2a7?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.prismalabs.ai/ar-synchronization-system-by-prisma-3c41ff55a2a7</a></p> <p><a href="https://t.me/chillhousetech/76?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> You’re with an AR/VR expert. https://darkolorin.com/posts/75.html https://darkolorin.com/posts/75.html Mon, 05 Jun 2023 19:26:15 GMT <p>You’re with an AR/VR expert. Or, in short, my thoughts:<br> - the name has PRO in it and the price is obviously way too high for a mass user. Which means the target market isn’t you and me.<br> - a release at a dev conference means the goal of this release is to slowly take over the dev market, because that’s the point: the device won’t be mass-market anytime soon, but when it does go mass, it’ll need a ton of adapted content and software (which just doesn’t exist right now), and therefore you need a lot of developers and partners (especially ad ones, who’ll be shoving car ads into your entire house)<br> - the presentation overall is very strong, the company’s focus on end users is tracked extremely clearly, the ecosystem really shows its power, it makes sense. It’s the only ecosystem in the world of devices and software and it works.<br> - still not entirely honest 4K per eye, still questions about what the quality will actually be and whether it’ll make you nauseous, still questions about 2 hours of battery life (won’t even last for a normal movie, let alone a flight)<br> - and the most important thing is that it’s not the first gen that should sell, but the second. They made it stylish and now we’re waiting for big improvements by 2030 to finally feel like we’re in a movie from the future.<br> <a href="https://youtu.be/TX9qSaGXFyg?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/TX9qSaGXFyg</a></p> <p><a href="https://t.me/chillhousetech/75?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I can’t access that Twitter link from here. https://darkolorin.com/posts/74.html https://darkolorin.com/posts/74.html Mon, 05 Jun 2023 19:14:13 GMT <p>I can’t access that Twitter link from here. Paste the Russian text of the post (or a screenshot), and I’ll translate it preserving the exact tone and style.</p> <p>Links from the original post:</p> <ul> <li><a href="https://twitter.com/joekndy/status/1665738946150531073?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/joekndy/status/1665738946150531073?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></li> </ul> <p><a href="https://t.me/chillhousetech/74?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Out of the interesting finds, I got an ad in one of the channels for Eduson’s… https://darkolorin.com/posts/73.html https://darkolorin.com/posts/73.html Mon, 05 Jun 2023 13:44:22 GMT <p>Out of the interesting finds, I got an ad in one of the channels for Eduson’s course academy. Well, I think, I’ll take a look at what the prices are like. The CEO course costs less than the QA tester course. Funny how the world works.</p> <p><a href="https://t.me/chillhousetech/73?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Weekend quotes https://darkolorin.com/posts/72.html https://darkolorin.com/posts/72.html Sat, 03 Jun 2023 22:22:55 GMT <p>Weekend quotes <br> <a href="https://twitter.com/sta_maths_stats/status/1664945024243572736?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/sta_maths_stats/status/1664945024243572736?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/72?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/71.html https://darkolorin.com/posts/71.html Sat, 03 Jun 2023 15:59:12 GMT <img src="https://cdn4.telesco.pe/file/q_vkRi1cKnVUZg-8dI7kkXl6-I67rU8bKH65Walunmwgr2Kk5S-KzYOvEh11T84vuYpLQTACXKvW4jWBfna2iIkEKDkir0oxv71jGFzvEtyqAiLOaFsimRvZEs0dSpPEs5LjsM76D4aWdoSloCsuWU19G0fnYpLoZZv4aZReKIJMp4Y-7Hmb8NebtpHEt7tQ9XgzoJkqrDkbFszkkS2mMhQRJT0hN9jP1z0Ljqft4bQ1bB4d-O-uNmyjZdOFBGvLvAiegxUx3POS9IzwhcJ3_18gPL2UdO7sQzgqtxBFUbMNP8pLA1thWZCVMONJeEjQRXccyoTPbiKxAw4xXxMjHw.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/71?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I’m liking this account more and more https://darkolorin.com/posts/70.html https://darkolorin.com/posts/70.html Thu, 01 Jun 2023 15:09:39 GMT <p>I’m liking this account more and more <br> <a href="https://twitter.com/soren_iverson/status/1664270647672082432?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/soren_iverson/status/1664270647672082432?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/70?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Well now it’s all different, huh https://darkolorin.com/posts/69.html https://darkolorin.com/posts/69.html Wed, 31 May 2023 13:45:25 GMT <img src="https://cdn4.telesco.pe/file/nyvdaRC9QDm-aYzwFawt9kHpZwANaJrKHoVMDTuhsDWA_CYwqUFmNpOBKm7BEFVcZmCAypaQvSP3GvCuNWng2sjSou3Gcxg89RQ3EzetJE_8W67sDjGCtkSa9BjZ5p2XR-LaUV9Xi-OVRQ-3B1-xLdVS9s68TkraxORd0TVKvGIChEp_4SG8Dubvm2jhBNkG2m7VbLxB41jJip6S8xzzSxlU8VuFgvFm5_O3EQtY6-XucIIvvR2AVPZqawE0ewSaNggv4ollfeuIhwT1xaLCjC3vox4A5ztHVENGmNP1GJtKgiauRKDiMgyiEQFt16-JR_wQcPkNrQ_aD7H3wUwx7Q.jpg" alt="" loading="lazy"> <p>Well now it’s all different, huh</p> <p><a href="https://t.me/chillhousetech/69?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Found on Twitter a thread by some CPO, I think, at the startup Color. https://darkolorin.com/posts/68.html https://darkolorin.com/posts/68.html Tue, 30 May 2023 13:41:51 GMT <p>Found on Twitter a thread by some CPO, I think, at the startup Color. Not really the point. What was interesting was seeing this person’s chart and what problems this kind of employee solves. As far as I understand the project is like 300 people.</p> <p>So it turns out calls with the “topic” people management take up at least 50% of her time.<br> <a href="https://twitter.com/clairevo/status/1663372768992456704?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/clairevo/status/1663372768992456704?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p>Saw the same thing with product leads at Yandex. Their whole day was always packed with endless syncs and meetings. The funniest part is that I was on the ops team of this kind of project/product person who actually touched every feature with their own hands and talked to all of dev. And you know what, I barely crossed paths with them.</p> <p>How’s it going for you? In the CEO role I try not to talk and call inside the team too much. Because to make decisions (wow) you need time to think. That’s how it is.</p> <p><a href="https://t.me/chillhousetech/68?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> This is awesome https://darkolorin.com/posts/67.html https://darkolorin.com/posts/67.html Fri, 26 May 2023 18:02:48 GMT <p>This is awesome <br> <a href="https://twitter.com/soren_iverson/status/1662096315445129218?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/soren_iverson/status/1662096315445129218?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/67?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A minute not about Zoom, but about tech https://darkolorin.com/posts/66.html https://darkolorin.com/posts/66.html Mon, 22 May 2023 21:17:16 GMT <p>A minute not about Zoom, but about tech <br> <a href="https://blog.google/technology/research/project-starline/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://blog.google/technology/research/project-starline/</a></p> <p><a href="https://t.me/chillhousetech/66?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The search engine is shutting down Was expected. https://darkolorin.com/posts/65.html https://darkolorin.com/posts/65.html Sat, 20 May 2023 21:38:42 GMT <p>The search engine is shutting down <br> <a href="https://twitter.com/neeva/status/1659973810047832066?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/neeva/status/1659973810047832066?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a> <br> Was expected. The standard approach unfortunately won’t work. And they poured plenty of money into it. Still, I keep pointing at the product. First it’s a subscription instead of ads, then LLM, then later something else. People who were doing this still didn’t get that consumer == use case perfection.</p> <p><a href="https://t.me/chillhousetech/65?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I love how a company worth god-knows-how-many yards makes iOS screenshots for… https://darkolorin.com/posts/64.html https://darkolorin.com/posts/64.html Fri, 19 May 2023 10:30:54 GMT <img src="https://cdn4.telesco.pe/file/DZ30gKE_ojc-fMDh07fz5ixdvZtRMmRlw46c42v69TY3zOT6UdZE7EXUQ6nWPj-o-LGik7ueYqcRGbb8wHmuA9W-FBwwJ3fvae6j5I37m-hZ6ED4v9DHXrZrjSQTgrmscuVKg9lYcPxocPO48XDqv1AmDGdL-LL7VsAfl5H-asQaq43Kf66Ph0wcoL074oqW3_K7rUd7EaCYuRiY5v4G6as7DNU9dNJx9nZRBYn_56TJDb9UF1ffTTxx7P7jduCPRFNkfKGnIk-yhTGEAGdej4pogUtVzwZ0n9f5zMfKU8hM7YeKTjFNyH68Tz60UZUfBouGFs7PHCoZ7wZpg2Dphg.jpg" alt="" loading="lazy"> <p>I love how a company worth god-knows-how-many yards makes iOS screenshots for its product from random templates off the internet. I even recognized them, we use the exact same ones 🙂<br> All this, by the way, to the point about the importance of optimization and “other super-mega-important pixels” at release.</p> <p><a href="https://t.me/chillhousetech/64?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, about the numbers in announcements like these I remember how Zen… https://darkolorin.com/posts/63.html https://darkolorin.com/posts/63.html Thu, 18 May 2023 13:32:15 GMT <p>By the way, about the numbers in announcements like these <br> <a href="https://t.me/ostorozhno_novosti/16120?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/ostorozhno_novosti/16120</a> <br> I remember how Zen was the fastest-growing product on the market. And you know why? Because everyone who saw Zen on Yandex’s homepage was counted as a Zen user. I remember how “Alice users” were considered those who use their voice to enter a destination in the navigator. <br> Can’t wait for them to start posting pure product metrics — that means meaningful usage with an understanding of what the product even is. A rarity in Yandex’s latest reports that are about consumer software. All the wins have been in O2O for a long time already.</p> <p><a href="https://t.me/chillhousetech/63?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A pretty funny case: Facebook bought a company for a yard in 2020, but… https://darkolorin.com/posts/62.html https://darkolorin.com/posts/62.html Tue, 16 May 2023 22:18:45 GMT <p>A pretty funny case: Facebook bought a company for a yard in 2020, but apparently the “cutlet” didn’t go down and they had to send it back to the cafeteria. Not sure how this works and how many cases like this there have been. Maybe someone knows? <br> <a href="https://www.kustomer.com/blog/new-chapter-standalone-company/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.kustomer.com/blog/new-chapter-standalone-company/</a></p> <p><a href="https://t.me/chillhousetech/62?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Everything you need to know about “post-Soviet product managers” https://darkolorin.com/posts/61.html https://darkolorin.com/posts/61.html Fri, 12 May 2023 18:34:21 GMT <p>Everything you need to know about “post-Soviet product managers” <br> <a href="https://twitter.com/eshear/status/1656740583367639040?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/eshear/status/1656740583367639040?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/61?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Btw what are your bets, who is it? https://darkolorin.com/posts/60.html https://darkolorin.com/posts/60.html Thu, 11 May 2023 19:57:08 GMT <p>Btw <br> <a href="https://twitter.com/elonmusk/status/1656748197308674048?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/elonmusk/status/1656748197308674048?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a> <br> what are your bets, who is it?</p> <p><a href="https://t.me/chillhousetech/60?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I really regret that Sequoia stopped putting out these kinds of basic and… https://darkolorin.com/posts/59.html https://darkolorin.com/posts/59.html Tue, 09 May 2023 21:36:18 GMT <p>I really regret that Sequoia stopped putting out these kinds of basic and super important product pieces like this one <br> <a href="https://medium.com/sequoia-capital/engagement-part-iv-activity-feed-ranking-40d786b9d479?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://medium.com/sequoia-capital/engagement-part-iv-activity-feed-ranking-40d786b9d479</a> <br> The whole series is worth your attention and gives you knowledge about how consumer products are built and their key components. <br> My experience dealing with them was always extremely positive. The folks at the fund are really badass product pros and that’s a huge rarity. The team understands and is genuinely good at pretty much every type of product on the market. Well, at least that’s how it was in 2016–18. Maybe it’s time for new top investors. But I still think the key partners should have hands-on experience building product with their own two hands. Without that, it gets really hard for a fund to make decisions.</p> <p><a href="https://t.me/chillhousetech/59?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Weekend post. https://darkolorin.com/posts/58.html https://darkolorin.com/posts/58.html Sat, 29 Apr 2023 23:42:56 GMT <p>Weekend post. Sometimes I read various monographs and lecture series on physics. This time I’m reading, just to get acquainted, a quantum mechanics course by Paul Dirac. There’s a pretty interesting quote in there (it’s given as a footnote in the book, by the way).<br> “The fact that the idea of superposition forces us to generalize the original notion of the state of translational motion, whereas in the previous paragraph no such generalization was required for the state of polarization, is purely accidental and has no deep theoretical significance.”<br> Yeah, sure :) “purely accidental” (c)</p> <p><a href="https://t.me/chillhousetech/58?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Those very rules of the golden club. https://darkolorin.com/posts/57.html https://darkolorin.com/posts/57.html Sat, 29 Apr 2023 18:17:09 GMT <img src="https://cdn4.telesco.pe/file/agFDgGh41cB_RKAug95VbfP2KCyQVOYV2-SjrtsFC7gSr_zMGNHDJWMtwxmW9KCEsknKWBoBAaNMT-JyEWeSrfCCKf3-4cC9kUZJzpx7yU7pJbXITpFU7emsH0rggF38jux5OMRGDyrdL7P7c3184sBBBuRtqdY83-STQEmFqqtsO5MVHQRztRR59ZUT-N2P4MOxnfrL2aLMhW-Nwi-MKs8Mml2neJ6YfWEWQC-ogOr-vhKOUaq5YvZDJwsAjpyaBCjlo-uuvZibBzrzdTOKSzIeEV8vnqZloD5_qRAgONRUAV2l1vbn1bkgT2KDgD11TbwiIvbFXh86RgUmZ0MQZw.jpg" alt="" loading="lazy"> <p>Those very rules of the golden club. Either they message you themselves because they know you, or you’ve got such insane growth that they message you themselves too. Eat this, as Paul Graham says. Who’s apparently been stuffed to the gills for a long time already.</p> <p><a href="https://t.me/chillhousetech/57?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> I never stop sending everyone who’s interested in AI to resources like this https://darkolorin.com/posts/56.html https://darkolorin.com/posts/56.html Thu, 27 Apr 2023 18:53:32 GMT <p>I never stop sending everyone who’s interested in AI to resources like this <br> <a href="https://e2eml.school/transformers.html?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://e2eml.school/transformers.html</a></p> <p><a href="https://t.me/chillhousetech/56?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Today I did the final interview for the a16z seed program for gaming startups. https://darkolorin.com/posts/55.html https://darkolorin.com/posts/55.html Tue, 25 Apr 2023 01:10:13 GMT <p>Today I did the final interview for the a16z seed program for gaming startups. It’s like 15 minutes of a very weird dialogue. And out of that, 5 minutes you ask your own questions.</p> <p>So they make the decision in two stages:<br> 1) you submit an application and it’s like ~5 questions that are very not representative, basically two sentences about the idea and a link to the founders’ LinkedIns, and the most important item is who referred you (we were referred in there by one of the partners, Jack Soslow)<br> 2) and the second stage is this kind of interview with their partners. Andrew Chen and co. I’d had the honor of talking with Andrew before too. And there it’s basically three questions. What’s the product. Why it’s gonna be huge (1B plus)? And some follow-up after you quickly explain what’s what. For me it was about UGC and how we use it to build a new game launcher.</p> <p>Turns out to pass stage one you already need to be in their orbit and be buddies. They have to know you and understand what you’re doing. And to pass stage two you probably need to be confident and visionary. We’ll see. I’m not that good at 10-minute pitches when it’s about games. And games aren’t always that simple, because who would’ve thought a raid voice chat thing would end up being worth 10 billion. Fingers crossed.</p> <p><a href="https://t.me/chillhousetech/55?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A good book with DL basics https://darkolorin.com/posts/54.html https://darkolorin.com/posts/54.html Mon, 24 Apr 2023 00:05:29 GMT <p>A good book with DL basics</p> <p>Document available in the original Telegram post.</p> <p><a href="https://t.me/chillhousetech/54?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Dedicated to all subscription fans. https://darkolorin.com/posts/53.html https://darkolorin.com/posts/53.html Sun, 23 Apr 2023 22:17:16 GMT <p>Dedicated to all subscription fans. Dustin Moskovitz posted on twi that he’s ultimately fine paying for a blue checkmark. Because yeah, premium features in a social network that matters to him are ok.</p> <p>I’ll allow myself to continue his thought and moreover warn all startupers against making extra monetization of influencers the core feature for acquisition.</p> <p>A social network at the beginning of its journey is always made of UGC. Usually very low quality (early stories on Snap, YouTube with its home videos, and so on), then as the platform grows people get a dopamine mechanism that pushes them to make content better and better to get more likes/views and the rest.</p> <p>And then the moment comes when the network accumulates a critical mass of UGC at a level you can call professional content. And that’s the turning point for monetizing consumer social products. Because now people see value and they’re ready to pay for it. And what you can pay for is extra features, access to content, even watching ads is a kind of payment.</p> <p>This is a very important thought. Because it pretty clearly explains why monetizing consumer social at an early stage is just murderously hard. And it also answers why messengers still can’t make money — because the share of private p2p low-quality content will always dominate. Unless, of course, you try to play the channels game ;)</p> <p><a href="https://t.me/chillhousetech/53?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> There’s now a bung in every barrel. https://darkolorin.com/posts/52.html https://darkolorin.com/posts/52.html Fri, 21 Apr 2023 10:44:54 GMT <p>There’s now a bung in every barrel. It’s AI. Be afraid of that word. It’s everywhere now, even where it isn’t.</p> <p>The startup raised a lot of money, btw, and apparently part of the recent pitch was about AI.<br> <a href="https://www.axios.com/2023/04/21/humane-ai-powered-wearable?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.axios.com/2023/04/21/humane-ai-powered-wearable</a></p> <p><a href="https://t.me/chillhousetech/52?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Weekend rec: I recommend this art history newsletter Pretty damn cool. https://darkolorin.com/posts/51.html https://darkolorin.com/posts/51.html Sat, 15 Apr 2023 14:33:12 GMT <p>Weekend rec: I recommend this art history newsletter <br> <a href="https://culturaltutor.com/areopagus?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://culturaltutor.com/areopagus</a> <br> Pretty damn cool. The folks there do a ton of work. They started out on Twitter, then made it into a newsletter. You can get it for free once a week, which is more than enough. There’s enough reading and material for a couple hours, well, if you’re into art and history.</p> <p><a href="https://t.me/chillhousetech/51?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Genius. https://darkolorin.com/posts/50.html https://darkolorin.com/posts/50.html Sat, 15 Apr 2023 14:21:13 GMT <p>Genius. The current boom of generative nets (in reality, all the ideas in them are like 5–6 years old) is loud enough. And a lot of people still haven’t even realized how to use them, well, aside from the obvious tasks this stuff was created for (see Microsoft vs Open AI). Working with text and the like is preferably with a ready-made distribution. But literally in 2–3 months we’ve got this too <br> <a href="https://chatgpt.aiacademy.me/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://chatgpt.aiacademy.me/</a> <br> P.S. This is not an ad. I generally recommend not going there at all. Unless, of course, you’ve got a lot of extra time and money — then you can.</p> <p><a href="https://t.me/chillhousetech/50?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> a small AI news aggregator https://darkolorin.com/posts/49.html https://darkolorin.com/posts/49.html Fri, 14 Apr 2023 15:42:47 GMT <p><a href="https://news.bensbites.co/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://news.bensbites.co/</a><br> a small AI news aggregator</p> <p><a href="https://t.me/chillhousetech/49?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Hi investment courses https://darkolorin.com/posts/48.html https://darkolorin.com/posts/48.html Mon, 10 Apr 2023 22:18:07 GMT <p>Hi investment courses <br> <a href="https://arxiv.org/abs/2303.17564?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://arxiv.org/abs/2303.17564</a></p> <p><a href="https://t.me/chillhousetech/48?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> What’s especially interesting is the difference between FB and TT. https://darkolorin.com/posts/47.html https://darkolorin.com/posts/47.html Fri, 07 Apr 2023 10:28:53 GMT <p>What’s especially interesting is the difference between FB and TT. FB never tried to push its third-party apps this aggressively, but TT does exactly that. A bit more detail in the article (they pulled it out from behind the NYT paywall) <br> <a href="https://gjenuine.notion.site/TikTok-s-Owner-Pushes-a-New-App-While-Under-Washington-s-Glare-0226aac5ec6f4f88b26313deb326610d?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://gjenuine.notion.site/TikTok-s-Owner-Pushes-a-New-App-While-Under-Washington-s-Glare-0226aac5ec6f4f88b26313deb326610d</a></p> <p><a href="https://t.me/chillhousetech/47?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Photos https://darkolorin.com/posts/46.html https://darkolorin.com/posts/46.html Thu, 06 Apr 2023 21:01:46 GMT <img src="https://cdn4.telesco.pe/file/PRzQTDKg5DXSLu88-y_R1dGMvvCYWyeXztAfb02osBKH69WTSxIY2mKlyYGk5cSHqR2t1432cIq81DxkA9ac81FSJZS0f2a3-eH8a65q8tF4oWgh5r5sp4tbZ2yvlpGiVGWgvi7kv8Ep5S5C0XnA44AECyJ7pSQ6yFSg1OV99U380eyctvebEZJhYg2KF5XlqJRvf1eZ8iMhGkQ-FDXC-RGfhbNy0R5IIM_IlyMznDkSOhK4Vx7QkPutFsn4v932zF2kdPcYwry_cTN10k3H3Lqf8co8ujuan3iddYM7v9dEd_zQK6Yu-K9WnDJIq-YG_TgTU7oHjwtGiHqs_NZujQ.jpg" alt="" loading="lazy"> <p><a href="https://t.me/chillhousetech/46?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A minute of venture math. https://darkolorin.com/posts/45.html https://darkolorin.com/posts/45.html Thu, 06 Apr 2023 13:23:36 GMT <p>A minute of venture math. <br> In 2014 Leta Capital became the lead investor in a $5.5M round for 365Scores founded in 2008. As Alexander Chachava noted, in 2023 the company’s monthly revenue is roughly 1000x higher than its 2014 numbers. <br> Let’s try to figure out what kind of revenue that is. They bought the startup for $160M, let’s assume the multiplier to annual revenue is fairly low and closer to buying an asset as a business — x5. That’s $32M a year, or per month 32/12, about $2.5M. Assume “a thousand times” and we get $2.5K a month was the startup’s revenue at Series A, 6 years after founding. Well then. Further spinning the numbers and analyzing the purchase or its pretty description — you can do on your own.</p> <p><a href="https://t.me/chillhousetech/45?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Beautiful https://darkolorin.com/posts/44.html https://darkolorin.com/posts/44.html Fri, 31 Mar 2023 23:25:57 GMT <img src="https://cdn4.telesco.pe/file/cdwW5Dddkm0lyFFJjoKNzMbT5jA2RRfiupSvnk_qS3Uymnqs7CAyBwvDrkWD9chUWQlV2vJOdzJ45ZC_nWr1JHYxG8cI6sQY1jwsZrN8qjgnpEKECVSvJu3gdSzY9gI-z5Xz9DeXqDyBW68SZsCY9M3zszb8s__g-cdn5NtUjxx4IF6KzmDOyKQGNlFj09L-kLLs0Dgw8IGykxYq8EbX9lL_akmKUNdOHzEm8UuH0C-_DETg0VorNNiieV1WpWJg6YTcWfpsXGlwaqnFEEbYEb8CC4S_8sMcRU1-1Etv_gCd4KqD7CC8OnK1jk5OGUBf2ml6XqEMIAOK5xWEff6xIg.jpg" alt="" loading="lazy"> <p>Beautiful</p> <p><a href="https://t.me/chillhousetech/44?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And yeah, everything violates privacy anyway, what’s there to even talk about https://darkolorin.com/posts/43.html https://darkolorin.com/posts/43.html Fri, 31 Mar 2023 13:37:40 GMT <p>And yeah, everything violates privacy anyway, what’s there to even talk about <br> <a href="https://www.politico.eu/article/italian-privacy-regulator-bans-chatgpt/amp/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.politico.eu/article/italian-privacy-regulator-bans-chatgpt/amp/</a></p> <p><a href="https://t.me/chillhousetech/43?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Keep going? https://darkolorin.com/posts/42.html https://darkolorin.com/posts/42.html Thu, 30 Mar 2023 14:16:56 GMT <p>Keep going?</p> <ul> <li>Yes</li> <li>No</li> <li>Add more metrics we’ve achieved</li> </ul> <p><a href="https://t.me/chillhousetech/42?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Part one I’ll tell you a bit about our original project idea. https://darkolorin.com/posts/41.html https://darkolorin.com/posts/41.html Thu, 30 Mar 2023 14:16:21 GMT <p>Part one <br> I’ll tell you a bit about our original project idea. We started by tackling the thing every game dev and game-adjacent service in the world wants to solve (a sea of projects). Finding buddies/a team to play with. <br> At first glance LFG — looking for group — is a pretty obvious and big problem. For real, in a lot of games you want to find a decent partner or a team, and one that fits your schedule too, and actually gets it. But then the question comes up: if this is the holy grail of gamedev, why does basically nobody — not publishers/devs inside the game, and even more so not external services that aren’t constrained by the content release pipeline — really manage to pull it off? Let’s try to answer. <br> We started testing the hypothesis on existing communities in Discord. We made a bot (by the way it’s got the checkmark and all that, and it’s still hooked up to thousands of servers), whose job was to let you assemble your own group/find someone by set criteria, and the “killer feature” was supposed to be that it’s not just a dumb post in a channel — the bot can post to other servers too and also look at your game experience and stats. And even on ready-made communities it had mixed success. <br> The closest analogy is dating, and yeah, it’s true — we’re looking for people to play. And here it’s important to understand: the criteria by which someone “fits” matter, and how critical they are. Also the timeframe in which the choice happens matters.</p> <p><a href="https://t.me/chillhousetech/41?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> It’s kinda interesting that the consumer social company Snap is going into… https://darkolorin.com/posts/40.html https://darkolorin.com/posts/40.html Fri, 24 Mar 2023 16:23:36 GMT <p>It’s kinda interesting that the consumer social company Snap is going into commerce <br> <a href="https://ares.snap.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://ares.snap.com/</a> <br> And not their own, but specifically partner-based, and they’re playing it as a technology. It’s hard to say with confidence that things are all bad on the ads side. But it says a lot anyway. <br> I think if you managed to build something in your product that looks like an ad platform and it works, then right now that’s the most promising startup in the world (yeah yeah not all these AI things on one idea and an API). Because over the last few years only TikTok and Amazon are showing revenue growth from their ads. Waiting for new b2c with an ads model, since attribution and all that is getting more and more complicated.</p> <p><a href="https://t.me/chillhousetech/40?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The bot builder is starting to sprawl But I still think text input is… https://darkolorin.com/posts/39.html https://darkolorin.com/posts/39.html Thu, 23 Mar 2023 18:40:36 GMT <p>The bot builder is starting to sprawl <br> <a href="https://openai.com/blog/chatgpt-plugins?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://openai.com/blog/chatgpt-plugins</a> <br> But I still think text input is bullshit, not a revolution</p> <p><a href="https://t.me/chillhousetech/39?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> My favorite picture about censorship and democracy. https://darkolorin.com/posts/38.html https://darkolorin.com/posts/38.html Wed, 22 Mar 2023 15:45:17 GMT <img src="https://cdn4.telesco.pe/file/AEThry44tf-0z3qWpEgiOOjI8Pfuj91VlzxqyoBMAqtiFT6twMy66iE4aaej2sKovqpnM8vu9OSePJ84eYVO9jjIFLVyOq3Q9mtTaIR51aS3Vnci4mIiToYF0MhPaqhOSH7X2m7f1yHlDLHp6CGkp5NA10ZtGP9t2kyiofgtMmUw4tFkGSfyYgsny9nC1trn6qmODo7kobkbuYJaNsMc9wf34Sbc-BJvKwpni-hLY7LE9W47D9-xtj0kTGy6sEAEsErYFJbiSYK8thTqyWqG_kXEPy37v6LFvGdlVY5dT2YCFq9CGVyspq4GXWC52Hh9ZAl7Wwga1VYR8bC6PNW6nA.jpg" alt="" loading="lazy"> <p>My favorite picture about censorship and democracy. Wonder why Russia isn’t there — probably because they left it themselves lol :) <br> But all of this says that the era of global social products has most likely passed.</p> <p><a href="https://t.me/chillhousetech/38?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> About that very magic of pricing. https://darkolorin.com/posts/37.html https://darkolorin.com/posts/37.html Tue, 21 Mar 2023 18:51:25 GMT <img src="https://cdn4.telesco.pe/file/oCrSzSOHD-_ZQGZ3tc4h_mRItoAV4WbZZQiejHlhO_TKyeSiHr9ShxG34QkWll7AWURw2CYXgKFFm__Qx83xGTNk4-LaVC2qtRjtl0Ggm058Ep_xbH_sitAV6upwcGuu5wLqBxxM-GARKbTzWL_8ghOzRNj_dsQ41jwjq9KxyiKcbEV-sddcGngVsBmy3elnkDlIdvXV4bbrNVXOzMEtFsswZf1-XfiYg-oUP8Kvhma-W1n_f_bOql6qVbdT6TFWit3UIbvRhY7W2vL96LYqIiTykv-kaibK4WSUbQ1AzIpygMST5QJAT4zeb1H3dqzpY2dU0-IOzRqTWxuhmeaL8A.jpg" alt="" loading="lazy"> <p>About that very magic of pricing. Or how marketing depends on the target audience even when it’s the price.</p> <p><a href="https://t.me/chillhousetech/37?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Tech right now is like working at the plants and factories of the early 20th… https://darkolorin.com/posts/36.html https://darkolorin.com/posts/36.html Mon, 20 Mar 2023 14:50:37 GMT <p>Tech right now is like working at the plants and factories of the early 20th century. Like, literally one-to-one, the whole thing is playing out.<br> - a crisis and layoffs (automation + overhiring)<br> - assembly-line-ification (AI Tools)<br> - education will now teach programming or at least understanding what it is everywhere (writing basic prompts or using AI Tools — read: screwdrivers/hammers as shop class)<br> - the value of a manager and a lead, with an understanding of goals and approaches, of creativity, is going up<br> - there’s a fuckton of money in the system and it’ll be enough to give birth, against that backdrop, to a new industrialization (most likely biotech)</p> <p><a href="https://t.me/chillhousetech/36?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Pretty often I hear this take that when you pitch an investor you should act… https://darkolorin.com/posts/35.html https://darkolorin.com/posts/35.html Tue, 14 Mar 2023 02:57:27 GMT <p>Pretty often I hear this take that when you pitch an investor you should act like you’re basically hiring a co-founder into your company. Like a “looking for a workhorse who’ll grind with me till they drop dead” kind of approach.</p> <p>But that’s not true at all. It’s some myth from the Valley of<br> dreams<br> near San Francisco. An investor is an investor precisely because, at this point, he doesn’t want to build anything with his own hands full-time — otherwise why the hell work at a fund if you want to do a startup.</p> <p>And yeah, his job is to help the company, find the right people, bring in capital and all that. But not to work with you permanently. His job is to make money off you. And that’s honest. And those are exactly the ones you should be looking for — the ones who’ll see in you the person who’ll make money and change the world, and who’ll help you with their experience and knowledge and yeah MONEY!</p> <p><a href="https://t.me/chillhousetech/35?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> But https://darkolorin.com/posts/34.html https://darkolorin.com/posts/34.html Sun, 12 Mar 2023 23:32:55 GMT <p>But <br> <a href="https://www.federalreserve.gov/newsevents/pressreleases/monetary20230312b.htm?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.federalreserve.gov/newsevents/pressreleases/monetary20230312b.htm</a></p> <p><a href="https://t.me/chillhousetech/34?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And another bank bites the dust. https://darkolorin.com/posts/33.html https://darkolorin.com/posts/33.html Sun, 12 Mar 2023 23:22:33 GMT <p>And another bank bites the dust. Mmm, interesting how this is unfolding.</p> <p><a href="https://t.me/chillhousetech/33?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> About how the startup bank SVB fell apart. https://darkolorin.com/posts/32.html https://darkolorin.com/posts/32.html Sat, 11 Mar 2023 00:54:26 GMT <p>About how the startup bank SVB fell apart. Prisma used to have an account there at some point btw, I remember. <br> <a href="https://twitter.com/compound248/status/1634105371198644224?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/compound248/status/1634105371198644224?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/32?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Yesterday Spotify dropped an announcement of a new home screen design for… https://darkolorin.com/posts/31.html https://darkolorin.com/posts/31.html Thu, 09 Mar 2023 20:59:30 GMT <p>Yesterday Spotify dropped an announcement of a new home screen design for their app with a TikTok focus. Like, scroll video clips and listen to music there. A wave of negativity rolled in on Twitter and in the тусовка.</p> <p>And meanwhile Apple is finally rolling out a separate app for their classical music in super quality. And that’s a very subtle move. Basically, people who listen to classical are pretty damn solvent (supposedly the average age there is way higher), and accordingly they’ll all be ready to use a service like that, which doesn’t cancel using regular music too (I’m guessing there’s a bundle discount). And honestly it’s a great move and I think Apple Music will, little by little, start beating Spotify in the long run.</p> <p><a href="https://t.me/chillhousetech/31?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Kinda wondering how things turned out for Musk with the Twitters after that… https://darkolorin.com/posts/30.html https://darkolorin.com/posts/30.html Wed, 08 Mar 2023 11:12:15 GMT <p>Kinda wondering how things turned out for Musk with the Twitters after that investor call <br> <a href="https://twitter.com/farzyness/status/1633223285411389440?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/farzyness/status/1633223285411389440?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a></p> <p><a href="https://t.me/chillhousetech/30?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> This weekend, something to think about, even though I can barely dredge part… https://darkolorin.com/posts/29.html https://darkolorin.com/posts/29.html Sun, 05 Mar 2023 13:42:41 GMT <p>This weekend, something to think about, even though I can barely dredge part of the math apparatus back up in my head already (need to freshen it up just a bit) <br> <a href="https://youtu.be/KYTFsoT2yCo?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/KYTFsoT2yCo</a></p> <p><a href="https://t.me/chillhousetech/29?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And as a follow-up https://darkolorin.com/posts/28.html https://darkolorin.com/posts/28.html Fri, 03 Mar 2023 22:34:06 GMT <img src="https://cdn4.telesco.pe/file/KvLAW5zOW7bK8z89kRvZdVBF9HI2-bRHAkz8pcYfjd37WvQEPO68EzLIxHEyz0pZRpOJd4o9JtR6E0uWmzV60xIAXN6jvNV61Qjd4pbheAwz0rtlc-DmQOUF34WQt7AjSUnJnoi7_R-KYkc036h7_g2aArm4frUbJUqC_pc8CZm4AnqDkJztdEBHEzzzYTXG0j0O49ZCZ2-yDeTROcXyUyTaigX1PaufbkNGjkL4TMRNiCeqqfYKNvcFXuxDUl3ZFnbaEgT5DvnvveKlLTQxlgHFot19mhnnsveXdwSFNjW4q284cHPklzOndvJ35CO3oea67K9gTSGP5DgGOW8Yiw.jpg" alt="" loading="lazy"> <p>And as a follow-up</p> <p><a href="https://t.me/chillhousetech/28?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> How fucking sick I am of all these venture “great minds,” who’ve worked, like… https://darkolorin.com/posts/27.html https://darkolorin.com/posts/27.html Fri, 03 Mar 2023 22:13:09 GMT <p>How fucking sick I am of all these venture “great minds,” who’ve worked, like, everywhere they haven’t worked. Who make up unreal benchmarks just to reject everyone left and right, indiscriminately. <br> <a href="https://twitter.com/kirbyman01/status/1631704147648196608?s=46&amp;t=i_QoJsht4-9pnay0009QEQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/kirbyman01/status/1631704147648196608?s=46&amp;t=i_QoJsht4-9pnay0009QEQ</a> <br> Some dude there writes that DAU/MAU being just OK is 1/4. Shock. And of course he cites Facebook and then Snap as examples. Where the main use case is, you guessed it, messaging someone close. As if this “great” strategy wouldn’t have made it impossible to invest in, say, <br> Discord <br> <a href="https://www.businessofapps.com/data/discord-statistics/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.businessofapps.com/data/discord-statistics/</a> <br> Reddit <br> <a href="https://www.businessofapps.com/data/reddit-statistics/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.businessofapps.com/data/reddit-statistics/</a> <br> Man, what a bullshit bingo, seriously.</p> <p><a href="https://t.me/chillhousetech/27?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> By the way, I’d been thinking for a long time about doing reviews of those… https://darkolorin.com/posts/26.html https://darkolorin.com/posts/26.html Thu, 02 Mar 2023 14:21:16 GMT <p>By the way, I’d been thinking for a long time about doing reviews of those commenters who are pretty detached from the landscape. Because to write stuff like that you need a bit more than just reading a piece of news and speed-running a breakdown. You need to be at least kind of, and preferably a lot, in the know.</p> <p><a href="https://t.me/chillhousetech/26?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Making comments on comments on news? https://darkolorin.com/posts/25.html https://darkolorin.com/posts/25.html Thu, 02 Mar 2023 14:21:14 GMT <p>Making comments on comments on news?</p> <ul> <li>Sure</li> <li>Nooo, better just what you think yourself</li> <li>Yeah and let’s also make a suggestion box too</li> </ul> <p><a href="https://t.me/chillhousetech/25?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> If you don’t get why the giants are chasing TikTok so hard: since early 2023… https://darkolorin.com/posts/24.html https://darkolorin.com/posts/24.html Thu, 02 Mar 2023 13:01:04 GMT <p>Forwarded from <a href="https://t.me/addmeto/5297?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">addmeto ( Grigory Bakunov )</a>.</p> <p>If you don’t get why the giants are chasing TikTok so hard: since early 2023 TikTok has made $200+ million more on in-app purchases than Facebook, Instagram, Snapchat, and Twitter combined.</p> <p>Now they’re going to take TikTok out however they can, first in the States, and then in Europe. And mostly with non-market methods, because by now it seems unrealistic to put out a phenomenon like that with market ones.</p> <p><a href="https://www.forbes.com/sites/johnkoetsier/2023/03/01/tiktok-earned-205-million-more-than-facebook-twitter-snap-and-instagram-combined-on-in-app-purchases-in-2023/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.forbes.com/sites/johnkoetsier/2023/03/01/tiktok-earned-205-million-more-than-facebook-twitter-snap-and-instagram-combined-on-in-app-purchases-in-2023/</a></p> <p><a href="https://t.me/chillhousetech/24?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> interesting, so where the hell are these guys’ in-app purchases in the revenue… https://darkolorin.com/posts/23.html https://darkolorin.com/posts/23.html Thu, 02 Mar 2023 13:01:03 GMT <p>interesting, so where the hell are these guys’ in-app purchases in the revenue line? somewhere up their ass, I guess, analytics for the sake of analytics</p> <p><a href="https://t.me/chillhousetech/23?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> And as an example, one interesting P&L from early Google https://darkolorin.com/posts/22.html https://darkolorin.com/posts/22.html Wed, 01 Mar 2023 18:03:28 GMT <img src="https://cdn4.telesco.pe/file/tGSG3k98ngucralKuFhPm9qJCu2X4N3rzKJOZA6_d-29J-C3MVxqNrpVSVNbTDQuu4nQ4mcOoN7eGG3uiLqIyb1gQSsAggySych9jnnuZylWutpunRFhuD_odBsHx0O1PBIxBPz1qHBq2ke8c1fO4EMaytXu1tn-iqQ0VWT5kUIrb8ctGeqTe_h1AMZWRga1Vdyb_-5_ZKpH0-FrBnpB75EqTuQQ_A7wZDSKf1FKT10bAkOJk9Xh3Qf5lA479cDzQAA2Lk1QBrNBzvNMnv7GMG627jYT3rMLO94Swo4ZUTNGT5AUTU-Jiv1VlaPM99NzYmM5VNCqYEKEfaNoJHRHLw.jpg" alt="" loading="lazy"> <p>And as an example, one interesting P&amp;L from early Google</p> <p><a href="https://t.me/chillhousetech/22?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> About hype cycles in venture https://darkolorin.com/posts/21.html https://darkolorin.com/posts/21.html Wed, 01 Mar 2023 18:02:08 GMT <p>About hype cycles in venture <br> <a href="https://apexventurestudio.substack.com/p/hype-cycles?utm_campaign=auto_share" rel="noopener">https://apexventurestudio.substack.com/p/hype-cycles?utm_campaign=auto_share</a></p> <p><a href="https://t.me/chillhousetech/21?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Thanks to my colleague Phil for sharing this story. https://darkolorin.com/posts/20.html https://darkolorin.com/posts/20.html Sat, 25 Feb 2023 12:50:46 GMT <p>Thanks to my colleague Phil for sharing this story. I recommend everyone read it. How the open source world lives. <br> <a href="https://github.com/zloirock/core-js/blob/master/docs/2023-02-14-so-whats-next.md?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://github.com/zloirock/core-js/blob/master/docs/2023-02-14-so-whats-next.md</a></p> <p><a href="https://t.me/chillhousetech/20?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Typical Valley. https://darkolorin.com/posts/19.html https://darkolorin.com/posts/19.html Thu, 23 Feb 2023 16:22:30 GMT <p>Typical Valley. And we’re still surprised about layoffs. <br> <a href="https://twitter.com/dantley/status/1628604242188713984?s=46&amp;t=IZFjzBvJYxlcQ3Sjkfz5HQ&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/dantley/status/1628604242188713984?s=46&amp;t=IZFjzBvJYxlcQ3Sjkfz5HQ</a></p> <p><a href="https://t.me/chillhousetech/19?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Came across a post on FB about launching on ProductHunt. https://darkolorin.com/posts/18.html https://darkolorin.com/posts/18.html Tue, 21 Feb 2023 11:25:02 GMT <p>Came across a post on FB about launching on ProductHunt. Work’s been done: hundreds of hours, materials, mailouts and other stuff. Dozens of guides read. Something like +3k signups there.</p> <p>I’m obviously not a growth hacker and I’m not gonna give lectures and webinars on fundraising. But ProductHunt is a wildly унылый source of traffic and even leads these days. The platform is packed with Russian growth hackers and other kinds of boosting/fake crap. None of my friends in the US at an early stage even consider it something worth paying attention to.</p> <p>Just throw it out of your pipeline, or in 40 minutes upload some screenshots and a description and you’re good.</p> <p><a href="https://t.me/chillhousetech/18?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> One of the important things in consumer social. https://darkolorin.com/posts/17.html https://darkolorin.com/posts/17.html Mon, 20 Feb 2023 10:14:58 GMT <p>One of the important things in consumer social. Products basically don’t verticalize. This is about that exact fact that making TikTok for “food, water, dogs, etc.” just makes no sense. There are rare exceptions like gaming and more “adult” apps. But mostly anyway it all happens off the content format.</p> <p>There are a bunch of reasons, I’ll try to just briefly point out the basic ones:<br> - consumer social is mechanics, certain sequences of actions with a strong utility function, so by just repeating it you won’t hook new people<br> - knowing point one, it’s easy to get that if you have utility, to build mass in a separate vertical you need to make the utility many times better, otherwise the new vertical will most likely just show up inside the original product<br> - often there’s no answer on distribution strategy, it’s usually there in the original (timing, luck, randomness plus solid execution and you only do what works)<br> - the strength of social ties and, accordingly, social capital, which over time makes switching from one app to another very painful<br> - and there’s very little time in a day for lots of videos :)</p> <p><a href="https://t.me/chillhousetech/17?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> For the weekend, to chill — to think https://darkolorin.com/posts/16.html https://darkolorin.com/posts/16.html Sat, 18 Feb 2023 15:29:07 GMT <p>For the weekend, to chill — to think <br> <a href="https://youtu.be/fQ4rc7npiXQ?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://youtu.be/fQ4rc7npiXQ</a></p> <p><a href="https://t.me/chillhousetech/16?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> So, Durov is posting there about the number of messages, but nobody’s posting… https://darkolorin.com/posts/15.html https://darkolorin.com/posts/15.html Sat, 18 Feb 2023 14:30:45 GMT <img src="https://cdn4.telesco.pe/file/GfjRDnX0gNR5_jfF3XhQGiaBnJGnYQp3htVlrktbl2A7SOLzsxa0JSI_jjtlnVIhIe18a5JkXNUZkm9GNXoVjhdELzyiVCeyZPwmVcuUSgFR7a-tx0gMzs2Q1-_v74fN9IG0pKWc8h5FFfuIXrLGOuWih68Jms-5jELykm3NMHyi225KRjozOSFvXqhoeiYpjITERE5OZx-ilPvqa-ZqFCcMUzpCz6dtZkLON4KntA5jIXJqAf_aWvFO5lmszZgApTZIPiUMywZUoYgN1IymMlkEfJi1VDDUh_lqi63m9HflpMqO7ysDx4uyp6T3I8mM4RkP4lq8m8e1O-ozO5cI8A.jpg" alt="" loading="lazy"> <p>So, Durov is posting there about the number of messages, but nobody’s posting that stat and there are no absolute numbers. And the number of messages is an extremely weird thing anyway, because you need to know how many per user, the number of bots, and so on. <br> Plus it’s aggregator data. In short, some kind of empty PR thing.</p> <p><a href="https://t.me/chillhousetech/15?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> The paradox of modern products. https://darkolorin.com/posts/14.html https://darkolorin.com/posts/14.html Sat, 18 Feb 2023 11:41:55 GMT <p>The paradox of modern products. Retention doesn’t matter without distribution, and distribution doesn’t matter without retention. <br> I think back in the 2000s nobody even particularly cared about this. It’s kind of wild how fast the approach to evaluating products changes. Feels like there’s big potential in this area for new projects. <br> P.S. And then on top of that they roll out “stacked cohorts” for you.</p> <p><a href="https://t.me/chillhousetech/14?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Really liked this thread about hiring at RIOT https://darkolorin.com/posts/13.html https://darkolorin.com/posts/13.html Fri, 17 Feb 2023 12:58:41 GMT <p>Really liked this thread about hiring at RIOT <br> <a href="https://twitter.com/justmazer/status/1626396319576391681?s=46&amp;t=pddonjs_EEVpGjYN09JKhA&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://twitter.com/justmazer/status/1626396319576391681?s=46&amp;t=pddonjs_EEVpGjYN09JKhA</a></p> <p><a href="https://t.me/chillhousetech/13?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Subscriber churn in the channel is 1%. https://darkolorin.com/posts/12.html https://darkolorin.com/posts/12.html Thu, 16 Feb 2023 22:11:18 GMT <p>Subscriber churn in the channel is 1%. Fine.</p> <p><a href="https://t.me/chillhousetech/12?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A pretty interesting framework for splitting up players’ motivation in games. https://darkolorin.com/posts/11.html https://darkolorin.com/posts/11.html Thu, 16 Feb 2023 02:21:44 GMT <img src="https://cdn4.telesco.pe/file/Mxj0aCXMdpL77CYnbvQV8DIuTspus8vLCJcrgv4hWsCb-Zz08K7Hia-dlXnU_bhaDbKZQsFbSxovqVqTmEk5FhL-de15vwy2yc4VZrRwraEb2cN-HYru9WhsrXAFQnmoHi76-bVY3VsKhyp5G2MAByw498txPcBbKb1-n89d13Wq_Ffhq7gIjWX7yBXwPrNlAzU7mXEMvNbH44sjuO1I_UXHNbk6JVoCNL9HWEKGsmg5XahCSpFRBFTXCngJsgn_ni5r3tiwVPPyVVQz_Tct0SA4QQgh2eF91orJMx1Z5V-O4ZtCjER1JnlwMJymZWDJ09DOqTDLlRlOVMJLFCPQjQ.jpg" alt="" loading="lazy"> <p>A pretty interesting framework for splitting up players’ motivation in games. Used at Riot &amp; Blizzard.</p> <p><a href="https://t.me/chillhousetech/11?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> A small case about networking. https://darkolorin.com/posts/10.html https://darkolorin.com/posts/10.html Tue, 14 Feb 2023 14:44:05 GMT <img src="https://cdn4.telesco.pe/file/C3VuaXWUrxLhW9VPzhmScdRKKNerhFo06cfFsoMOQ4Pxu5DlLAU9wYb_qPvUJGnd_Xxao6TaOCEuJ9NxXSC2rxcsGpFpWKe81eMtmvviHfPbcabGkfp4dbC8T6wliETk5O0YNvYhpJ-I3C5k6VU70h_1TfS-iBmHRC4QzTztRBGRd71PBNQpfCMLL9LAyYqC__Z6gSYwNyWJkYJv1meKuIStwWM5iwGceJ7x5k3g_vRcFSVjQyaY4ZNhl82PBHjCxcpORj14TBKjh7FLkdyU0qYqunlTK4SCSVPY-4HrqthFtJ2hOiHMltc6k-WejmmNDMQlmRgwfUODr3DuLsNdrQ.jpg" alt="" loading="lazy"> <p>A small case about networking. Everyone talks a ton about it, but nobody knows jack shit about anyone and can’t make an intro. I’ll give you a simple example. <br> Run a Twitter, just don’t do that bullshit bingo content like “10 more ways to live better/product better/launch worse and so on.” Drop short but leading comments. And voila, it works. <br> I left a comment like that under the CEO of Neeva (like, subscription search and the dude himself is a super experienced guy from Google) and he replied, then I added him on LinkedIn, described my experience and said I’d like to send him some proposals by email. And he gave it. <br> P.S. Haven’t figured out yet how to use the guy and haven’t even had time to write, but I think I’ll write this weekend. I actually have ideas on search distribution.</p> <p><a href="https://t.me/chillhousetech/10?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> All I want to say about chatGPT and the others for search: - it’s a massive… https://darkolorin.com/posts/9.html https://darkolorin.com/posts/9.html Tue, 14 Feb 2023 11:05:50 GMT <p>All I want to say about chatGPT and the others for search:<br> - it’s a massive shift in the algorithm and now it’s not clear what even counts as SEO, and how to get into the results<br> - most likely it’s a big change in the ads market, now you need to get inside the bot’s results, not the old SERP<br> - it runs slow and it’s a very expensive thing resource-wise, hell knows how they’re planning to scale this<br> - it’s probably a shift from an artificial “search query language” (though there are a lot of questions here, maybe not a fact) to a natural one<br> - and finally I still think text input and output is outdated bullshit, we need to get to input without text and results that are purely visual answers (generate it or search it)</p> <p><a href="https://t.me/chillhousetech/9?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Channel photo updated https://darkolorin.com/posts/8.html https://darkolorin.com/posts/8.html Mon, 13 Feb 2023 16:36:04 GMT <img src="https://cdn4.telesco.pe/file/X9dzcaSagFaUVJOvS3hcKXq361ciqUe0I7mWslZKj8HHzyrldMxcY8IIv6npcdoWjSa_JDxqUSc50OIV0i95G_3qIBKdKbn9gcArmsn4sYtsqQPkpHfAcyF9vTk2AGXN94-lPD3z6XPcm56uRliRRTYq5dZqu8vLtf6Le1wZd2YxyYoUqkdZ7LrXODBRdIp4LZ0kkmXQW56GRGjnf3SF4ttZB2tXzYUH6r6Zyw7oeQ51EGqbQSF8tfKZBVLxt7x4ldlIRvPy2zu_F8h94MejO5k31qjvpX_4W0pdnOgdhtoSMLzAuXDfhG16WnYYbdy3gdbChPY1-SsHV-Hzl57L8w.jpg" alt="" loading="lazy"> <p>Channel photo updated</p> <p><a href="https://t.me/chillhousetech/8?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> App-fiction. https://darkolorin.com/posts/7.html https://darkolorin.com/posts/7.html Mon, 13 Feb 2023 16:35:24 GMT <img src="https://cdn4.telesco.pe/file/KHztmdEEpHGnAOVr-MxfjJQ_g__Lx8tVll7x1EyBa5rFzYH_GSjaVb4v_xAMcxGNbc8EcQ_70JPQH61wJzLaziTe59x6dRdcOhbAaxjrWpJT-oZX75J413WytjuJpHC7pVflVHg348of320PeF9USXyjhAm0ykoSdCrkmnbNYx4H5_gAiWeO5OGMXdJBEtJOWryR5f9SAZ_bgEZgjWuXFEmVlIoxtBLVt44mZZY0q0Va1lHdPD68A2ituVkUt4NMckjgUbphEN7WnVyhqZHTU41oissgOYbCbmHQA0VQQOJDoCh3YclwgXJklEDf3fojAy07dFwfjRuuhDkIxnVAuw.jpg" alt="" loading="lazy"> <p>App-fiction. <br> What’s interesting. We’ve kinda arrived at a time where consumer apps are like TV shows. You built it, found the top content (a mechanic), people played around, and that’s it — the app died. The product lasts, like, 2–3 months up to a year. After that, making “seasons” gets hard and complicated, and everything starts sliding downhill. All these show-products are sponsored by venture (pretty much like in the film industry), and then the hope is there’ll be box office (only here there probably won’t be any). <br> Here’s the chart for the everyone-knows-it BeReal</p> <p><a href="https://t.me/chillhousetech/7?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Since it’s Sunday, there’s time to do some graphomania. https://darkolorin.com/posts/6.html https://darkolorin.com/posts/6.html Sun, 12 Feb 2023 14:20:41 GMT <p>Since it’s Sunday, there’s time to do some graphomania. A bit of analytics on consumer social. <br> One of the most important metrics is the so-called hourly pattern. Basically any b2c has a time when the audience is most active. And it’s important to keep an eye on which hours which cohorts show up. I’d even say DAU isn’t that important — what matters is this условный “online.” We constantly track the hourly online and try to grow it. <br> Usually it’s in GA and other analytics (you can do it off logs).</p> <p><a href="https://t.me/chillhousetech/6?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Post #5 https://darkolorin.com/posts/5.html https://darkolorin.com/posts/5.html Sun, 12 Feb 2023 12:20:40 GMT <p>Links from the original post:</p> <ul> <li><a href="https://t.me/chillhousetech?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://t.me/chillhousetech</a></li> </ul> <p>A message was pinned on Telegram.</p> <p><a href="https://t.me/chillhousetech/5?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Feature request for all apps with followers, it’d make onboarding via word of… https://darkolorin.com/posts/4.html https://darkolorin.com/posts/4.html Sun, 12 Feb 2023 12:11:53 GMT <img src="https://cdn4.telesco.pe/file/Nkubun524dt6VpCi_9DN2aUTl-uQvtL7y3fbBW5XgsG0TG5FrMYsKXaMKenPcoXtGMd7htZDSfBhOXlyGNTiIUB_oiwmJ0vASja-yRUmrJthyT4v0_dRPqpwWz-1mIe7fKdq8Dw-5wvLpRdf_O0uizRlqm4ECiVGW1J2JjwAtriRJZLtR6ppsBC7KD6KGnSyfd50XwOR9BAq8iDySSpqc64H_b5Y1ScisF_D-ybax6F2vFFUR6dn-BFZLkd3v-QZNWI2Z_btB4O0PHFUEra6nkMoy6lEOcTj1vpfjFsbfXbGwaj14TIW9Cf2fh_C7aNIK66vtzaVkhPaXwNDHwumBQ.jpg" alt="" loading="lazy"> <p>Feature request for all apps with followers, it’d make onboarding via word of mouth easier. Sharing the list of who you yourself follow, that’s a killer thing. Basically personalized onboarding from a friend. You can spice it up with customization and picking interests.</p> <p><a href="https://t.me/chillhousetech/4?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Lately I’ve been watching how the big founders of consumer social giants… https://darkolorin.com/posts/3.html https://darkolorin.com/posts/3.html Sun, 12 Feb 2023 11:54:58 GMT <img src="https://cdn4.telesco.pe/file/S6q3CzvO1mANq2fKMA2Z2OM109rSFyB7QjC4uu-39t0FZX1CSpxa9y3eDYALtU9pc5zYd_dRgSwsN8wJkgDGNEOlgCTr5f2o3cwkF3CmIVDwi8o7rUpgIOnou5i7iMQC-TeEwuzCyuBAVbNhAFC34pm2vU8gfQLGdhrbEhW2rEjkpu-gtcUGUteWXl-gmArlA3gE22shxKoi2R97K4dUZNS7wnGANMIOg3fIq8-3uVpn8Pgvs2lQ-VnoJNKJ1N_0t-2T9VRrdJegEGfZMdmPBVJYyckR4TNyzFVqHZIA0qpzauTow6BrzpaZORXOZpXKivHQh7MkxgW2Hu44dLxdpg.jpg" alt="" loading="lazy"> <p>Lately I’ve been watching how the big founders of consumer social giants started doing their new projects. Whether it’s Kevin with his Artifact, or the guys from Vine with some SUP xyz bullshit during the NFT boom. <br> In short. There’s this persistent feeling that consumer social is like the news: you’re either in it, or you already don’t get it. The market shifts way too fast. And you can’t build new products off what used to work. I’ve seen 100M users myself. <br> And one more thing. I think it’s even more important. Basically everything runs on the founders’ motivation, and I don’t really get what it is for people who spent 5–6 years sitting in a $10M house and, overall, being bored (that’s often exactly how it is, I remember a conversation at home with the Tinder founder Sean Rad, literally said there’s fuck-all to do). Attaching a pic of the little house in Bellair from, like, visiting, so to speak.</p> <p><a href="https://t.me/chillhousetech/3?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> My name’s Alexey and I made Prisma, Capture (not as successfully, but I did… https://darkolorin.com/posts/2.html https://darkolorin.com/posts/2.html Sun, 12 Feb 2023 11:45:07 GMT <p>My name’s Alexey and I made <br> Prisma, (<a href="https://en.wikipedia.org/wiki/Prisma_(app)?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://en.wikipedia.org/wiki/Prisma_(app)</a>) <br> Capture (not as successfully, but I did raise from top-tier funds in the Valley), <br> <a href="http://lf.group/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">http://lf.group/</a> <br> (acquired by Xsolla). 2 exits, 3 startups. <br> I’ll be writing all sorts of random stuff. Just thoughts, because sometimes it hits. I can change my mind over time, admit mistakes. Welcome. <br> My links: <br> <a href="https://www.linkedin.com/in/alexeymoiseenkov?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.linkedin.com/in/alexeymoiseenkov</a> <br> <a href="https://www.facebook.com/darkolorin?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://www.facebook.com/darkolorin</a> <br> <a href="https://x.com/darkolorin?s=21&amp;utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://x.com/darkolorin?s=21</a> <br> Right now I’m working on <br> <a href="https://trymirai.com/?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site" rel="noopener">https://trymirai.com/</a></p> <p><a href="https://t.me/chillhousetech/2?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p> Post #1 https://darkolorin.com/posts/1.html https://darkolorin.com/posts/1.html Sun, 12 Feb 2023 11:43:08 GMT <p>Channel created on Telegram.</p> <p><a href="https://t.me/chillhousetech/1?utm_source=darkolorin.com&amp;utm_medium=referral&amp;utm_campaign=personal_site">Original on Telegram</a></p>