There’s this common take that AI will just make every link in the tech…

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):

• the bottleneck in any economic production process (software) is always time — basically, how long it takes to do something useful
• 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
• 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)
• introduce a sync element — manager/lead
• the number of sync elements grows like crazy — you get processes/responsibility (a “reduction/folding” function)
• and that’s basically our whole industry today

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).

There are tons of examples of this kind of compression:
- the early-20th-century assembly line
- industrialization (people leaving the fields)
- etc

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.