Column “read, but not fully understood”

Column “read, but not fully understood”

1/ really like the idea of
model merging (https://crisostomi.github.io/blog/2025/model_merging/)
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)

2/
no new ideas, there are new datasets, (https://blog.jxmo.io/p/there-are-no-new-ideas-in-ai-only)
and right away some architecture rolls up,
mamba3 (https://openreview.net/pdf?id=HwCvaJOiCj)
(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

3/ from philosophy trying to finish off Marcus Aurelius “Meditations” (“alone with yourself”)

4/ about causality, statistics, connections between phenomena, basically
an excellent chapter (http://web.cs.ucla.edu/~kaoru/BIB5/ch2.pdf)
from the book The Book of Why

5/ sometimes it’s interesting to read
what respectable analysts wrote a year ago (https://stratechery.com/2024/enterprise-philosophy-and-the-first-wave-of-ai/)
about AI, helps you read and listen soberly to statements by big managers

6/
about how they made the legendary (https://share.google/pexNwvYmAtedMP2fX)
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.