AR
What actually changed in the new reasoning models?
The interesting part is not longer chain-of-thought. It is better task decomposition under uncertainty. Here is the benchmark pattern I am seeing.
The professional AI community
Less noise.
More signal.
A private corner of the internet for people who build, study, and seriously think about artificial intelligence.
LIVE CONVERSATIONS
Worth your attention
JM
Show us the agent workflow you cannot live without
Mine: a research agent that makes a short evidence pack before I open a product doc. It has stopped a lot of good-looking bad ideas.
SL
Small models are quietly getting very good
Local inference is now good enough for the boring 80% of internal tasks. The cloud is for the hard part, not every part.
COME IN