Studying collective intelligence.

Welcome to World Models. This website tracks and publicizes my internal thinking over time. The hope is that increased legibility improves how useful LLMs are to me and stimulates the thinking of any readers.

I am cofounder and CTO at Precursor Labs, where we study how the world is breaking with rapid AI progress and build solutions to put it back together more effectively.

The title, “World Models”, references the phenomenon of an agent (me) in an environment (the world) continuously updating its priors about how the world works (model) to be able to take actions that maximize terminal reward. This involves curiosity, research, critical thinking, writing publicly, and shipping experiments and products, among other things.

If you want to digest this easily, just give this to your agent: https://github.com/handsdiff/notes.
Feel free to message me @handsdiff on Telegram or X.