Writing

I write occasionally about AI safety, mechanistic interpretability, and reasoning — things I'm actively working on or thinking through.

Most of it is on LessWrong.