Product manager, technologist, and builder. I write about product management, AI, books, and other fun things!
I trained Echo, a 1.38B-parameter GPT, from scratch on 8 H100s using nanochat and Claude Code. She beats 2019's GPT-2 on CORE for $69, after $104 of failures.
OpenAI dumped solutions to hundreds of open maths problems and mathematicians aren't thrilled. Ted Chiang saw elements of this in 2000, the same feelings are coming for many of us, the future is going to get weird
Sebastian Mallaby's biography of Demis Hassabis and DeepMind: a book of the year contender, and the clearest window yet into how the people building AGI think about it.
AI tools for learning have never been better, but the temptation to skip the learning has never been stronger. On gyms, construction sites, and what the evidence says about working your mind.