There's so much handwaving about how "AI will cure cancer" or "AI will revolution science" What we need are more specific, actionable ideas for how AI could accelerate science & security Dropping next week: @IFP will publish 15 "AI for science" ideas from experts.

we have signed a deal for an additional 4.5 gigawatts of capacity with oracle as part of stargate. easy to throw around numbers, but this is a _gigantic_ infrastructure project. some progress photos from abilene:

every city in history built strong walls before its people could build cathedrals

in 2010 for my undergrad electrical engineering capstone thesis “learning image compression” i used an mlp to do causal predictive on-line image compression, with a primitive convolution. it got a b- and i didn’t touch ai again until 2015. the lesson is always bet on yourself.

“what would you do if you knew you couldn’t fail” — bland, secular, wine-aunt coded “what will you do if you knew only a glorious death earns a seat in odin’s hall” — powerful, infused with divinity, lindy

the most interesting take on gpt psychosis might just be this one from @LapsusLima: part psychedelic experience, part method acting in a theater of hyperstition. this piece takes us beyond the ‘simulators’ frame and into llm-as-dramaturge.

Today, we at @OpenAI achieved a milestone that many considered years away: gold medal-level performance on the 2025 IMO with a general reasoning LLM—under the same time limits as humans, without tools. As remarkable as that sounds, it’s even more significant than the headline 🧵

develop a trusted group of people who are smart enough to argue against you effectively. this is one of the functions of “the boys”.

no fucking way a man named Eli Dourado is breaking this story

might make a few hundred billion dollars though