Teaching AI to do science and engineering.

Science doesn't move at the speed of computation. It moves at the speed of insight. A handful of researchers in each generation see what others miss — they ask the question no one thought to ask, design the experiment no one thought to run, connect findings across fields that don't talk to each other. The rest of progress follows from that.

If AI can learn to do that kind of work — not just crunch data, but reason, hypothesize, and iterate the way a great researcher does — the timeline of human progress changes. Decades of work become years. Problems we'd written off as generational become tractable.

We recruit exceptional scientists and engineers — researchers at the top of their fields, people who've made discoveries and built things that matter — and pair them with frontier AI. They teach the system how they actually work: how they form intuitions, choose between approaches, recognize when something's wrong, and know when to push harder.

This is the tacit craft that doesn't live in textbooks or papers. Capturing it is the work.

Faster paths to new medicines. Better materials. Cleaner energy. Real progress on problems that have outpaced the supply of human attention for too long. The upside of getting this right is enormous, and we think it's one of the most important things to be working on now.