principia labs
Most scientific breakthroughs begin as math problems. A new drug, a faster algorithm, a better material— somewhere at the root is a piece of mathematics someone had to work out by hand. That's slow. There aren't enough mathematicians, and the problems are getting harder.
We train our own models, from scratch, to do original mathematical and scientific reasoning.
We're a small research lab in San Francisco, backed by Modern Capital, Pebblebed, Align, and others. Our researchers have contributed to DeepSeek V3, FrontierMath, Kimi 1.5, and Tinker.
Follow us on twitter, or find upcoming talks and reading groups on our events page. If you want to work on this, see our open roles.