Interpretability is the most important scientific problem in the world.

At Tilde, we want to do moonshot, applied interpretability research to make maximally safe and performant AI. We envision a future where trillion parameter models can be ~ approximated by humans with surgical precision.