Source: Anthropic Research
Anthropic uses Claude to write formal proofs for 100+ open-source libraries
Why it matters
With Theorem, Anthropic's Frontier Red Team had Claude prove security properties at the RISC-V assembly level in over 100 real libraries, finding long-missed bugs along the way. Formal proofs used to take years of expert time; if you maintain security-critical code, model-written proofs are becoming a realistic step after tests, though the post lists limitations worth reading first.