Anthropic: Claude Produces 13M-Line Proof of Fermat’s Last Theorem

This milestone shifts the focus from AI assisting human mathematicians to AI independently producing verifiable proofs at a scale no human could manually check. The 13-million-line output suggests formal verification is becoming a practical tool for tackling problems where human-written proofs are impractical, which could accelerate progress in fields relying on rigorous logic. The open question is whether such machine-generated proofs will be accepted as mathematical knowledge or remain confined to computer-checkable systems, and how this capability scales beyond isolated problems to broader research agendas. It also raises the stakes for AI reliability, since the proof's validity rests entirely on the verification system's integrity.