Science News reports that AI is accelerating mathematical formalization — the process of translating proofs into computer-verifiable code — which was previously a niche effort practiced by only a handful of mathematicians. Kevin Buzzard at Imperial College is using this approach on Fermat's last theorem not to re-prove it (a proof exists from 1998) but to demonstrate that computers can eventually help mathematicians find and solve a much wider range of problems. The key insight: the computer doesn't fill in gaps — the mathematician writing the proof must be far more rigorous than usual, making formalization a demanding but ultimately more verifiable approach to mathematics. This shift could reshape how mathematical knowledge is organized and verified once a digital library of formalized proofs exists.