Pedro Domingos: Masullo lost $1M by having his proof checked by a human instead of an LLM

pmddomingos · x · 2026-11-22

AI researcher Pedro Domingos quips that Masullo's mistake was having his proof checked by a human rather than an LLM — with an LLM review, he would likely have finalized it before Alpöge and won the million-dollar prize. A pointed claim that frontier LLMs are already more reliable than humans at verifying mathematical proofs, sure to spark debate in both math and AI circles.

Related event: AI-Assisted Math Proof Sparks Fight Over Lean Verification and Gatekeeping(7 posts)→

Original post →

More from Fun

Fun channel →