Anthropic mathematician claims Riemann Hypothesis proof
Original: AI claimed a solution to the 167-year-old Riemann hypothesis — and a human did, too
Why This Matters
Proving the Riemann Hypothesis would reshape number theory and cryptography foundations.
Levent Alpöge, an Anthropic employee, claims a Lean-verified proof of the 167-year-old Riemann Hypothesis. Compiled successfully by MIT's Andrew Sutherland, the result is under community review.
Levent Alpöge, a mathematician employed by Anthropic, announced this week that he has produced a Lean-compilable proof of the Riemann Hypothesis — one of mathematics' seven Millennium Prize Problems, unsolved since 1859. The hypothesis concerns a hidden structure governing the distribution of prime numbers.
Alpöge messaged Anthropic CEO Dario Amodei on a Sunday night to ask how to announce the finding, describing the timing as 'horrible' given that Anthropic simultaneously became the first publicly traded frontier AI lab. He released the proof without a full human-readable explanation, saying he 'could not live' waiting longer.
MIT mathematician Andrew Sutherland and others on the Zulip mathematics discussion platform confirmed the Lean code compiles successfully — a meaningful signal, since Lean is a formal verification language that flags logical errors mechanically. Full mathematical community consensus will take considerably longer. Number theorist Chris Birkbeck noted the proof appears to draw on automorphy-lifting arguments, Galois representations, and results by Deligne–Ribet, among other heavy machinery.
Alpöge also credited Anthropic's Claude extensively in his writeup, describing the AI model's role in working through the argument. He and NYU's Tristan Buckmaster separately released a proof related to the Navier-Stokes equations.