- Anthropic mathematician Levent Alpöge says he used Claude to help produce a Lean-verified proof of the Generalized Riemann Hypothesis, posting the paper and code on his own site.
- Hours later, Stanford mathematician Alessandro Masullo published an independent, entirely human proof attempt addressing the same problem.
- Alpöge said he messaged Anthropic CEO Dario Amodei last Sunday night to ask how to announce the result, days after Anthropic’s stock began trading following its IPO.
- The Riemann hypothesis, open since 1859, is one of the Clay Mathematics Institute’s Millennium Prize Problems and underlies hundreds of dependent results in number theory.
What Happened
Levent Alpöge, a mathematician at Anthropic, said he worked with Claude to produce a proof of the Generalized Riemann Hypothesis, publishing the paper and Lean-verification code on his personal website, according to The Decoder (Nov 21, 2026). Hours later, Stanford’s Alessandro Masullo released a competing, entirely human derivation of the same result.
Why It Matters
The claim lands days after Anthropic’s Class A stock began trading on Nasdaq following its IPO, and after OpenAI’s own attempted Millennium Prize proof collapsed into a dispute over credit. It also follows GPT-5.5’s verified Lean proof of a new Ramsey number result, part of a run of AI-assisted attempts at unsolved mathematics problems.
Technical Details
The Riemann hypothesis, dating to 1859, conjectures a hidden structure underlying the distribution of prime numbers and is one of the Clay Mathematics Institute’s Millennium Prize Problems. Alpöge submitted his proof in the Lean programming language, allowing computer verification of each logical step rather than relying on human peer review alone. Alpöge wrote in his announcement essay that “the precise timing is horrible,” referring to the proof landing the same week as Anthropic’s public listing.
Who’s Affected
Mathematicians working in analytic number theory now face two independent claims to verify instead of one, since both Masullo’s human draft and Alpöge’s Claude-assisted proof require line-by-line review before either is accepted. Anthropic’s research credibility is also tied to the outcome, since the proof is attributed to one of its own researchers using its flagship model.
What’s Next
Formal verification of Alpöge’s Lean code and peer review of Masullo’s human draft will determine whether either proof holds, and number theorists are already working through both claims in public. A confirmed proof of the Generalized Riemann Hypothesis would validate hundreds of dependent theorems across mathematics.