Anthropic mathematician claims AI-aided Riemann hypothesis proof
Ongoing: AI Hot
Levent Alpöge says he found a solution to the 167-year-old Riemann hypothesis that a computer proof checker called Lean accepts, with help from Anthropic's Claude. Mathematicians in a chat group said the code compiles, but nobody has shown the proof is correct.
Alpöge, a mathematician at Anthropic, announced the claim in an essay. Andrew Sutherland, a mathematician at MIT, said he confirmed that the Lean solution compiles. Other participants in a Zulip chat said the same, and most appeared confident that the result will hold.
A separate proof appeared at the same time. Ravi Vakil, a Stanford professor and president of the American Mathematics Association, posted a proof of the General Riemann hypothesis. Stanford postdoc Alessandro Masullo wrote it, and Vakil said it is all human work with no AI.
Masullo said his proof took 12 years. Whether either proof is correct is not yet established.
Key facts
- Who
- Levent Alpöge; Dario Amodei; Chris Birkbeck (number theorist at the University of East Anglia); Andrew Sutherland (MIT mathematician who said he confirmed the Lean solution compiles)
- What
- Alpöge posted a four-page paper on his personal website claiming an AI-assisted proof, supported by 17 million lines of code that formally verify a broader version of the conjecture. He has not released the AI-generated explanation, which he says runs over 250,000 pages, and mathematicians expect it could take years for humans to digest; a separate draft proof developed without AI was announced…
- When
- 20 November 2026
- Status
- Reported