Anthropic mathematician Levent Alpöge claims Riemann hypothesis proof made with Claude
Archived — this story has rotated out of today’s deck. It is kept here in full.
The gist
Anthropic mathematician Levent Alpöge posted a four-page note claiming he and Claude proved the Riemann hypothesis.
Mathematicians are still checking the Lean code, so the 1859 problem stays officially open.
Background
Levent Alpöge is a 34-year-old Turkish-American mathematician and member of technical staff at Anthropic, the AI company behind the Claude model; he won the Morgan Prize in 2015 and has worked on Hilbert's tenth problem and the Jacobian conjecture. Since joining Anthropic in 2024 he has become a public face of AI-assisted math, posting high-profile announcements often directly to X outside traditional peer review. The Riemann hypothesis, proposed by Bernhard Riemann in 1859, holds that the locations of the prime numbers follow an intricate order and is the most famous open problem in mathematics. The claim drew attention partly because the first crack in the problem appeared in October, when one proof in OpenAI's large release of math results made progress on the primes.
How it unfolded
- 2010Named a finalist in the Intel Science Talent Search for a computer-science project analyzing MRI scans of blood vessels; an asteroid, 25898 Alpöge, was later named after him.
- 2014Graduated summa cum laude from Harvard with a bachelor's in mathematics and a master's in physics.
- 2015Studied at Cambridge as a Churchill Scholar and was awarded the Morgan Prize.
- 2020Earned a Ph.D. in mathematics from Princeton, supervised by Fields Medalist Manjul Bhargava.
- 2022With Bhargava and Ari Shnidman, established the first proven upper bounds on the proportion of integers expressible as the sum of two rational cubes.
- Jan 2024Began as a member of technical staff at Anthropic.
- 2025With Bhargava, Wei Ho and Shnidman, published a second proof in Inventiones Mathematicae that Hilbert's tenth problem has a negative answer over the integers of every algebraic number field.
- Jul 19, 2026Announced an explicit counterexample to the 87-year-old Jacobian conjecture, found using Anthropic's Claude Fable 5 model; other mathematicians checked the arithmetic within days.
- Aug 12, 2026Alongside Philippe Voinov and Saul Reynolds-Haertle, used Claude to construct Hadamard matrices.
- Aug 23, 2026Posted a 100-plus-page document with a purported proof that the 6D sphere has a complex structure, addressing the 1947 Hopf problem; the result remains unverified.
- Nov 20, 2026Posted a four-page note on his website claiming he had discovered a proof of the Riemann hypothesis and the broader Generalized Riemann Hypothesis with Claude, linking to computer code rather than a traditional paper.
Key terms
- Riemann hypothesis
- A conjecture that all nontrivial zeros of the Riemann zeta function have real part one-half.
- Lean code
- Formal proofs written in Lean, a proof assistant and functional programming language.
- Morgan Prize
- An annual award for outstanding undergraduate mathematics research by students in the US, Canada, or Mexico, founded in 1995.
Source: Wikipedia
Who’s saying what
- Claimant
- Alpöge says he believes he and Claude discovered a proof of the Riemann hypothesis, and that what he saw during the work amazed him as 'one dawn across science of something far greater.'
- Expert
- Cornell mathematics professor Steven Strogatz called the result 'the Mount Olympus' and said Alpöge seemed to play a central role in directing Claude, likening him to 'a very skillful jockey riding a really fast horse.'
- Caution
- Independent mathematicians are still verifying the large archive of formal Lean 4 verification code, and experts have yet to confirm the hypothesis is properly coded and the proof correct.
Still unverified
Alpöge's Riemann claim is unconfirmed: he posted no traditional paper, only a four-page note linking to Lean code, and experts have yet to verify the coding and proof; commenters say the proof is reportedly 250,000 pages and that no one else appears to have finished compiling it. His August 2026 Hopf problem result also remains unverified.