Levent Alpöge, a mathematician employed by Anthropic, says he and Claude have found a proof of the Riemann Hypothesis, but the released files have not been publicly validated by independent mathematicians.
In a four-page note dated Nov. 20, Alpöge wrote, “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis.” He said the method extends to the Generalized Riemann Hypothesis and may reach the Extended Riemann Hypothesis.
Alpöge said he and Claude “have also formalized the arguments without assumptions” in Lean, a proof assistant that checks mathematical statements and deduction steps against formal rules. The accompanying material was uploaded in 14 compressed archive parts.
The README for the repository hosting the Lean files said that, as of the Nov. 20 mirror, no independent build or review had been published. Any review would need to determine whether the encoded theorem matches the standard hypothesis and whether the project is complete and free of hidden assumptions.
The 1859 hypothesis concerns where the nontrivial zeros of the Riemann zeta function lie. The Clay Mathematics Institute defines it as the claim that all those zeros have a real part of 1/2 and lists it among the seven unsolved Millennium Prize Problems.
Alpöge described the argument as “incredibly long.” He said a machine-generated written version exceeded 250,000 pages and that he planned to study and rewrite it. He also credited Claude with determining that a chain of theta lifts used for a von Mangoldt sum of length X had a length of √X.
A completed Lean check would establish only that the theorem encoded in the files follows from their formal premises. It would not alone show that the formalization faithfully represents the accepted mathematical problem. Clay’s prize rules require publication, at least two years of review and general acceptance by the mathematical community before it considers a claim; each problem carries a $1 million prize.
Anthropic said in an Aug. 10 research post that an earlier Claude effort had not solved the hypothesis, although it increased the share of zeta-function zeros known to lie on the critical line from 41.6% to 67.2%. “Claude did take a real stab … it didn’t succeed,” the company said. Alpöge’s files now await independent examination under the standards required for a mathematical proof claim.



