Money and business in the Middle East.

AI

Anthropic mathematician says Claude helped prove the Riemann hypothesis

A proof of the Riemann hypothesis, unsolved since 1859, has been posted by Anthropic mathematician Levent Alpöge, who says he reached it with help from the company’s Claude model. He published the argument Friday in Lean, a language that lets computers check proofs line by line.

· Originally published by ontime+ · Last verified: 21 Nov 2026

Key Points

  1. Anthropic's Levent Alpöge says he found a proof of the Riemann hypothesis using Claude.
  2. The 1859 conjecture resisted mathematicians for 167 years and is a Millennium Prize Problem.
  3. If verified, it would be the second such problem solved with artificial intelligence this year.

The latest:

A proof of the Riemann hypothesis, unsolved since 1859, has been posted by Anthropic mathematician Levent Alpöge, who says he reached it with help from the company’s Claude model. He published the argument Friday in Lean, a language that lets computers check proofs line by line. It landed one day after Anthropic’s stock market debut.

Details:

  • The claim: Alpöge said on his website that he believes he has found the solution to the Riemann hypothesis with Claude’s assistance, according to The Wall Street Journal. He posted the argument in Lean, a programming language that allows machine verification line by line, and mathematicians began working through it immediately.
  • The problem: The Riemann hypothesis is a conjecture in number theory proposed in 1859. For 167 years no mathematician produced a proof, and in 2000 it was named one of the seven Millennium Prize Problems, a set of challenges that became the field’s best-known open questions.
  • The scale: Alpöge described the full argument as both incredibly long and, in his own words, “some massive unreadable travesty of ‘slop,'” with the AI-written version running to more than 250,000 pages. That length is itself the main obstacle facing anyone trying to assess the work by reading it.
  • The timing: The result was reached late Sunday night into Monday, Alpöge said, days before Anthropic listed publicly. He called the timing horrible and said he messaged chief executive Dario Amodei for permission to announce it, writing that there was a constraint to respect.
  • The company: Anthropic did not respond to a request for comment after Alpöge posted his findings Friday morning, according to the Journal. The company’s shares fell on their first full day of trading, dropping about 10% as the claim circulated.
  • The precedent: If the proof holds, it would be the second Millennium Prize Problem taken down with artificial intelligence, after an OpenAI model and millions of dollars in computing power solved one in September. Humans working alone have cracked only one of the seven.
  • His reasoning: Alpöge wrote that he could not live with waiting any longer before going public, and framed the release as part of what he called a revolution in mathematics that he hopes points toward discoveries still ahead in the simplest things.
  • The caveat: The claim rests on Alpöge’s own statement that he believes he has a proof; no outside verification has been announced. He did not say how much of the argument Claude generated versus how much he directed, and Anthropic has not endorsed the result.

Background:

The Millennium Prize Problems were selected in 2000 as seven of mathematics’ hardest open questions, each carrying a million-dollar award. The Riemann hypothesis, concerning the distribution of prime numbers, has long been treated as the most prominent of them.

Between the lines:

Two facts sit uneasily together: the proof was reached days before the IPO, and announced the day after it. Alpöge himself called the timing horrible and said he sought Amodei’s permission before publishing, indicating the delay was a company decision rather than a mathematical one. The 250,000-page length also means confirmation will not be quick, and Anthropic’s silence leaves the claim resting on one employee.

What’s next

Watch for mathematicians reporting results from running the Lean file, which can verify the argument mechanically, and for any statement from Anthropic or Amodei. A Clay Mathematics Institute assessment would be the formal test of the claim.

Read on ontime+ ↗