Discoveries

Anthropic mathematician claims Claude helped prove the Riemann hypothesis, not yet verified

Levent Alpöge, a number theory researcher at Anthropic, has posted a claimed proof of the Riemann hypothesis on his website. The proof exists so far only as Lean code, and independent experts have yet to check it.

AI text based on reports by TechRadar, The Independent, EL MUNDO and other publishers. Published . How it is made

For membersThe first 30 seconds for everyone. Synthetic voice (AI).

In brief

How it came to this
  1. September 2026OpenAI claims a solution to the Navier-Stokes problem
  2. 19 NovAnthropic makes its stock market debut
  3. 20 NovAlpöge publishes the claimed proof on his website

The claim

Levent Alpöge, a number theory researcher at Anthropic, posted a statement on his personal website on Friday, 20 November, saying he had found a proof of the Riemann hypothesis while working with Claude, the company's AI model. He wrote that the same method extends to the Generalized Riemann Hypothesis and that he expects, with more time, to reach the Extended Riemann Hypothesis as well. The hypothesis, proposed by Bernhard Riemann in 1859, concerns how prime numbers are distributed and has resisted proof for 167 years. The Independent reports that Alpöge holds dual United States and Turkish citizenship and degrees from Harvard, Cambridge and Princeton. In the statement he described the work in these words:

“I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis.”

Levent Alpöge, anthropic mathematician

Proof and verification

Alpöge released the argument in Lean, an open-source programming language used to check mathematical reasoning. TechRadar reports that no human-readable version exists yet, which Alpöge partly attributes to the amount of mathematics in the argument and to his wish to avoid publishing an unreadable document. An early draft reportedly ran to more than 250,000 pages. The Independent notes that experts must now examine whether the work is properly coded and whether the hypothesis has actually been proved. Spiegel Online reports that Alpöge does not yet fully grasp the proof himself and cannot yet explain it in understandable terms. Until other mathematicians vet the argument, the claim rests on the statement and the code Alpöge published himself.

Anthropic and the timing

Anthropic has not commented on the announcement or taken part in it, The Verge reports. Alpöge's document says he texted Anthropic CEO Dario Amodei about the solution late on Sunday night and held back publication until after the company's stock market debut on Thursday, 19 November. Earlier, in July, Alpöge posted a counterexample to the Jacobian conjecture, a problem in algebraic geometry, which he said he had also reached with Claude. Anthropic said in August that it was working on the Riemann hypothesis with Claude, and it added then that it did not expect the approach to prove it:

“We don't expect that the techniques Claude used will lead to proving the Riemann hypothesis.”

Anthropic

The prize and the problem's history

David Hilbert included the hypothesis in his list of 23 mathematical problems in 1900, and in 2000 the Clay Mathematics Institute offered $1 million for a proof. Under the institute's rules, a published proof must withstand at least two years of scrutiny and win general acceptance in the global mathematics community before it can be considered for the prize. The Verge reports that only one of the seven Millennium Prize problems has been officially solved in the quarter-century since the prize was set up, the Poincaré conjecture. In September OpenAI claimed to have cracked another of them, the Navier-Stokes problem, after hearing that rival researchers were closing in on a solution.

Ask about this text

Sources and context

Pollar is free and carries no ads in the text. Its members keep it going.

Support Pollar