Anthropic Mathematician Levent Alpöge Posts Unverified Riemann Hypothesis Proof Claim

Levent Alpöge, a mathematician employed by Anthropic, posted a four-page note on his personal website on Friday saying he had completed a proof of the Riemann hypothesis by collaborating with Claude, the company’s artificial intelligence system. The claim has not been verified. Dr. Alpöge published no conventional mathematical paper; he linked instead to computer code that he said would verify the result, and other researchers said they were checking it. The Riemann hypothesis has resisted proof since 1859 and is one of the seven Millennium Prize Problems listed by the Clay Mathematics Institute, which attaches a $1 million prize to each.

What the note says, and what it does not

The New York Times reported that Dr. Alpöge, described as a mathematician at Anthropic, posted the four-page note on Friday and said he had completed the proof by collaborating with Claude. The Times reported that he did not include a traditional mathematical paper to describe the proof, but instead linked to computer code that he claimed verified the result.

The Information, in its account of the document, described the work as a claim to have solved the most famous open problem in math with help from Anthropic’s AI models — a plural framing. The Times account names Claude, the company’s AI system, as the collaborator. The Financial Times headlined its story “Anthropic staffer claims solution to mathematics’ Riemann hypothesis.” Neither account described a journal submission, a referee report or a formal adjudication of the claim.

Dr. Alpöge’s own language, quoted by the Times, credited the machine with the final stretch. “Claude and I have had a great time together, with such breadth in recent months,” he said. “But this time, Claude far outpaced me near the finish line. No, Claude flew, and danced, and sparkled near that finish line.” In the note itself he wrote of a “revolution in mathematics” and of hoping the argument “provides a glimpse towards the miracles that lie ahead.”

Lean code, not a paper: how the claim is meant to be checked

The Times reported that the code Dr. Alpöge linked to is written in Lean, a language that turns mathematical statements into a form a computer can check. That choice moves the dispute out of the seminar room and into a compiler: a Lean file either builds against its stated theorem or it does not, and the check can in principle be repeated by anyone with the toolchain.

It also shifts the question being asked. A successful build establishes the theorem as formalized in the file. Whether that formal statement is the Riemann hypothesis as mathematicians phrase it is a separate reading, and one the researchers now examining the code would have to make. The Times reported that other researchers said they were checking Dr. Alpöge’s code. No result of that checking has been reported.

Steven Strogatz, a professor of mathematics at Cornell University, told the Times the problem was “the Mount Olympus.” He said it seemed clear that Dr. Alpöge had played a central role in telling Claude which mathematical ideas to explore to pull the proof together, and described the division of labor as “a very skillful jockey riding a really fast horse” — something, he said, “we might see a lot more of in the future.”

The note landed the day after Anthropic’s IPO

The Wall Street Journal reported that the mathematician went public with the claim “the day after Anthropic’s historic IPO,” and headlined its story “Anthropic Mathematician Says AI Has Solved the Most Famous Problem in Math.” By the Journal’s account, the gap between the listing and the posting was one day.

A market-data module displayed on the Journal’s article page showed the stock, ticker ANT, down 13.04%. The passage available did not link that move to the proof claim, and no source in the material reviewed connected the two. The percentage is a quote reading attached to the page, not a stated cause, and it carries no timestamp in the text reviewed.

Anthropic had not, as of the accounts reviewed, published its own assessment of the argument. The company’s name appears in the coverage chiefly as the employer of the mathematician and the maker of the system he credits.

A crowded scoreboard for AI in mathematics this year

The Information reported that in August, Anthropic used an unreleased version of Claude to make progress on a related problem. In October, OpenAI said it had solved a relaxed version of the Riemann hypothesis. In September, OpenAI announced that it had solved the Navier-Stokes problem, another of the Millennium questions.

If Dr. Alpöge’s argument survives checking, The Information reported, it would be the second Millennium problem solved with AI this year — after the Navier-Stokes announcement. The Times framed the stakes in the same register, writing that a confirmed proof would cement the field’s sense that no unsolved problem is safe from artificial intelligence, and that mathematicians have been repeatedly stunned this year by machines conquering problems that had eluded human geniuses.

The sequence matters for how the claim is being read. Three of the four data points in it — August, September, October — are company announcements about company systems. The fourth is an individual posting a note and a code file on his own site, without a company release attached.

The $1 million sits with the Clay institute, not the posting

The Riemann hypothesis is one of the seven Millennium Prize Problems, a list that carries a $1 million prize for each, as The Information reported. The prize is defined by the Clay Mathematics Institute’s rules. Nothing in the accounts reviewed said the institute had received, reviewed or recognized the argument, and no award was reported.

The problem is 167 years old in the framing both outlets used: Bernhard Riemann, a German mathematician, formulated it in 1859. It concerns when a mathematical equation called the Riemann zeta function is equal to zero, a question tied to the nature of prime numbers and to how those primes are distributed along the number line as it stretches toward infinity.

The Times noted how quickly primes thin out. Up to 100, some 25% of the numbers are prime; up to one million, only about 7.8% are. Riemann hypothesized a formula for estimating, or counting, how many primes exist below any given number — the counting question the hypothesis governs.

Questions and answers

Has the Riemann hypothesis been proven? No. A mathematician has posted a claim that he proved it, and other researchers said they were checking his code. No verification, referee report or institutional confirmation has been reported.

Who is Levent Alpöge? The Times and the Journal both identify him as a mathematician at Anthropic, the AI developer. The Journal reported that he went public with the claim the day after the company’s IPO.

What did he actually publish? A four-page note on his personal website, posted on Friday, November 20, 2026, plus a link to computer code written in Lean that he said verifies the result. He did not publish a conventional mathematical paper.

Why does Lean matter? Lean turns mathematical statements into a form a computer can check, so the file either builds against its stated theorem or it does not. That makes the check repeatable by outsiders, but it also means the formal statement in the file has to be read against the Riemann hypothesis itself.

What is the Riemann hypothesis? Formulated by Bernhard Riemann in 1859, it concerns when the Riemann zeta function equals zero and bears on how prime numbers are distributed. Primes thin out as numbers grow: roughly 25% of numbers up to 100 are prime, against roughly 7.8% up to one million.

Is there a $1 million prize? The Riemann hypothesis is one of the seven Millennium Prize Problems, each carrying a $1 million prize under the Clay Mathematics Institute’s programme. No award, adjudication or official review was reported in the accounts reviewed.

How does this fit the rest of 2026? The Information reported that Anthropic used an unreleased Claude version in August to make progress on a related problem, that OpenAI said in October it had solved a relaxed version of the Riemann hypothesis, and that OpenAI announced a Navier-Stokes solution in September.

The claim now rests on a four-page note, a Lean file and the patience of the mathematicians reading both. Until that reading is published, the Riemann hypothesis remains open and Dr. Alpöge’s posting remains a claim.

Share your love