A mathematician at Anthropic says he has a Riemann hypothesis proof. In a four-page note on his website, Levent Alpöge says Claude helped him develop the argument. “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis,” he wrote. The hypothesis, posed by Bernhard Riemann in 1859, describes how prime numbers are distributed. AFP reported that he released the note on Friday. Clay has allocated $1 million to the problem. It alone decides whether to award the prize.
No journal paper accompanies the announcement. Alpöge describes the proposed argument as formalized in Lean, a programming language that checks each step of an argument for errors, and released it only in that form. His note identifies 14 archive parts of the code. In a footnote, he says an experiment produced more than 250,000 pages, and that posting such a text would be “obviously inappropriate” because it would be “some massive unreadable travesty of ‘slop.'”
He also set limits on what he understands. “The argument is incredibly long at the moment,” Alpöge wrote. He is “presenting the world with a complete argument before having mastered a massive neighbourhood around its mathematics.” Alpöge expects a shorter argument to emerge quickly, writing that the proof “will be massively simplified, likely immediately.” He plans a proper write-up and talks, and the note gives no timeline for either.
The claim reaches further than the original problem. Alpöge wrote that the method “easily extends much further, resolving the so-called Generalized Riemann Hypothesis,” and that he expects the Extended Riemann Hypothesis with more time. He said Claude and he “formalized the arguments without assumptions.”
What Anthropic’s researcher said about the timing
Alpöge’s note appeared one day after Anthropic's Nasdaq debut. He addressed the optics himself. “The precise timing is horrible,” he wrote. “There is an obvious cynical view, to do with the timing, but it is false.” The work finished late on a Sunday night, he said, and he messaged Dario Amodei, listed on Anthropic’s board of directors, “eventually asking about my ability to announce.” He added that “there was a constraint to respect,” without saying what it was.
How the $1 million prize actually works
The Clay Mathematics Institute set aside a $7 million fund for its seven Millennium Prize Problems, with $1 million allocated to each. Its rules, last approved on September 26, 2018, put several steps between an announcement and a payment. A proposed solution normally must appear in a refereed mathematics publication of worldwide repute. Clay’s board can approve a relaxed set of conditions, on its Scientific Advisory Board’s recommendation. Qualifying outlets also need a named, contactable editorial board and a published refereeing process. Inclusion in MathSciNet’s list of publications is another of the required traits.
Then the waiting starts. At least two years must pass after publication in a qualifying outlet, and the solution must gain “general acceptance in the global mathematics community.” In judging that acceptance, CMI may consider independent publications, conference discussion and awards. It may weigh any other factor it sees as indicative of consensus. Only then does the institute decide whether a solution merits detailed consideration. If it does, CMI convenes a special advisory committee of at least three members, including at least two experts in the field.
Other rules shape who gets credit. CMI will not accept a solution sent to it directly, and it will not consider supplementary material from the author. It may divide a prize among several solvers. It also pays “special attention” to whether a solution depends crucially on insights published earlier, and may recognize or include the authors of that prior work.
What mathematicians have checked
Clay’s news page carried no item on the new claim when it was checked on November 22, 2026. The institute has spoken about a comparable claim this year. After an announcement that the Navier-Stokes problem had “apparently been settled” in September, CMI said the process for evaluating what has been achieved “is deliberately unhurried,” and it promised updates.
A second claimed proof appeared the same day. Ravi Vakil wrote on his Math 216 blog that Alessandro Maria Masullo has a proposed proof of the Generalized Riemann Hypothesis. Brian Conrad has examined it for months, Vakil wrote, and “it is so far looking good to him. It is all human, with no AI.” In an update about eight hours later, Vakil said several people reported problems in the proposed proof, “identified with the assistance of different AI models.” All issues must be “resolved to the satisfaction of the community before this proof would be accepted as true,” he wrote. The sources reviewed do not establish either proof.
The Lean files are public, so the code can be checked before any paper appears. Lean checks the logic, but Alpöge says he is still working to understand the mathematics around his argument. He says he plans to write up the argument and give talks about it.














































