Levent Alpöge, a mathematician at Anthropic, says he and Claude found a proof of the Riemann hypothesis and the Generalized Riemann Hypothesis. He posted only a Lean formalization. Outside experts have not confirmed it.
The Riemann hypothesis, the most famous open problem in mathematics, may have fallen to a human working with Claude. Levent Alpöge, a mathematician who works at Anthropic, said on November 20 that he believes he has a proof. The claim is unverified. He has released only a machine-checkable version of the argument, and outside experts had not confirmed it at publication.
What Alpöge claims
In a two-page statement on his website, Alpöge writes: “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis.” He says the method extends to the Generalized Riemann Hypothesis, a broader version, and that he expects it to reach the Extended Riemann Hypothesis “with more time.”
The Riemann hypothesis dates to 1859. It concerns how prime numbers are spread out, and a long list of results in number theory assume it is true. It is one of the problems the Clay Mathematics Institute pays $1 million to solve.
Alpöge says the argument is “incredibly long” and that he and Claude have formalized it in Lean, a programming language that checks every step of a proof by machine. He posted the Lean files as a 14-part download. He chose not to post a full written proof yet, calling a machine-generated document of that size “unreadable.” He says he will write the argument up properly and expects it to be “massively simplified.”
What Claude did, according to Alpöge
Alpöge says he steered the work toward the right setting and Claude did much of the technical work near the end. “This time Claude far outpaced me near the finish line,” he writes. He gives one example for mathematicians: Claude proposed the length of a chain of “theta lifts,” a step he says lets the argument copy a trick that already works in a simpler setting.
He also addresses the timing. The statement came one day after Anthropic’s stock market debut. Alpöge writes that the result came together “late Sunday night,” that he asked Anthropic’s leadership about announcing it, and that “there was a constraint to respect.” He calls the cynical reading of the timing “false.”
Who has checked it
According to Scientific American, which reported the claim, experts had not yet confirmed the Lean code or the proof. Lean confirms that a proof is logically valid. It does not confirm that the formal statement in the code says exactly what the Riemann hypothesis says. That part takes human reviewers, and Scientific American expects the check to be quick.
Anthropic had not published a statement as of the afternoon of November 20. Alpöge has made big claims before: Scientific American notes he posted a Claude-generated solution to the Jacobian conjecture in July.
A Lean file shifts the burden of proof
A written proof this long could take experts months to read. A Lean formalization can be checked by machine, so the question narrows fast to one thing: does the code state the right theorem with no hidden assumptions? That check should take days.
For people who buy AI tools, the takeaway is narrower than the headline. Anthropic already reported that Fable 5.2 agents formalized Fermat’s Last Theorem in Lean in 30 hours. If this proof holds, a Claude model will have done original research-level work on mathematics’ hardest open problem with one expert guiding it. That says a lot about long, multi-step reasoning, the same skill behind coding agents and analysis work. It says nothing yet about the Claude plan your team pays for. Wait for the verification before drawing conclusions, and discount any marketing that leans on this claim before then.
Anthropic began trading publicly this week; its earlier filing is covered in Anthropic’s IPO filing shows a $17 billion quarter.
