Riemann Hypothesis Proof Claimed With Claude’s Help
Anthropic mathematician Levent Alpöge says he and Claude have proved the Riemann Hypothesis, the 167-year-old problem with a $1 million prize on it, and has released the proof only as Lean code. MIT’s Andrew Sutherland says the code compiles; whether it proves exactly the right statement is still to be checked.
Levent Alpöge, a mathematician at Anthropic, posted a four-page note on his website on Friday, November 20, that opens: “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis.” He goes further. The method, he writes, “easily extends much further,” resolving the Generalized Riemann Hypothesis as well, and he expects to reach the Extended Riemann Hypothesis with more time. He also says that he and Claude have formalized the arguments “without assumptions” in Lean, a programming language in which a computer checks every step of a proof.
The Riemann Hypothesis, posed by Bernhard Riemann in 1859, says that all the nontrivial zeros of the zeta function lie on a single line, the one where the real part equals one half. If it holds, mathematicians get the best possible control over how prime numbers are spread out, and a large body of results that currently start with “assuming the Riemann Hypothesis” becomes unconditional. It is one of the Clay Mathematics Institute’s Millennium Prize Problems, each carrying a $1 million award.
Only Lean, for now
Alpöge has not released a written paper. The proof is published as a Lean project split into fourteen archive files on the same site. In the note he says the argument is “incredibly long at the moment,” and that he is presenting a complete argument before having mastered much of the mathematics around it. Producing a huge written version would be routine, he adds, but he chose not to post what he calls an unreadable “slop” document; a footnote says one machine-written draft ran past 250,000 pages. He plans to study the argument, simplify it and write it up properly, and expects much shorter proofs to follow quickly.
The note describes how the proof was found. Alpöge says he steered the work toward combining two ideas he had seen in a simpler setting, and that “Claude far outpaced me near the finish line.” The argument descends from Stepanov’s method, which proves the analogue of the hypothesis over function fields, and uses chains of theta lifts to imitate a step that has no obvious counterpart over the integers. A 20-step timeline that Claude compiled runs from October 6 to November 16, the date given for the Generalized Riemann Hypothesis. Alpöge warns that his summaries of the parts where Claude was the expert may be inaccurate, and says the Lean code is the authority. The note does not say which Claude model was used.
What mathematicians say so far
The first checks are on the code, not the mathematics. Scientific American reports that Andrew Sutherland of MIT independently compiled the Lean files and called it “an historic moment,” while adding that there is “much work to be done” to understand the details. A successful compile shows that Lean accepts every step. It does not show that the theorem stated in the code is the Riemann Hypothesis rather than something subtly weaker, and that is what experts now need to read. Hector Pasten of the Pontifical Catholic University of Chile told the magazine that, if correct, the proof works like “a Rosetta Stone” between branches of mathematics.
Brian Conrey, a leading specialist on the hypothesis, told AFP it was “a monumental moment in human history,” though he had seen only the posted remarks, not a manuscript, and expected simpler routes to appear. AFP notes the claim has not been independently verified or reviewed by outside experts. Under Clay’s rules, a solution has to be published in a peer-reviewed journal and survive two years of scrutiny before the prize can be awarded.
The timing
The note arrived a day after Anthropic’s record stock market debut, and Alpöge addresses that directly. He writes that the final step came late on a Sunday night, that he then messaged Anthropic chief executive Dario Amodei to ask whether he could announce it, and that there was a constraint to respect. He calls the timing “horrible” and says the obvious cynical reading of it is false. Anthropic did not respond to AFP’s request for comment, and its newsroom carried no post about the claim when BitsMinds checked.
The claim lands in a year of AI mathematics results that are still being sorted out. In August, Anthropic said an unreleased Claude model had raised a long-standing bound related to the zeta function, while stressing that it had not proved the hypothesis. In September OpenAI said its agents had solved the Navier–Stokes problem, another Millennium problem, a result followed by a credit dispute involving the NYU mathematician Tristan Buckmaster and Alpöge himself. Alpöge’s note praises OpenAI’s mathematical work, and thanks colleagues including Manjul Bhargava and Peter Sarnak, without suggesting that any of them has checked the proof.
More on Claude
Evergreen coverage we keep current — start here.
Want AI news before everyone else?
The morning's most important AI stories, straight to your inbox. No fluff.