A Riemann hypothesis proof claim landed on Friday 20 November 2026 — and within hours it was the most talked-about story in mathematics. Levent Alpöge, a mathematician who works at the AI company Anthropic, posted a four-page note on his own website saying he believes he has, working with the company’s Claude chatbot, discovered a proof of the Riemann hypothesis, the 167-year-old puzzle about prime numbers that has resisted every mathematician since 1859. The catch, and it is a large one: the proof is not yet fully verified. Independent mathematicians have now compiled and rechecked the Lean code and say it is almost certainly real, but a final comparator check is still pending. What Alpöge has released is not a conventional paper but computer code — a formalisation in the Lean programming language that machines can check step by step. Until independent experts confirm that the code really proves what it claims to prove, the hypothesis stays unsolved, and the $1 million prize stays unclaimed. This article explains what was actually claimed, what the hypothesis says in plain language, how the checking will work, and what happens next.
What the Riemann Hypothesis Proof Claim Actually Says
Alpöge’s words leave little room for doubt about what he believes he has done. “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis,” he wrote in the note posted on his website on 20 November 2026, as reported by Scientific American. The claim goes further than the original problem: Alpöge says the same method also settles the Generalized Riemann Hypothesis, a far broader family of related conjectures about Dirichlet L-functions, which track how primes fall into arithmetic progressions.
The unusual part is the form the proof has taken. There is no conventional paper explaining the full argument yet. Instead, Alpöge released the proof as code written in Lean, a programming language and proof assistant in which every logical step is written in a formal language and checked mechanically by a computer. The archive was posted on his site in 14 parts and runs to roughly 239 megabytes and 175,187 source files, according to Interesting Engineering. A readable human account of the mathematics would be extremely long: Alpöge said he has not yet mastered all of the mathematics involved and plans to write it up properly, and one report put the unpublished human-readable version at more than 250,000 pages.
The note sketches a route that starts from Stepanov’s method, a technique devised by Sergei Stepanov in the 1960s, and runs through long chains of mathematical constructions called theta lifts, using weighted prime sums known as von Mangoldt sums. It closes with a 20-step timeline, running from 6 October to 16 November 2026, which Alpöge says Claude assembled. Phys.org reports that Alpöge consulted Anthropic chief executive Dario Amodei before announcing the result, and acknowledged that its timing could be viewed cynically — without elaborating. The announcement came a day after Anthropic’s record stock-market debut, when its shares closed around $97, about 17 percent above the $83 offer price.
The Riemann Hypothesis in Plain Language
Prime numbers — 2, 3, 5, 7, 11 and on — look random at first glance. They thin out as numbers grow, but nobody can predict exactly where the next one will turn up. The Riemann hypothesis, first proposed in 1859 by the German mathematician Bernhard Riemann, says that behind that apparent randomness there is a precise hidden order. It claims that every “nontrivial” zero of a mathematical object called the Riemann zeta function lies on one particular vertical line in the complex plane, the line where the real part equals one-half.
That sounds abstract, but the stakes are concrete. The zeros of the zeta function encode the distribution of the primes: each zero contributes a finer detail to the sequence of primes, as Anthropic itself explains. If the hypothesis is true, mathematicians gain an exact grip on how primes are spread out. If it is false, that grip loosens. The hypothesis has become the North Star of number theory. Computers have checked the first ten trillion zeros and found every one of them on the line — Gourdon and Demichel verified the hypothesis up to the ten-trillionth zero — but checking cases, even trillions of them, is not a proof. A proof has to cover all infinitely many zeros at once, and that is what has defeated every attempt for 167 years.
Thousands of other results in mathematics have been proved on the assumption that the hypothesis is true. As Scientific American puts it, a correct proof would “instantly unlock a deluge of truths contingent upon the ambitious conjecture.” Hector Pasten, a mathematician at the Pontifical Catholic University of Chile, told the magazine: “If the proof is correct and understood, this completes a dictionary between different branches of math where the connections were only partial. Something like a Rosetta Stone.”
How Experts Will Check the Riemann Hypothesis Proof
Releasing the argument as Lean code gives the mathematical world a concrete way to test it, and that is a large part of why the claim is being taken seriously. A proof written in Lean can be checked against explicit definitions and logical rules by any computer running the same software. But a successful check has several stages, and each one can fail.
First, the files have to build — the code must compile without errors. Second, and most important, the theorem the code states has to be the real Riemann hypothesis: the formal statement must faithfully match the conjecture as mathematicians understand it, because a proof of a subtly different statement proves nothing. Third, the proof’s dependencies have to hold, with no unsupported assumptions smuggled in. The explainx.ai analysis of the claim makes the same point: as of 20 November 2026, no independent expert had confirmed the proof, and no expert report existed.
The early signs are encouraging but narrow. MIT mathematician Andrew Sutherland told Scientific American he had independently compiled Alpöge’s result in Lean, calling it “an historic moment” and adding: “There is much work to be done to understand the mathematical details of the proof.” An independent verification project has reportedly confirmed that the downloaded files match Alpöge’s published archive, and early accounts say the main theorem statements compile against Mathlib, Lean’s standard mathematics library. Compiling, though, is not the same as being right. Scientific American notes that issues can arise in Lean, and that it should not take experts long to establish whether the hypothesis is properly coded and whether the proof is correct. The first independent verdicts are expected within days.
It is worth remembering that the internet already holds many claimed proofs of the Riemann hypothesis, including papers on arXiv and Zenodo. What sets this Riemann hypothesis proof claim apart is the combination of a machine-checkable formalisation, a claimant with serious standing in the field, and coverage by established science outlets. None of that makes it correct. It makes it checkable, quickly, which is exactly what will happen next.
A second proof claim arrived the same day. Stanford mathematician Ravi Vakil, president of the American Mathematical Society, uploaded a 940-page manuscript by Stanford postdoctoral scholar Alessandro Masullo proposing a human-written proof of the generalized Riemann hypothesis. Mathematician Brian Conrad has been reviewing Masullo's manuscript for months, and Vakil says it is so far looking good. It is a separate claim from Alpöge's, and it has not been verified either.
Why the $1 Million Prize Is Not Won Yet
The Riemann hypothesis is one of the seven Millennium Prize Problems named by the Clay Mathematics Institute, each carrying a $1 million award. That is the figure in this article’s headline, and it is also the reason the prize is not won yet. Clay’s rules are slow by design: a proposed solution must be published in a qualifying outlet, then survive at least two years of scrutiny and gain general acceptance by the global mathematics community before the institute will even consider it.
Alpöge’s Riemann hypothesis proof claim clears none of those bars today. It has not been published in a peer-reviewed outlet. Two years have not passed. The mathematical community has not accepted it — it has barely begun to read it. So even in the best case, the $1 million is years away, and in the worst case it never comes into play at all. If the proof turns out to be wrong, nothing changes: the hypothesis stays open, the prize stays unclaimed, and the search continues.
That is also why the wording around this story matters so much. “Solved” and “claimed” are different words with different consequences, and several headlines on Friday blurred them. This article uses “claimed” throughout, because that is what the evidence supports right now: a claim, published as code, awaiting judgment.
Claude’s Year of Mathematical Firsts
The claim did not arrive out of nowhere. It caps a year in which AI systems have moved from unreliable maths students to something mathematicians had not expected to see. In July 2026, Alpöge tweeted a Claude-generated solution to the famous Jacobian conjecture, kicking off what Scientific American calls maths’ “unprecedented summer.” In August, Anthropic said an unreleased version of Claude had improved a long-standing result related to the Riemann hypothesis itself, raising the share of zeta zeros proven to lie on the critical line from 41.6 percent to 67.2 percent — 67.25 percent in the paper’s precise terms. Anthropic stressed at the time that Claude had not proved the hypothesis, and added that it did not expect those techniques to lead to a full proof. Mathematician Youness Lamzouri later gave a different, more direct proof of a related bound.
September brought a row over the Navier-Stokes equations, when New York University mathematician Tristan Buckmaster implied that OpenAI had taken work belonging to him and Alpöge. In October, one of the proofs in OpenAI’s large release of mathematics results made significant progress on making sense of the primes. Against that background, the November claim looks less like a bolt from the blue and more like the next step in a fast-moving sequence — though it is a far larger step than any before it.
MIT’s Andrew Sutherland once summed up the stakes in a line Scientific American repeated on Friday: “An AI that can prove the Riemann hypothesis is not one I’d want to meet. If that happens, mathematicians having jobs will be the least of our problems.” Carlo Pagano, a mathematician at Concordia University and a visiting researcher at Google DeepMind, told the magazine that if the claim holds up, “the new ideas in the proof will open up brand new avenues of thinking in number theory and have an immense impact beside the obviously extremely influential result in itself.”
What Happens Next — and Why It Matters
The short term is mechanical: independent teams will run the Lean code to test the Riemann hypothesis proof, check that the theorem statement matches the real hypothesis, and hunt for gaps. The first verdicts on the code should arrive within days. The longer term is human: mathematicians will need to understand the argument, Alpöge will need to produce a readable account of it, and the work will need to pass peer review before the Clay Mathematics Institute’s two-year clock can even start.
Alpöge’s own note strikes an elegiac tone, recognising what the end of a 167-year quest would feel like. “It is natural to feel that this is an end, in my case a loss of a principal guiding light,” he wrote, “but what I saw during the course of progress so amazed me that I believe it is one dawn across science of something far greater.”
If the proof is correct, the payoff is not a faster computer or a new product — Scientific American is explicit that it has no immediate real-world applications, and will not upend global banking or make the internet run faster. The payoff is inside mathematics: a dictionary between branches of the subject that were only partly connected, thousands of conditional results turned unconditional, and new lines of research opened by the ideas in the proof itself. If the proof is wrong, the field loses little: the August work on the 67.25 percent bound already showed that even a failed attempt at the full problem can leave behind real progress.
Either way, the next few days will tell the world more than the last 167 years have about whether this Riemann hypothesis proof claim holds up. The code is public. The checks are running. And until they report back, the honest sentence is the modest one: a mathematician and an AI say they have proved the most famous problem in mathematics — and nobody yet knows whether they have.