Short answer: no one has confirmed it yet. On Friday, 20 November 2026, Levent Alpöge, a mathematician who works at Anthropic, posted a four-page note saying he and Claude found a proof of the Riemann Hypothesis, and of the much stronger Generalized Riemann Hypothesis (GRH). He released the proof only as Lean code, a format a computer can check line by line. Early checks by other mathematicians have gone well, but the final automated checks were still running when we wrote this, and no human has yet read and understood the argument. The same day, a separate, human-only proof claim by Alessandro Masullo was announced. That one has not been verified either.

vibegamer.io is a site about browser games made with AI. We have no connection to Anthropic, Alpöge or Masullo. Everything below comes from the linked sources.

The claim at a glance

Who Levent Alpöge, a mathematician at Anthropic, working with Claude
When Posted Friday, 20 November 2026
What A claimed proof of the Riemann Hypothesis and the Generalized Riemann Hypothesis
Format A four-page note plus Lean code in 14 download parts; no full human-readable paper
Size 17.4 million lines of Lean, according to Scott Aaronson
Status Not verified yet. Independent Lean builds passed; a final comparator check and human review are still pending
A second claim Alessandro Masullo, a human-only proof announced by Ravi Vakil the same day, also unverified

What is the Riemann Hypothesis?

It is a guess Bernhard Riemann made in 1859 about the prime numbers. Primes look scattered at random, and the hypothesis says their spacing actually follows a precise hidden pattern, tied to where a function called the zeta function equals zero. It is the most famous open problem in mathematics, and the Clay Mathematics Institute offers a $1 million prize for it. The "generalized" version extends the same claim to a whole family of related functions, so proving it would settle many results that mathematicians have long assumed.

What Alpöge's note says

The note, posted on his website, opens with this sentence: "I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis." He writes that the method extends to the Generalized Riemann Hypothesis, and that he expects it to reach the Extended Riemann Hypothesis with more time.

  • Timeline. The note lists 20 steps in a timeline it says Claude put together. The first entries are dated 6 October 2026, and the last, GRH itself, is dated 16 November 2026.
  • Who did what. Alpöge says he set up the general approach and that Claude far outpaced him near the finish, especially in areas of advanced number theory he says he hasn't mastered himself.
  • Lean only. He says the formalization is finished and that he chose to release only the Lean code instead of a huge machine-written paper. In a footnote he says that draft passed 250,000 pages. The code is split into 14 files on his site.
  • Not polished. He says the argument is extremely long for now, says he may have made odd choices that shorter proofs will replace, and says he plans to write it up properly.

Scientific American describes Alpöge as an Anthropic employee and says he was behind a Claude-generated solution to the Jacobian conjecture in July. Terence Tao writes that he used an internal AI model. As of 21 November we found no post about the claim on anthropic.com.

A separate human proof, the same day

Hours later, Stanford mathematician Ravi Vakil, president of the American Mathematical Society, put up a hurried blog post saying Alessandro Masullo has a proposed proof of GRH. Vakil writes that Brian Conrad, a Stanford expert, had been checking it for months and that so far it looks good. He also writes: "It is all human, with no AI." The draft, titled The Music of the Primes, is linked from the post, and Vakil asks others to check it and formalize it.

Scientific American reports that Masullo is a young mathematician also at Stanford, that Conrad confirmed the details but said the paper is undergoing major revision, and that the proof has not been verified by anyone. It also says some mathematicians are already skeptical.

What experts have said so far

  • Scott Aaronson, a computer scientist, wrote on his blog that Alpöge says Anthropic models proved GRH in 17.4 million lines of Lean. He points out that Lean has had bugs in the past and that a motivated AI could in theory exploit one, but says he sees no sign of that here and bets the proof will stand, while saying he's not 100% sure. He adds that mathematicians won't be satisfied until they can explain the ideas themselves, which may take years.
  • Terence Tao listed it as the least digested, meaning least understood, of three AI-assisted results Alpöge was part of. He calls it an enormous Lean-only proof, says first looks suggest it uses the full power of modern Langlands-program mathematics, and expects understanding it to take time. He points readers to a thread on the Lean community chat where the checking is happening.
  • Andrew Sutherland of MIT and others compiled the Lean code independently without problems, Scientific American reports, and Sutherland said the kernel check used only the three standard axioms. Kevin Buzzard of Imperial College London said they were still waiting for a comparator run, a check that the code doesn't quietly change the meaning of the theorem and that it passes several independent checkers.

What happens next

  1. Finish the Lean checks. The comparator run tests that the theorem stated in Lean really is the Riemann Hypothesis and that nothing in the code redefines its terms.
  2. Understand the proof. A computer-checked proof that nobody understands is not the end. Mathematicians will try to pull out the ideas, as Aaronson and Tao both say, and Alpöge says he will write a proper paper.
  3. Peer review. Neither claim has been through peer review. For Masullo's paper, Conrad says it needs to reach a more final revision before anyone can judge it.

Why it matters for AI

If it holds up, this would be one of the strongest signs yet that an AI model can produce new mathematics, not just explain known results. Scientific American's explainer adds that it would not directly break encryption like RSA. It is also a big step from August, when Anthropic reported that Claude raised a known bound on zeta-function zeros from 41.6% to 67.2%, and said then that it didn't expect those techniques to lead to a proof of the Riemann Hypothesis.

On Hacker News, the thread on Alpöge's note had 2,377 points and 1,603 comments when we checked on 21 November.

Claude and VibeGamer

Claude is also the model behind many of the browser games on VibeGamer. If you're curious about Anthropic's newest model, read what Fable 5.2 is, or play some of the games made with Claude. They're games, not math proofs.

FAQ

Did Claude prove the Riemann Hypothesis?

Not confirmed. Levent Alpöge says he found a proof working with Claude, and early Lean checks have passed, but the final checks and human review are not done. Until they are, it is a claim.

Who is Levent Alpöge?

A mathematician who works at Anthropic, according to Scientific American. His note says he has been planning to work on this problem since he was a student.

Is there a paper I can read?

Not a full one from Alpöge. He posted a four-page note and the Lean code. Masullo's draft, The Music of the Primes, is linked from Ravi Vakil's blog post.

What is Lean?

A programming language and proof checker. If a Lean proof compiles, the computer has checked every step. Experts still have to check that the theorem written in Lean is the one people care about.

Which claim came first, Alpöge's or Masullo's?

Alpöge's note was posted on the morning of 20 November (US time), and Vakil's post about Masullo followed hours later, according to Scientific American. Vakil says Conrad had been reading Masullo's work for months before that.

Does this win the $1 million prize?

Not yet. Neither proof has been verified or peer reviewed.

Written on 21 November 2026, one day after the announcements. This is a developing story; we will update this page when the checks finish.