Person looking up at a Nasdaq digital billboard displaying a group of people and the text "ANT NasdaqListed" in a cityscape.

Anthropic researcher says he and Claude proved the Riemann hypothesis, but no one has verified it yet

Anthropic researcher Levent Alpöge says Claude helped prove the Riemann hypothesis. The claim is unverified and under scrutiny.

In short

Anthropic researcher Levent Alpöge says he and Claude proved the Riemann hypothesis, but the claim has not yet been independently verified. The announcement has intensified scrutiny of AI-assisted mathematics and the race among labs to claim breakthroughs in famous open problems.

  • Levent Alpöge says he and Claude proved the Riemann hypothesis and the Generalized Riemann Hypothesis.
  • The claim has not yet been independently verified, and Anthropic has not commented.
  • Clarín says the proof archive includes 175,187 Lean files and more than 250,000 pages of argument.
  • The Riemann hypothesis is one of the seven Millennium Prize Problems with a $1 million prize.
  • Mint says OpenAI recently made a separate claim about solving Navier-Stokes using 10,000 AI agents.

Levent Alpöge, a mathematician on Anthropic’s technical staff, says he and Claude have found a proof of the Riemann hypothesis and the broader Generalized Riemann Hypothesis. The claim, posted publicly on November 20, 2026, has not yet been independently verified, which means the central question is not just whether the proof exists, but whether the mathematical community can confirm it.

What Alpöge says he found

Alpöge’s announcement is unusual even by the standards of AI research news: it is not about a model solving a benchmark or drafting code, but about one of mathematics’ most famous unsolved problems. According to The Verge, Alpöge said he discovered the proof with Claude and posted the supporting materials only on his own website so far.

He also said he texted Anthropic chief executive Dario Amodei late Sunday night about the result, and suggested he waited to publish until after Anthropic’s Thursday IPO. The Verge reported that Anthropic has not commented on the announcement.

“I believe I have, working with Claude, discovered a proof1 of the Riemann Hypothesis.”

The claim matters because the Riemann hypothesis sits in a rare category of mathematical celebrity. First proposed in 1859, it is one of the seven Millennium Prize Problems, each carrying a $1 million bounty from the Clay Mathematics Institute. But that prize is not awarded on announcement alone: the institute must accept the proof after publication, and the result must survive two years of scrutiny.

Why this claim is drawing attention

The first reason is obvious: if the proof holds, it would resolve one of the most famous open questions in mathematics. The second is more immediate to the AI industry. A verified proof would become another landmark in a year already marked by aggressive claims about what large language models and AI agents can do in specialized scientific work.

Clarín identified Alpöge as a 34-year-old Turkish-American mathematician on Anthropic’s technical staff and said he posted the claim on November 20, 2026. The same report said the proof archive includes Lean 4 verification code and has been under independent checking, though verification remains underway. That detail matters: Lean is a proof-assistant system used to formalize mathematical arguments, and formal checks can expose gaps that a human-written write-up might miss.

Still, the status is not the same as a completed mathematical victory lap. Independent verification has not happened yet, and the supporting files are not the same thing as broad consensus. In practice, the real test is whether other mathematicians can reproduce the result, inspect the Lean files, and decide the argument is sound.

How the proof is being presented

One reason the claim has drawn unusually close scrutiny is the scale of the materials involved. Clarín reported that the archive contains 175,187 Lean source files and totals 239 megabytes, while the full written argument reportedly exceeds 250,000 pages. Those numbers do not prove the theorem, of course, but they do signal that the claim is being packaged as a substantial formal project rather than a brief informal note.

According to Clarín, the Lean files compile against the standard math library, but proof verification remains underway. That is an important distinction. A file that compiles can still leave open questions about whether the proof actually establishes the theorem it claims to establish, especially in a problem as deep and historically contested as the Riemann hypothesis.

Alpöge also reportedly extended the claim to the Generalized Riemann Hypothesis, which would raise the stakes even further if confirmed. That broader statement is still subject to the same constraint as the original: no independent verification, no accepted result.

What is the Riemann hypothesis?

The Riemann hypothesis is a famous unsolved problem about prime numbers and the zeros of a complex function. It has fascinated mathematicians for more than a century and a half because it sits at the intersection of pure theory and the distribution of primes, one of the most studied patterns in mathematics.

The Clay Mathematics Institute included it in the Millennium Prize Problems in 2000, turning an already legendary question into one with a formal public challenge attached. The prize is $1 million per problem, but the money is secondary to the status: a verified proof would instantly become one of the most significant mathematical results of the era.

Background knowledge also helps explain why the result is so difficult to assess quickly. High-level mathematics is not validated by press release. It requires a proof that can be read, checked, and stress-tested by experts, often over long periods. That is especially true when the proof is said to involve AI assistance, because the community will want to know exactly which steps were generated by the model, which were formalized by the human researcher, and where the crucial insight actually came from.

Why it matters

If the claim is correct, the implications would reach well beyond one theorem. It would add a major proof to the record of AI-assisted science and mathematics, showing that frontier models may be useful not only for drafting text or code, but for navigating one of the hardest categories of abstract reasoning.

For readers, the immediate significance is less about the theorem itself than about the verification bottleneck. The story shows how the public conversation around AI increasingly turns on proof, auditability, and trust. In other words, the question is not just whether a model can suggest a path to truth; it is whether humans can independently certify that path.

For the market and the research community, the timing is also sensitive. If Anthropic is linked to a verified mathematical breakthrough, that would strengthen the company’s brand at a moment when prestige, technical credibility, and lab identity matter intensely. But if the claim stalls under scrutiny, it will become another reminder that ambitious AI headlines often outrun confirmed results.

How did this become part of an AI race?

Clarín described Alpöge’s announcement as part of a 2026 streak of AI-assisted math claims. Mint added a broader comparison point: in September, OpenAI claimed it had solved the Navier-Stokes existence and smoothness problem, another of the world’s toughest mathematical challenges.

According to Mint, OpenAI said it used up to 10,000 AI agents and millions of dollars of compute on the Navier-Stokes work. Mint also reported that OpenAI said the effort took 88 hours, used 130 billion output tokens, and involved 2.7 million messages sent by agents. Those figures do not validate the claim on their own, but they show how the scale of AI-assisted research is becoming part of the story.

Jakub Pachocki, OpenAI’s chief scientist, framed the result this way:

“We see this as a demonstration of just how far AI has gotten—and how quickly.”

That line captures the competitive undercurrent running through these announcements. Each claim is not only about mathematics; it is also a public signal about who can push AI systems furthest into expert domains.

Anthropic, for its part, has been more restrained. The Verge reported that the company did not comment on Alpöge’s announcement, and also cited a company statement saying, “We don’t expect that the techniques Claude used will lead to proving the Riemann hypothesis.” That puts a useful check on the hype: even inside the same organization, there appears to be caution about treating the result as settled.

Background and context

Why mathematicians care about verification

Mathematics has always relied on proof, but modern formal systems such as Lean add another layer of checking. They allow arguments to be encoded in a way that software can verify step by step. In principle, that can make a very complex proof more reliable, not less. In practice, it also raises a new question: can a massive formal artifact be trusted if only a few people can fully understand it?

That is the core tension in Alpöge’s claim. A proof that exists only as a website upload is not yet part of the mathematical canon. A proof that compiles in Lean is promising, but still not the same as broad acceptance. The community wants both rigor and clarity, and it will be watching closely for whether the materials can survive outside the original researcher’s own checking process.

How the Clay prize changes the stakes

The Millennium Prize Problems were created to spotlight major unsolved questions and attach a concrete reward to them. But the rules also emphasize patience. A proposed solution must be published and then endure at least two years of scrutiny before the Clay Mathematics Institute can consider it for the prize. That delay is deliberate: it protects the prize from being handed out on the basis of a premature announcement.

So even if Alpöge’s proof were ultimately accepted, there would still be a long institutional process ahead. The prize clock does not start with excitement; it starts with publication that can withstand a sustained review process.

Milestone What happened Source
1859 Riemann hypothesis first proposed The Verge
2000 Millennium Prize Problems established The Verge
Late Sunday night Alpöge says he texted Dario Amodei about the result The Verge
Thursday Alpöge says he waited until after Anthropic’s IPO to publish The Verge
November 20, 2026 Alpöge publicly posted the claim Clarín
September 2026 OpenAI claimed a Navier-Stokes breakthrough Mint

What the outlets agree on — and where they differ

The three reports line up on the basic facts: Alpöge says he and Claude found a proof, the claim has not yet been independently verified, and Anthropic has not publicly endorsed it. They also agree that the result, if real, would be a major mathematical event.

The differences are mostly in emphasis and detail. The Verge focused on the announcement, the lack of verification, and the prize context. Clarín added the most technical detail about the archive, the Lean files, and the reported scale of the write-up. Mint widened the frame by comparing the claim with OpenAI’s earlier Navier-Stokes announcement and the massive computational effort that accompanied it.

Those differences matter because they show how the story is evolving. One outlet frames it as a singular claim, another as a formalized archive, and another as part of a competition among AI labs to demonstrate scientific reach. Taken together, they suggest the headline is not just about a theorem; it is about a new kind of prestige contest.

What happens next?

The next phase is straightforward to describe, even if it may take time to unfold. Independent mathematicians will need to inspect the proof, including the Lean verification code, and determine whether the argument is valid. If the proof does not survive that process, the claim will fade into the long history of attempted solutions to famous problems. If it does survive, it could move toward formal acceptance and eventual Clay consideration.

There are several open questions to watch:

  • Will independent mathematicians verify the proof in Lean?
  • Will the Clay Mathematics Institute accept any submission tied to the claim?
  • Has Anthropic confirmed Alpöge’s account or offered technical comment?
  • Does the proof hold up under peer review and broader mathematical scrutiny?

For now, the most important fact is the least dramatic one: the claim exists, but the verdict does not. Until outside experts finish their review, Alpöge’s announcement remains a potentially historic mathematical breakthrough wrapped in an unresolved verification problem.

By the numbers

The figures attached to the claim help explain why it has spread so quickly across AI and science circles.

Number What it refers to
7 Millennium Prize Problems
$1 million Clay Mathematics Institute prize for each problem
2 years Minimum scrutiny period before prize consideration
175,187 Lean source files in the reported archive
239 megabytes Reported archive size
250,000+ pages Reported length of the written argument
10,000 AI agents OpenAI said it used on Navier-Stokes
88 hours Time OpenAI said it took to find a solution

Frequently asked questions

Has the Riemann hypothesis been proven by Claude and Anthropic researcher Levent Alpöge?

No. Alpöge says he and Claude found a proof, but the claim has not yet been independently verified. Until other mathematicians review the materials and confirm the result, it remains an unconfirmed announcement.

Why is the Riemann hypothesis such a big deal?

It is one of mathematics’ most famous unsolved problems and one of the seven Millennium Prize Problems. A verified proof would be a major breakthrough in number theory and would carry a $1 million prize from the Clay Mathematics Institute.

What evidence has Alpöge provided so far?

According to Clarín and The Verge, Alpöge posted proof materials on his website, including Lean 4 verification code. Clarín also reported that the archive contains 175,187 Lean source files and that the full written argument reportedly exceeds 250,000 pages.

Has Anthropic responded to the claim?

The Verge reported that Anthropic has not commented on Alpöge’s announcement. It also cited a company statement saying the techniques Claude used were not expected to lead to proving the Riemann hypothesis.

What happens if the proof is valid?

If the proof is valid, it would still need to be published and then survive at least two years of scrutiny before the Clay Mathematics Institute could consider it for the prize. The key hurdle is formal, independent verification.

Share this 🚀