Technology•

Anthropic Mathematician Says Claude Helped Prove the 167-Year-Old Riemann Hypothesis, a Day After the IPO

Levent Alpöge says he and Claude found a proof of the Riemann hypothesis, posted with a Lean archive on Friday. Independent checks are pending, and the timing has raised questions.

Written By :MINRK
Anthropic Mathematician Says Claude Helped Prove the 167-Year-Old Riemann Hypothesis, a Day After the IPO

Photo by Thomas T on Unsplash

1. A proof claim arrives with Claude's name on it

Levent Alpöge, a mathematician at Anthropic, said on Friday, November 20, that he and the company's Claude AI model have found a proof of the Riemann hypothesis, a 167-year-old problem that carries a $1 million prize. The Claude Riemann hypothesis claim, posted one day after Anthropic's Nasdaq debut, has not been independently verified.

Alpöge set out the claim in a four-page note in which he wrote that he believes he and Claude have arrived at a proof. He went further than the original conjecture: according to the note, the same method also settles the Generalized Riemann Hypothesis, a broader statement that carries the original question over to a much larger family of mathematical functions.

If the argument holds, it would be one of the most significant results in the history of mathematics, with an AI system playing a central role in the work. That is a very large if. A proof of this kind is not accepted because its author announces it. It is accepted only after other experts have worked through it line by line and found no gaps, a process that for major results routinely takes months or years.

For now, what exists is a claim, a short explanatory note and a machine-checkable archive. The mathematical community has not yet delivered a verdict, and the circumstances of the release have drawn nearly as much attention as the mathematics itself.

2. What is the Riemann hypothesis?

The Riemann hypothesis is a prediction about where certain zeros of the Riemann zeta function lie, and because those zeros are tied to how prime numbers are spread out along the number line, it is one of the most important open questions in mathematics. The German mathematician Bernhard Riemann put it forward in 1859.

Prime numbers, the whole numbers divisible only by one and themselves, look scattered almost at random: 2, 3, 5, 7, 11, 13 and onward, with gaps that grow and shrink unpredictably. Yet across very large ranges they thin out in a remarkably regular way. Riemann showed that the fine detail of that pattern is encoded in the zeta function, and in particular in the locations of its so-called nontrivial zeros. His conjecture is that all of those zeros sit on a single vertical line in the complex plane.

Mathematicians have tested the prediction for an enormous number of zeros and never found an exception, but no amount of numerical checking counts as a proof. Generations of researchers have attacked the problem for well over a century without success, which is part of why a credible proof would be regarded as a historic event.

The question also matters far beyond pure mathematics. A great many results in number theory are written in conditional form, beginning with the assumption that the Riemann hypothesis, or its generalized version, is true. A confirmed proof would turn those conditional statements into established theorems in a single stroke. Prime numbers also sit at the heart of widely used public-key cryptography, which is one reason the problem draws interest from people who work on digital security and blockchain systems.

3. How Alpöge and Claude packaged the argument

Rather than relying only on a written paper, Alpöge and Claude encoded the argument in Lean, a software system known as a proof assistant that checks a proof one logical step at a time. In a proof assistant, every deduction has to be spelled out in a formal language, and the software refuses to accept any step that does not follow strictly from what came before.

The note points readers to a proof archive split into 14 downloadable parts. Making the archive public is an invitation for others to run it, inspect it and test whether it does what its authors say.

Formal verification has become an increasingly important tool as AI systems produce long and intricate mathematical arguments. A human reader can miss a subtle error buried on page 80 of a dense manuscript, while a proof assistant will not let it pass. For that reason, a Lean archive is a stronger form of evidence than a written claim alone.

Even so, publishing the files is not the same as the mathematical community accepting the result. Lean can confirm that each formal deduction obeys its rules. It cannot, on its own, guarantee that the statement being proved is exactly the famous conjecture that mathematicians care about.

4. A Sunday message to Dario Amodei and a Friday release

The timeline Alpöge described has become part of the story. He said he contacted Anthropic chief executive Dario Amodei about a possible announcement late on Sunday night, November 15, four days before the company's shares began trading.

After that message, Alpöge said, he consulted the company's leadership and was told there was a constraint that had to be respected. He did not say what that constraint was.

The claim was ultimately made public on Friday, the day after the listing. Alpöge acknowledged that the timing looks bad, while pushing back against a cynical interpretation of it. He did not state outright that the IPO was the reason the announcement was held back.

Anthropic began trading on Nasdaq on Thursday, November 19, after raising $81.34 billion in its offering. Companies preparing to go public typically face strict limits on what they can say in the period around a listing, and any major announcement that could move investor sentiment draws scrutiny. Alpöge did not tie his description of the constraint to those rules, so the reason for the delay remains unexplained. The result is that a potential landmark in mathematics arrived on the same news cycle as one of the most closely watched stock debuts of the year, a coincidence that has invited questions about the release's unusual timing.

5. The 'mathocalypse' debate that came before

Alpöge is not new to controversy over AI and mathematics. In September, he was caught up in a dispute that followed OpenAI's claim to have solved the Navier-Stokes problem, another of the famous unsolved questions that carry a $1 million prize, concerning the equations that describe how fluids move.

At the time, Alpöge and New York University mathematician Tristan Buckmaster were working on related research into fluid equations, and the overlap led to a disagreement over who deserved credit and which work came first.

In October, OpenAI released 722 manuscripts produced by an unreleased model, sorted into 372 families of results. Only 162 of those included main results that had been checked in Lean. Mathematicians responded by asking for stronger evidence before accepting the output as genuine progress.

Those episodes fed a wider anxiety in the field that some have nicknamed the 'mathocalypse': the fear that AI systems will knock down the discipline's great problems before human researchers have the chance to learn from the long struggle of working on them. Toronto mathematician Daniel Litt has warned that mathematics, as a human pursuit, could largely disappear within 50 years if the profession does not adapt. Buckmaster took a different view at an MIT panel this week, arguing that mathematics still depends on human understanding and cannot be reduced to producing correct outputs.

6. Why the Riemann hypothesis claim is not settled

The Riemann hypothesis claim is not settled because independent mathematicians have not yet confirmed it, and Lean's checks alone do not answer every question that matters. Several steps still lie ahead before anyone can treat the problem as solved.

First, other researchers need to reproduce the archive, running the 14 parts themselves to confirm that the formal proof compiles and checks as described. Second, and just as important, experts have to confirm that the theorem formalized in Lean actually matches the original conjecture. In formal mathematics, a small mismatch in how a statement is defined can mean that a proof establishes something subtly different, and far less important, than what was intended.

Third, the mathematics has to be understood. A proof of a problem this central is expected to come with ideas that specialists can follow, extend and apply elsewhere. Researchers will want to examine the method closely, especially since the note says it reaches as far as the generalized version of the conjecture.

Anthropic has previously reported progress on questions involving the zeta function, but it did not claim to have solved the full hypothesis. Alpöge's note goes much further than anything the company has said before, which is one reason caution is warranted until outside experts weigh in.

7. The $1 million Millennium Prize and its rules

The Clay Mathematics Institute includes the Riemann hypothesis among its Millennium Prize Problems, a set of famous open questions that each carry a $1 million award. The institute introduced the prizes in 2000, and the Riemann hypothesis was already one of the most celebrated unsolved problems long before then.

Winning the money is not a matter of posting a solution. Under the institute's rules, a proposed solution must appear in a qualifying publication, must then survive at least two years of scrutiny, and must win general acceptance in the mathematical community.

That means that even in the best case for Alpöge and Claude, any formal recognition is years away. It also raises a question the prize rules were never designed to answer: how credit should be assigned when an AI system played a central part in producing a proof.

For now, both the verification and any prize recognition remain pending.

8. What happens next

The immediate next step lies with independent mathematicians and formal verification specialists, who are expected to download the archive, check whether it runs as described and scrutinize whether its central theorem truly captures the Riemann hypothesis. No date has been set for any formal review, and the Clay institute's process would only begin once a qualifying publication appears.

For Anthropic, the announcement lands in its first days as a public company, when investors are paying close attention to anything that bears on the capability of its models. A verified proof would be a powerful demonstration of what Claude can do. A claim that falls apart under examination would add to the skepticism that followed earlier AI mathematics announcements across the industry.

Either way, the episode sharpens the questions the field has been wrestling with all autumn: what counts as evidence when machines produce proofs, who deserves credit, and how fast the profession needs to adapt. Until outside experts deliver their verdict, the claim that Claude helped crack the Riemann hypothesis remains exactly that: a claim.

Related Articles

NEWSLETTERS

Don't miss another story.

Subscribe to the MINRK Newsletter today.

Tick the newsletters you want. No selection? You'll get MINRK Headlines.

By signing up, you will receive emails about MINRK products and you agree to our terms of use and privacy policy.

MINRK Daybook

Market analysis for crypto traders and investors.

EVERY WEEKDAY

MINRK for Advisors

Defining crypto, digital assets and the future of finance for financial advisors.

EVERY THURSDAY

MINRK Protocol

Exploring the tech behind crypto one block at a time.

WEEKLY

MINRK Long & Short

A must read for institutions. Insights, news and analysis delivered weekly.

EVERY WEDNESDAY

MINRK Headlines

The biggest crypto news and ideas of the day.

EVERY WEEKDAY

MINRK Policy

Examining the intersection of cryptocurrency and government.

WEEKLY

MINRK Research

Join thousands of readers who rely on MINRK for data-driven insights on the latest digital asset trends.

MONTHLY