Home · Posts · AI news
AI news

Riemann hypothesis AI proof: what was claimed and what is checked

By wiredsmarter · · 2 min read

Riemann hypothesis AI proof: what was claimed and what is checked

A Riemann hypothesis AI proof was posted this week. TechCrunch reported on Nov 20, 2026 that an Anthropic mathematician announced, in an essay, a solution to the Riemann hypothesis that he found working with Claude. Many videos already call the problem solved. What is confirmed so far is narrower: the proof is written so a computer can check it, and outside mathematicians say that check passes. Nobody has a human explanation of it yet.

What the Riemann hypothesis is

The hypothesis dates back to 1859. It says there is a hidden structure behind where prime numbers appear, and it is one of the Millennium Prize problems. Many results in number theory assume it is true, so a proof would matter well beyond one paper.

What was posted, and how it was checked

The proof is written in Lean, an open source language that expresses mathematics as code so a computer can verify every step. According to TechCrunch, several mathematicians in the public Lean discussion, including MIT mathematician Andrew Sutherland, say they compiled the proof successfully. The article was later updated to say independent sources had checked it.

A Lean check confirms that the steps follow from the statement as written. It does not explain the idea behind the proof. The author wrote that he felt uneasy releasing it without a version people can follow, and he has committed to write one. Until that exists, mathematicians can confirm the code runs but can't yet say they understand why it works.

A human proof the same week

On the same days, a Stanford professor shared a separate draft proof of the generalized Riemann hypothesis by a Stanford postdoc, the result of 12 years of work. He wrote that "it is all human, with no AI." That draft is not in Lean, so there is no computer check yet, and it is too early to say whether it holds.

How this differs from August

In August, Anthropic said a research version of Claude raised a known lower bound for the share of zeta function zeros on the critical line from 41.6% to 67.2%. Anthropic said then that it did not expect those techniques to prove the hypothesis itself. This week's claim is a full proof, which is why the checking matters more.

What to watch next

The next step is a written explanation from the author and review by number theorists. Expect weeks or months before the mathematics community says it understands the result. Headlines that call it settled today are ahead of the evidence.

FAQ

Did AI solve the Riemann hypothesis?

A proof found with Claude was posted and its Lean code compiles, according to outside mathematicians. Experts still need to understand and review it before calling it solved.

What is Lean in math?

Lean is an open source programming language for writing proofs as code. A computer then checks that every step is valid.

Is there a human proof too?

Yes. A Stanford postdoc's draft, made without AI, was shared the same week. It has no computer check yet.

Sources

Tools in this article

Claude No commission

Anthropic's AI assistant: chat, files, artifacts and the Claude Code app. Free plan; paid plans start with Pro.

Open Claude

We earn nothing from these links. All tools

Making something to sell?

Our guide shows how to make and sell one small PDF with AI, with the real numbers. Chapter 1 is free.

Follow wiredsmarter for the next one

Read next

Checked against official sources on Nov 21, 2026. The character in our posts is made with AI; the text is written with AI help and checked by hand rules.