---
title: "Riemann Hypothesis Proof Made With Claude Passes Lean Checks"
description: "An Anthropic mathematician says he proved the Riemann hypothesis with Claude. The Lean code checks out; a readable proof and prize review are still to come."
url: https://publishmyblog.com/ai/riemann-hypothesis-proof-claude/
section: "AI"
author: "Saeed Shahriar"
published: 2026-11-22T10:31:31.663Z
---

# Riemann Hypothesis Proof Made With Claude Passes Lean Checks

The Lean and Mathlib communities say the statement and the proof are correct, but a human-readable version is still months away.

Levent Alpöge, a mathematician at Anthropic, says he's proved the Riemann hypothesis, the most famous open problem in maths, by working with Claude.

**The Lean code behind his Riemann hypothesis proof has passed independent checks.** Bernhard Riemann posed the problem in 1859.

Do you need number theory to care? No.

If you use Claude for your writing, your code or your client reports, I think this is the clearest sign yet of what it can do when an expert steers it for months.

Alpöge announced it on Friday, 20 November, [Scientific American reports](https://www.scientificamerican.com/article/anthropic-mathematician-claims-to-prove-riemann-hypothesis-most-famous-problem-in-all-of-math/). He says it also covers the Generalized Riemann Hypothesis, a much wider version.

## How the Riemann hypothesis proof was checked

So how would you check a proof no one could read in full?

He released the proof itself only as Lean code, where a computer checks each logical step. **Printed, the Riemann hypothesis proof would run past 250,000 pages**, he says, [The Decoder reports](https://the-decoder.com/hours-after-anthropic-claims-a-claude-proof-of-the-riemann-hypothesis-stanford-releases-an-all-human-competing-draft/).

Then the checks came in, fast:

- The Lean and Mathlib communities said in a joint statement that both the statement and the proof are correct, [per Scientific American's follow-up](https://www.scientificamerican.com/article/the-ai-riemann-hypothesis-proof-is-almost-certainly-true-independent-mathematicians-attest/).
- MIT's Andrew Sutherland compiled it himself without issues, and put the chance of anything fishy at "approximately zero".
- A comparator run, which looks for tricks like a redefined symbol in the statement, has now been completed.

Still, a computer check isn't the same as understanding, and Alpöge admits Claude was far ahead of him in some areas.

He plans a shorter, human-readable version over the coming months.

Which Claude was it? The reports don't name a model version, so don't assume it's the one in your chat window today.

## A human proof landed the same day

Meanwhile, Stanford professor Ravi Vakil posted a draft proof of the Generalized Riemann Hypothesis by Alessandro Masullo, [TechCrunch reports](https://techcrunch.com/2026/11/20/ai-claimed-a-solution-to-the-167-year-old-riemann-hypothesis-and-a-human-did-too/). **Vakil called it "all human, with no AI".**

That draft isn't in Lean yet.

The timing is awkward for Anthropic too, since the claim came a day after [Anthropic's IPO](https://publishmyblog.com/ai/anthropic-ipo-lower-valuation/).

"The precise timing is horrible," Alpöge wrote, [TechCrunch reports](https://techcrunch.com/2026/11/20/ai-claimed-a-solution-to-the-167-year-old-riemann-hypothesis-and-a-human-did-too/). He rejects any idea of a PR stunt, according to The Decoder.

## What we checked

On 22 November we opened the [Clay Mathematics Institute's Riemann Hypothesis page](https://www.claymath.org/millennium/riemann-hypothesis/). **It still says "Unsolved".**

That's expected. Clay's [Millennium Prize rules](https://www.claymath.org/millennium-problems/rules/) need three things before a solution is even considered:

- publication in a qualifying outlet
- at least two years since that publication
- general acceptance among mathematicians

We also read [Vakil's own post](https://math216.wordpress.com/2026/11/20/hurried-announcement-alessandro-masullo-grh/). About eight hours in, he added that people had reported problems in Masullo's draft, found with help from AI models.

## Why it matters

You won't run your blog differently on Monday.

Scientific American says the result has no immediate real-world use.

But the method is worth copying. Claude took over near the finish, Alpöge says, and **a machine checker let experts back the output** within about a day.

My read: AI output is only worth what you can check, whoever made it.

In practice, your checker is a test suite you run before every deploy, a site crawl, a GA4 report you trust or the source link behind each claim in a post.

**If you can't check a claim, it doesn't go on your site.**

Though I'd expect every AI lab to quote this proof for months. When a pitch does, ask one thing.

What checks your output?

## Common Questions

**Has the Riemann hypothesis been proved?**

A Riemann hypothesis proof made with Claude has passed Lean checks: the Lean and Mathlib communities said in a joint statement that both the statement and the proof are correct. Alpöge plans a shorter, human-readable version over the coming months, and Clay still lists the problem as unsolved.

**Will Alpöge get the $1 million Millennium Prize?**

Not soon. Clay has put a $1 million bounty on the problem, Scientific American notes, but its rules need publication in a qualifying outlet, at least two years since that publication, and general acceptance among mathematicians, so even if every check you read about keeps passing, Clay can't consider it before late 2028.

**Which Claude model was used?**

The reports we read don't name a model version. Alpöge's quoted statement says only that he was working with Claude, so don't assume it's the one in your chat window today.

## Sources

- [AI claimed a solution to the 167-year-old Riemann hypothesis, and a human did, too](https://techcrunch.com/2026/11/20/ai-claimed-a-solution-to-the-167-year-old-riemann-hypothesis-and-a-human-did-too/), TechCrunch
- [Anthropic mathematician proves Riemann hypothesis, most famous problem in all of math](https://www.scientificamerican.com/article/anthropic-mathematician-claims-to-prove-riemann-hypothesis-most-famous-problem-in-all-of-math/), Scientific American
- [The AI Riemann hypothesis proof is true, independent mathematicians confirm](https://www.scientificamerican.com/article/the-ai-riemann-hypothesis-proof-is-almost-certainly-true-independent-mathematicians-attest/), Scientific American
- [Hours after Anthropic claims a Claude proof of the Riemann hypothesis, Stanford releases an all-human competing draft](https://the-decoder.com/hours-after-anthropic-claims-a-claude-proof-of-the-riemann-hypothesis-stanford-releases-an-all-human-competing-draft/), The Decoder
- [Riemann Hypothesis](https://www.claymath.org/millennium/riemann-hypothesis/), Clay Mathematics Institute
- [Rules for the Millennium Prize Problems](https://www.claymath.org/millennium-problems/rules/), Clay Mathematics Institute
- [Hurried Announcement: Alessandro Masullo GRH](https://math216.wordpress.com/2026/11/20/hurried-announcement-alessandro-masullo-grh/), Math 216
