Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)

Anthropic Researcher Claims He Proved the Riemann Hypothesis with Claude in Six Weeks

Anthropic mathematician Levent Alpöge says he and Claude have proved the Riemann Hypothesis, the 1859 problem for which the Clay Institute offers $1 million. Lean works like a compiler for mathematics: it checks a proof step by step. Alpöge has published that Lean record, but the proof text itself remains closed. For now, this is only a claim. If the code passes verification, the approach of "expert poses the problem, AI searches for steps, machine checks them" would become a working model.

AI-processed from TG: @hiaimedia (Hi, AI • Новости технологий); edited by Hamidun News
Anthropic Researcher Claims He Proved the Riemann Hypothesis with Claude in Six Weeks
Source: TG: @hiaimedia (Hi, AI • Новости технологий). Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceTG: @hiaimedia (Hi, AI • Новости технологий) ↗
Editor-in-Chief
Key metric #1The work ran for about six weeks, from early October to November 16
Key metric #2The Clay Prize is not awarded until at least 2 years after publication
Key metric #3In September, OpenAI said it had solved the Navier–Stokes problem, and that claim is under review

Imagine writing a program of a million lines and having it checked not by a reviewer but by a compiler with tests: if everything builds, the logic has no errors. This is roughly how Levent Alpöge, a mathematician at Anthropic, proposes that his claim be checked. He says that with Claude he proved the Riemann Hypothesis, one of the most famous unsolved problems in mathematics. For now, it is only a claim, and there are reasons to examine it carefully.

What is claimed and what is not yet available

Alpöge has posted a note on his website asserting that he proved the Riemann Hypothesis with Claude. According to him, the same method also proves the generalized Riemann Hypothesis. He worked on the problem for about six weeks, from early October to November 16.

Alongside the note, he released a formalization in Lean. Lean is a language in which mathematical statements are written so that a computer can verify them. If a proof is written in Lean and the code passes verification, the machine confirms that each step follows from the previous ones. It is like green tests in CI: a person does not have to take the author's word for it.

But the text of the proof in a readable form that can be studied and discussed is not yet available. There are also no details about which specific model was used or how many resources the work consumed. Alpöge admits that the hypothesis occupied him for many years. Even as a student, he kept quiet that his plans were tied to this problem for decades ahead, and he set out to master every tool that might someday prove useful.

Why mathematicians care about the Riemann Hypothesis

Bernhard Riemann formulated it in 1859. It states that all nontrivial zeros of the Riemann zeta function ζ(s) have real part equal to 1/2. In everyday life, a proof would change almost nothing. For mathematics, the consequences are enormous: the hypothesis determines how precisely the distribution of prime numbers can be described.

A good way to grasp the scale is to think of a library running on a "temporary patch." A whole body of modern work rests on the assumption "suppose the Riemann Hypothesis is true...". It is like code that works until you update a dependency. If the proof holds up, all those conditional results would become unconditional.

The hypothesis is one of the seven Millennium Prize Problems, for whose solution the Clay Institute has set a $1 million prize. So far only one has been solved: the Poincaré conjecture, proved by the Russian mathematician Grigori Perelman, who declined the prize. In September, OpenAI said that an internal model of theirs had solved another such problem, concerning the existence and smoothness of solutions to the Navier–Stokes equations. That result is still being checked.

What this means in practice

No quick verdict should be expected. The Clay Institute awards the prize only if the solution has been published in a recognized scientific journal, at least two years have passed since publication, and the community as a whole has accepted the result. Mathematicians around the world will most likely start examining Alpöge's note now, but the process will be long.

For those building products on AI, there are two lessons here.

The first concerns automated verification. The most convincing AI results appear where the output can be run through a strict mechanism: a compiler, tests, a formal verifier. If your agent has such a "judge," its work is easier to trust. The remaining question is whether the judge checks the right thing: one has to confirm that the formalized statement really matches the Riemann Hypothesis and that the code contains no extra assumptions.

The second lesson concerns the division of roles. An expert who had spent years preparing posed the problem, and Claude, according to the description, worked as a tool in his hands. For teams this is a useful benchmark: a strong specialist plus a model plus automated verification delivers more than any of the three alone.

Caveats

At this point there is neither a full text, nor independent expert review, nor data on the model and the computation spent. The claim comes from an Anthropic employee, that is, from a company that has an interest in showcasing Claude's capabilities. This does not mean the result is wrong, but caution is warranted. The Navier–Stokes example shows that after a loud announcement, verification can drag on for months.

If the proof survives scrutiny, the question of who gets the million dollars, the mathematician or Anthropic, will be the least of the problems. The academic community will have to develop rules for work found together with AI.

Story

Key Questions Answered

What is claimed and what is not yet available?

Alpöge has posted a note on his website asserting that he proved the Riemann Hypothesis with Claude. According to him, the same method also proves the generalized Riemann Hypothesis. He worked on the problem for about six weeks, from early October to November 16. Alongside the note, he released a formalization in Lean. Lean is a language in which mathematical statements are written so that a computer can verify them.

Why mathematicians care about the Riemann Hypothesis?

Bernhard Riemann formulated it in 1859. It states that all nontrivial zeros of the Riemann zeta function ζ(s) have real part equal to 1/2. In everyday life, a proof would change almost nothing. For mathematics, the consequences are enormous: the hypothesis determines how precisely the distribution of prime numbers can be described. A good way to grasp the scale is to think of a library running on a "temporary patch." A whole body of modern work rests on the assumption "suppose the Riemann Hypothesis is true...".

What this means in practice?

No quick verdict should be expected. The Clay Institute awards the prize only if the solution has been published in a recognized scientific journal, at least two years have passed since publication, and the community as a whole has accepted the result. Mathematicians around the world will most likely start examining Alpöge's note now, but the process will be long. For those building products on AI, there are two lessons here. The first concerns automated verification.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…