The AI Assistance Behind An Anthropic Mathematician’s Riemann Hypothesis Claim
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: The AI Assistance Behind An Anthropic Mathematician’s Riemann Hypothesis Claim on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

Anthropic mathematician Levent Alpöge posted a note on Nov. 20 saying he believes he and Claude found a proof of the Riemann Hypothesis and the stronger Generalized Riemann Hypothesis. The proof is a 17.4-million-line Lean archive that has not been independently verified or accepted; early checks are encouraging, but a key test remains pending.

Levent Alpöge, a mathematician who works at Anthropic, posted a note on Nov. 20 saying he believes he and the company’s Claude AI found a proof of the Riemann Hypothesis, one of mathematics’ best-known unsolved problems. The proposed proof has not been accepted: its central argument is contained in a large Lean code archive that mathematicians are still checking.

Alpöge’s four-page note says the approach also extends to the Generalized Riemann Hypothesis, a broader conjecture involving a family of related functions. The note is a sketch, not a conventional proof paper. Alpöge linked a 14-part Lean archive, a computer-checkable formalization of the argument. Computer scientist Scott Aaronson estimated the archive contains 17.4 million lines of Lean code. Alpöge says a human-readable version would exceed 250,000 pages and did not publish it.

The account of how the work was done comes from Alpöge. He says he guided the strategy while Claude handled much of the technical work, and wrote that “Claude far outpaced me near the finish line.” He also says he has “little mastery” of one of the areas the argument uses. The source material does not identify which Claude model was involved.

Initial technical signals are not the same as verification. MIT mathematician Andrew Sutherland told Scientific American that the code compiled independently “with only the three standard axioms.” Kevin Buzzard, however, wrote that a “comparator run” was still awaited. That check is intended to establish that the formalized code proves the intended mathematical statement rather than a different or weakened one. The proof’s mathematical content is not yet broadly understood.

At a glance
reportWhen: Claim posted Nov. 20, 2026; scrutiny on…
The developmentLevent Alpöge has published a claim that he and Anthropic’s Claude AI proved and formalized the Riemann Hypothesis, prompting initial scrutiny from mathematicians.

Explainer · Riemann hypothesis claim

From claim to accepted proof: six checks, and where this one stands

What the hypothesis says, and the path every proof claim has to take before mathematicians accept it.

The claim in one picture

0 1/2 1 real part

Riemann’s guess (1859): every non-obvious zero of the zeta function lies on this one line. If true, the primes are spread out as evenly as possible.

The checks a proof claim must pass

  1. 1Claim postedA 4-page note plus Lean code, November 20, 2026✓ done
  2. 2Code compiles elsewhereIndependent compile reported by an MIT mathematician✓ reported
  3. 3Comparator runRight statement? Only standard axioms? Several Lean kernels?… pendingWe are here · November 21, 2026
  4. 4Humans understand itA readable proof that experts can follow○ open
  5. 5Journal publicationA refereed mathematics journal of worldwide repute○ open
  6. 6Two years + general acceptanceWhat the Clay prize rules require before a $1 million award○ open

Sources: Clay Mathematics Institute (Riemann Hypothesis page; Millennium Prize rules); Levent Alpöge, remark.pdf (November 20, 2026); Scientific American, November 20, 2026 (independent compile; comparator run awaited); Kevin Buzzard, Xena Project, November 12, 2026 (the checks). Status as reported in these sources by November 21, 2026.

Graphic: Powered by Thorsten Meyer AI

Why a Verified Proof Would Matter

The Riemann Hypothesis concerns the distribution of prime numbers, which appear irregularly among the whole numbers. It makes a precise claim about the non-obvious zeros of the zeta function: their real parts all equal one-half. A proof would settle a central question in number theory and establish many results that currently depend on the hypothesis being true.

The scale of the claim makes verification especially important. A formal proof checker can test whether each encoded step follows under its rules, but specialists must still determine whether the code represents the intended argument and whether the underlying mathematics is sound and meaningful. Until that work is done, the announcement is a significant claim to examine, not a settled result.

The episode also draws attention to how AI might contribute to research. Alpöge attributes substantial technical work to Claude, but his account alone cannot establish what the system contributed, how independently the work can be reproduced, or whether the proof is correct. Those questions are relevant beyond this problem as mathematicians evaluate AI-assisted research.

Samsung 65-Inch 4K OLED Smart TV with Dolby Atmos, 120Hz Refresh Rate, and Amber Protection Bundle

Samsung 65-Inch 4K OLED Smart TV with Dolby Atmos, 120Hz Refresh Rate, and Amber Protection Bundle

Breville Barista Express Espresso Machine with Built-In Grinder, Brushed Stainless Steel

Breville Barista Express Espresso Machine with Built-In Grinder, Brushed Stainless Steel

As an affiliate, we earn on qualifying purchases.

The Claim and Its Formal Proof

Bernhard Riemann introduced the hypothesis in 1859. In the Clay Mathematics Institute’s description, it “asserts that all the ‘non-obvious’ zeros of the zeta function are complex numbers with real part 1/2.” The institute has listed it among its Millennium Prize Problems, each associated with a $1 million prize. The prize is not triggered by posting a proposed proof: Clay’s rules require publication in a refereed journal, at least two years of scrutiny, and general acceptance by mathematicians.

Computers have checked the hypothesis for the first 10 trillion solutions, according to Clay. Such checks cover many cases but cannot settle a statement about infinitely many zeros. The distinction matters here: the Lean archive may provide a way to check formal steps, but early compilation reports do not by themselves demonstrate that the whole claim has been verified.

Alpöge’s note says he “decided to only present Lean” rather than publish the very large human-readable document he says he produced. Terence Tao described the submission as an “(enormous) Lean-only proof.” The result is a file-based claim whose assessment depends on careful technical review, rather than a paper whose reasoning can readily be followed from start to finish.

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

— Levent Alpöge, in his note

The Proof Still Awaits Review

No journal has published the proof, and no mathematical body has accepted it. The comparator run reported as pending is one outstanding check; the source material does not give a timetable for its completion or a full account of the subsequent review process. A successful run would be evidence about the formalization, not a substitute for expert assessment of the claim.

It also remains unclear how the proof’s ideas will be evaluated by mathematicians who have not worked directly with the code. Andrew Granville called Alpöge’s released note “garbled,” while Tao characterized the submission as an enormous Lean-only proof. Anthropic had not announced the claim on its news page as of Nov. 21, and Nature reported that the company did not immediately respond to its request for comment. The model used and the precise division of labor have not been independently established in the supplied reporting.

Independent Checks and Mathematical Review

The next near-term step is to complete the reported comparator test and continue independent examination of the Lean archive. Reviewers will need to determine whether the code formalizes the Riemann Hypothesis as stated and whether the argument withstands mathematical scrutiny. Alpöge’s note does not set out a public schedule for those checks.

Any path to formal recognition would take longer. Under the Clay Mathematics Institute’s stated prize rules, a proposed proof must appear in a refereed journal, undergo at least two years of scrutiny, and gain general acceptance among mathematicians. Until those conditions are met, the claim remains under review and the Riemann Hypothesis remains an open problem in the accepted mathematical record.

Key Questions

Has the Riemann Hypothesis been proved?

Not as an accepted result. Alpöge says he believes he and Claude found a proof, but mathematicians are still checking the Lean archive and no mathematical body has accepted it.

What did Claude reportedly contribute?

Alpöge says he guided the strategy and Claude performed much of the technical work, including a step involving the length of a chain of theta lifts. That division of labor is Alpöge’s account; the model used has not been identified in the source material.

What is Lean, and why is it relevant?

Lean is a proof assistant that checks formally encoded mathematical steps against its rules. The archive’s reported successful compilation is an initial technical check, not by itself proof that the intended theorem has been established.

Does this claim qualify for the Clay prize?

Not on the information available. Clay’s rules require a proof to be published in a refereed journal, scrutinized for at least two years, and generally accepted by mathematicians before the $1 million prize can be awarded.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
You May Also Like

The Economics Of AI: Why GLM-5.3-Flash Is Gaining Attention

GLM-5.3-Flash, a 320B multimodal AI model, is attracting interest due to its performance, open release, and low-cost API for agent workflows.

Using Claude Meetings In Google Meet? Review These Points First

Anthropic’s Claude Meetings beta adds live notes, answers and documents to Google Meet. Review consent, plan defaults, access and retention questions.

Can One Memory Help AI Agents Collaborate?

Thorsten Meyer describes a shared, file-based memory for his AI agents, reporting 97.4% top-one recall accuracy as of Nov. 21, 2026.

Frontier Lab’s Innovation Journey: From Land To Energy With AI

Frontier Lab is expanding its focus from research to capacity building, including land, energy, and infrastructure, to support large-scale AI development.