# Claude and the Riemann Hypothesis: What Alpöge Claims

Canonical: https://ournews.ai/news/claude-riemann-hypothesis-alpoge-proof-claim?lang=en
Published: 2026-11-21T02:07:15.719247+00:00
Updated: 2026-11-21T04:51:43.439549+00:00

*Cover: Levent Alpöge presents an earlier research project in an archival photograph from his personal website. The photograph does not depict the Riemann work. Photo: Nick Knupffer.*

**Has Claude helped prove the Riemann hypothesis?** Anthropic mathematician Levent Alpöge says it has. On November 20, he released a claim and linked the supporting Lean proof. Mathematicians are working to verify the argument; the result is not yet established. [Read Alpöge’s original note](https://alpo.ge/remark.pdf) · [Terence Tao’s assessment](https://terrytao.wordpress.com/2026/11/20/unforced-euler-singularity-unforced-navier-stokes-singularity-and-the-generalized-riemann-hypothesis/)

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

His four-page note claims the broader Generalized Riemann Hypothesis as well. The supporting files come in fourteen downloads. The public account names Claude but does not identify Fable 5.2 as the model used.

**What is the Riemann hypothesis, and why does it matter?**

Prime numbers—2, 3, 5, 7 and so on—are the building blocks of whole numbers. Mathematicians know how many primes to expect on average. The Riemann hypothesis would tightly constrain how far the actual count can stray from that average.

The precise claim concerns the zeros of the Riemann zeta function: all its nontrivial zeros should have real part one-half. Riemann proposed it in 1859. Checking trillions of examples is still different from proving that every one must obey the rule. [Clay Mathematics Institute’s explanation](https://www.claymath.org/millennium/riemann-hypothesis/)

**What Claude contributed**

Alpöge describes guiding the work through mathematical ideas before Claude supplied steps that outpaced his own understanding. He says they completed a formalization in Lean and plans to produce a clearer written explanation. His timeline ends with the generalized result on November 16. [Alpöge’s account](https://alpo.ge/remark.pdf)

A Lean proof is written so software can check its logical steps. That gives other researchers something concrete to inspect. They still need to establish that the code proves the intended mathematical statement and understand how the argument works. [Lean’s explanation of formal verification](https://lean-lang.org/)

**What Terence Tao says about verification**

Terence Tao describes the submission as an enormous Lean-only proof produced using an internal AI model. He points readers to the Lean community’s work to verify it and extract its strategy, warning that understanding the argument will take time. His initial description connects it to the Langlands program and arithmetic geometry, major areas of modern number theory. [Tao’s November 20 assessment](https://terrytao.wordpress.com/2026/11/20/unforced-euler-singularity-unforced-navier-stokes-singularity-and-the-generalized-riemann-hypothesis/)

The claim arrived one day after Anthropic’s stock-market debut. Alpöge addresses the awkward timing himself: he says he contacted Dario Amodei about announcing the result and faced a constraint on when he could do so. [His note, page 2](https://alpo.ge/remark.pdf#page=2)

For now, there is a specific claim and a released argument for mathematicians to examine. Whether it settles the 167-year-old problem depends on that examination.


Source: https://alpo.ge/remark.pdf
