2026-11-20 · ← News
Claude helped produce a claimed Riemann proof. The verdict will take years
Anthropic mathematician Levent Alpöge says he and Claude proved the Riemann and Generalized Riemann Hypotheses and has released a Lean formalization. The result has not been independently verified, while the million-dollar prize requires at least two years of scrutiny.
The image could not be loaded.
Levent Alpöge, a mathematician working at Anthropic, has claimed that he found proofs of the Riemann Hypothesis and its generalized form with Claude. Instead of a conventional paper, he released a Lean formalization in a fourteen-part archive and a short note describing the route.
Alpöge released a Lean artifact, not an accepted result
The Riemann Hypothesis dates to 1859 and is one of the seven Clay Mathematics Institute Millennium Prize Problems. It concerns zeros of the Riemann zeta function and is deeply connected to the distribution of primes. An accepted solution carries a $1 million prize.
Alpöge writes that Claude moved far ahead of him near the end of the work. He also says the method reaches the Generalized Riemann Hypothesis. His note is not itself a full readable proof. He says the current human-facing output would exceed 250,000 pages and offers the Lean code as the checkable artifact.
Formalization turns the first review into an artifact check
Lean can mechanically verify whether formal steps follow the rules of its logical kernel. That can accelerate the first examination: experts can test whether the archive builds the declared theorems without hidden assumptions and whether the formal statement matches the hypothesis.
A successful build alone would not explain the mathematical idea, the origin of each step or any gap between the informal claim and the exact statement. The field will still need a readable paper and independent reconstruction of the argument.
The million-dollar process does not begin with a downloaded archive
The Clay Mathematics Institute requires publication in a qualifying outlet, general acceptance by the mathematics community and at least two years of rigorous examination. A website note and Lean archive do not satisfy those conditions by themselves. Anthropic has not joined the announcement according to current reporting, and the result remains independently unverified.
Early headlines therefore run ahead of the evidence. The precise description is that a respected mathematician has made an extraordinary claim and provided a formal artifact. It is too early to describe the proof as an accepted mathematical result.
Reproduction comes first, community acceptance much later
The immediate signal will be whether independent teams can download, build and confirm the exact theorem statements without unexpected assumptions. A readable account that specialists can judge, simplify and use must follow.
Only long-term expert agreement can open the Clay Institute process. If the formalization survives, it will be both a mathematical breakthrough and evidence that AI can actively shape an argument rather than merely translate it into formal language.
Lilith's verdict
There is no $1 million cheque on the table, only fourteen archives that other mathematicians must unpack without the author's hands on the keyboard. If they rebuild the same proof, more than one textbook will need rewriting.