Anthropic Mathematician Says Claude Helped Prove the Riemann Hypothesis, $1 Million Prize at Stake
Anthropic mathematician Levent Alpöge says he and Claude proved the Riemann Hypothesis, the 1859 puzzle about the pattern in the distribution of prime numbers. The proof is written in Lean, which works like a compiler with tests for mathematics: it checks every step on its own. But the proof currently sits on Alpöge's personal website, and the Clay Institute, which offers a $1 million prize, requires at least two years of review by the community. For now, this is a claim, not a recognized result.
AI-processed from The Verge; edited by Hamidun News
Imagine an AI assistant sends you a huge pull request of thousands of lines, and you can't tell whether it works. That's roughly the situation mathematicians now face. Anthropic employee Levent Alpöge wrote on his website that he and Claude found a proof of the Riemann Hypothesis, one of the most famous unsolved problems in mathematics. No one has independently verified it yet, and Anthropic has not publicly commented on the claim.
What the problem is and why it matters
Prime numbers are 2, 3, 5, 7, 11, and so on: numbers divisible only by one and themselves. At first glance they appear chaotic, sometimes clustered together and sometimes separated by large gaps. The Riemann Hypothesis, formulated in 1859, asserts that this chaos hides a strict order, and it ties that order to the zeros of a special function. No one has been able to prove it in more than 160 years. It is one of the seven "Millennium Prize Problems": the Clay Institute offers $1 million for a solution to each. In the quarter century since the prize was established, only one of these problems has been officially solved: the Poincaré conjecture.
Alpöge says he worked together with Claude. According to him, the method extends easily: he claims it closes the generalized Riemann Hypothesis and, possibly with more time, extends further. He also says that he and Claude formalized the arguments without any additional assumptions.
What Lean is and why it is a strong argument
The proof is published in Lean. In programmer terms, Lean is a compiler and test suite for mathematics. Every step of reasoning is written out rigorously, and the program checks by itself that each step follows from the previous ones. A mistake of the kind "here the author quietly skipped a step" gets caught automatically.
There is a caveat. Lean confirms that the conclusion holds for the statement as written in it. That means mathematicians still have to check that the problem itself and all definitions are correctly encoded in Lean. It's like tests that are all green but check the wrong requirement.
The work has no formal recognition yet. Under the Clay Institute's rules, a solution must be published and must withstand at least two years of scrutiny, earning "general acceptance in the global mathematical community." For now, the proof exists only on the author's personal website.
According to Alpöge, he wrote to Anthropic CEO Dario Amodei "late on a Sunday evening" and postponed publication until the company's IPO on Thursday. In August, Anthropic said it was working on the Riemann Hypothesis with Claude, but at the same time added that it did not expect the techniques the model used to lead to a proof.
Context: the race of mathematical claims
This is not the first claim of its kind. In September, OpenAI announced a solution to the Navier–Stokes problem on fluid flow. According to The Verge, the company spent millions of dollars and ran 10,000 agents because it had heard rumors that competitors were close to a solution. Many consider the result convincing, but it has no formal recognition either. In addition, OpenAI has been accused of questionable research practices, including using unpublished work from mathematicians who shared ideas with its AI tools. The company categorically denies this.
Alpöge previously worked on Navier–Stokes with New York University professor Tristan Buckmaster. Buckmaster claims that OpenAI, using Codex and Claude, arrived at a solution via a route that he and Alpöge had been pursuing. And in July, Alpöge said he had solved the Jacobi conjecture with Claude.
What a developer should take from this
First: the most reliable way to work with AI on tasks where errors are costly is "the model proposes, an independent verifier checks." In mathematics that is Lean; in code it is tests, types, and review; in finance and legal texts it is independent calculation and cross-checking.
Second: even a green check doesn't settle what was actually checked. Tests have to be written against the right requirement.
Third: AI labs' scientific results are becoming part of competition and reputation. Publication timing tied to an IPO and disputes over authorship illustrate this well.
For now, neither mathematicians nor the Clay Institute have confirmed the claim. If the verification passes cleanly, it will be an event of decade-scale significance. If not, it will be another example of loud announcements outrunning verification. Wait for analyses from independent mathematicians, not press releases.
- Nov 21, 2026A Developer Had Claude Install 900 Apps and Built a Free Mac Cleaner
- Nov 21, 2026Apple Is Building an AI Agent and Making It Harder for Outsiders to Get Access: Permission Now Takes 14 Steps, Not 4
- Nov 21, 2026Qwen3-0.6B Fine-Tuned on a Mac mini: Errors Dropped to 3 out of 30 from 9
- Nov 21, 2026The Programmer Stopped Being an AI Agent Dispatcher: Multica Takes Over the Handoff Routine
Key Questions Answered
What the problem is and why it matters?
Prime numbers are 2, 3, 5, 7, 11, and so on: numbers divisible only by one and themselves. At first glance they appear chaotic, sometimes clustered together and sometimes separated by large gaps. The Riemann Hypothesis, formulated in 1859, asserts that this chaos hides a strict order, and it ties that order to the zeros of a special function. No one has been able to prove it in more than 160 years. It is one of the seven "Millennium Prize Problems": the Clay Institute offers $1 million for a solution to each.
What Lean is and why it is a strong argument?
The proof is published in Lean. In programmer terms, Lean is a compiler and test suite for mathematics. Every step of reasoning is written out rigorously, and the program checks by itself that each step follows from the previous ones. A mistake of the kind "here the author quietly skipped a step" gets caught automatically. There is a caveat. Lean confirms that the conclusion holds for the statement as written in it.
What are the key details on Context: the race of mathematical claims?
This is not the first claim of its kind. In September, OpenAI announced a solution to the Navier–Stokes problem on fluid flow. According to The Verge, the company spent millions of dollars and ran 10,000 agents because it had heard rumors that competitors were close to a solution. Many consider the result convincing, but it has no formal recognition either.
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.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.