Early responses range from cautious excitement to open bafflement, and Lean experts have already started checking the formal proof.
On Friday, a note claiming a proof of the Riemann Hypothesis appeared on the website of Anthropic mathematician Levent Alpöge. Within hours it had spread across the mathematics community. The note says Alpöge found the proof working with Claude, that the method also settles the Generalized Riemann Hypothesis, and that the whole argument has been formalized in the Lean proof assistant. Only the Lean code has been released, and there is no human-readable paper yet.
That unusual format has shaped the reactions. With no paper to read, mathematicians and scientists have focused on two questions: does the Lean code really state the Riemann Hypothesis, and does it compile? Here is what some of them have said so far.

Mathematician Elliot Glazer gave one of the most concrete early assessments. He said the formalized statement in Alpöge’s release is in fact the Generalized Riemann Hypothesis, which answers the first worry about any Lean proof: whether the theorem being proved is the one being claimed. He said the code was still compiling and would take a few hours, but that he expected it to compile, since that would surely have been checked ahead of time. He put the chance of the result resting on a bug in Lean itself at well under 1%, arguing that frontier models virtually never trigger such bugs by accident.
Thomas Bloom, a Royal Society University Research Fellow at the University of Manchester who works in number theory and additive combinatorics, was struck less by the claim itself than by its size. He said he couldn’t decide which surprised him more: that AI claims to have proved the Riemann Hypothesis, or that it claims to have produced a coherent proof that would run past 250,000 pages if written out for humans.
Bloom later pointed people to the Lean community’s Zulip chat, where Lean experts are investigating whether the claimed formal proof is valid. That public check is likely to be the first real verdict on the claim.
Harry Stoltz, a mathematician at NYU’s Courant Institute, was among the first to flag the note. His post, which simply said that Alpöge and Claude were claiming a proof of the Riemann Hypothesis, has been viewed nearly 200,000 times.
Ayush Khaitan, who works on math and AI at Princeton Language and Intelligence, read the result as a case for keeping humans in the loop. Quoting Stoltz’s post, he said that he and a collaborator had discussed only the day before that a human in the loop is still the best way to solve math problems with LLMs. He added that Alpöge deserves as much credit as Claude for the result.
The note itself gives Claude a very large share of the credit, describing the model as having outpaced Alpöge near the end. That makes the question of how to divide credit between person and model a live one.
Ellie Sleightholm, an applied maths PhD student at Cambridge’s Department of Applied Mathematics and Theoretical Physics (DAMTP) and a popular science communicator, struck the cautious note that many shared: a claimed proof, which would be huge if correct.
Pierre Richemond, multimodal lead at Cohere and a former Google DeepMind scientist, described the situation as a strange new state of knowledge. He said the hypothesis may be solved pending Lean review, which he called not totally unexpected but raising more questions than it answers. In his view, AI-driven mathematics has created something that barely existed before 2026: work that is unmistakably serious mathematics but has not been verified in human terms at all.
Kit Yates: excitement and caution
Mathematician and author Kit Yates wrote on his Substack that a claimed proof is not yet an accepted one, and that the consequences would be enormous for both mathematics and our understanding of AI if it survives scrutiny. He noted that the hedge in Alpöge’s opening line, “I believe,” carries less weight than it would have a few years ago. He also stressed that a Lean check still needs humans to confirm that the formalized statement really is the Riemann Hypothesis.
What the reactions have in common
Almost nobody is dismissing the claim, which says a lot about how far AI mathematics has come this year. Alpöge has a strong track record. He used Claude to disprove the Jacobian conjecture in July, and he is also behind a newly posted claim that the Navier-Stokes equations can blow up without forcing. The Lean code also gives the community something to check mechanically, which wasn’t the case for earlier claimed proofs of the hypothesis.
Equally, almost nobody is treating it as settled. The concerns are the same ones Terence Tao raised about AI proofs that arrive as black boxes: a machine-checked proof that nobody can read may settle whether the hypothesis is true while saying almost nothing about why. A declaration signed by 25 Fields medalists warned in September against exactly this kind of rushed release.
The first verdict should arrive soon, and it will come from the Lean community rather than from social media. If the archive compiles, the final theorem matches the Riemann Hypothesis, and it relies only on Lean’s standard axioms, the most famous problem in mathematics will have been settled, at least formally, by a mathematician working with an AI model. That would come only weeks after OpenAI’s quasi-Riemann result was hailed as the biggest step toward it in generations.