In Simple Terms
Two possible solutions to a famous math problem have made headlines: one was reportedly produced by an AI system, and the other was developed by a human mathematician. Neither has been checked thoroughly enough to know whether it is correct. For now, the big story is not that the mystery has been solved, but that both claims still need careful review.
A Problem About Simple Numbers and Hidden Order
Prime numbers, such as 2, 3, 5, and 7, can be divided evenly only by themselves and by 1. Mathematicians know a great deal about primes, yet their distribution among whole numbers can look irregular: sometimes primes appear close together, and sometimes larger gaps separate them. That is why mathematicians have long sought rules that describe how primes are spread out, even when the pattern is not obvious from a list of numbers.
The Riemann Hypothesis is one of the best-known ideas linking the distribution of primes to a deeper mathematical structure. It concerns the properties of a function called the zeta function and the locations of particular points known as its zeros. Put simply, the hypothesis says that all of its nontrivial zeros lie on a specific line in the complex plane. The statement is brief, but proving it requires working through a vast web of mathematical relationships.
This does not mean mathematicians know nothing about prime numbers, or that the Riemann Hypothesis explains every detail of their distribution. Its importance is that it makes a precise prediction about an underlying order. If proved, it would provide a powerful foundation for many results involving primes. If a counterexample were found, it would pose a major challenge to current understanding. So far, the reports circulating do not show that either outcome has occurred.
Claude’s Claim Raises Expectations
The first part of the story began when Levent Albuge, an employee at Anthropic, posted a proof that he said had been generated by the company’s Claude language model. The claim goes beyond the Riemann Hypothesis in its familiar form: it says the proof also addresses the Generalized Riemann Hypothesis, a broader and more difficult version. That raised the prospect of a purported achievement extending beyond the solution of a single problem.
But attributing a proof to an AI system does not establish its mathematical value. A model can produce a long, well-organized text that seems convincing, but every step must be correct, and the conclusion must rest on arguments that independent experts can examine. For a problem this complex, confident language, elegant explanations, or a quick announcement are not enough to show that a solution works.
That is why the post should not be treated as a declaration that the problem has been settled. What is currently known is that Albuge presented work he said Claude had generated, and that the claim covered two hypotheses. It is not known that the mathematical community has verified and accepted the proof. There is a substantial difference between producing a text labeled a proof and showing that it closes every logical gap in one of mathematics’ most difficult open problems.
A Human Effort May Have Come First
After Claude’s claim emerged, another account of a possible proof came to light. On Friday, Ravi Vakil, president of the American Mathematical Society and a mathematician at Stanford University, wrote that Alessandro Masullo was responsible for the human work. Masullo was described as a young Stanford mathematician whose project had been followed and discussed over the past several months.
According to information Vakil shared, Brian Conrad, a Stanford mathematician who specializes in the subject, had been reviewing Masullo’s work and offering feedback for several months. Conrad confirmed that the details in the post were accurate. This suggests the work did not suddenly appear in response to Claude’s post, but it does not mean the human proof has passed review or been accepted.
The account matters because it puts a different possibility at the center of the discussion: a human effort to solve the problem may have begun before the AI-related announcement. Still, being earlier in time does not make a piece of work correct, and feedback from an expert is not the same as a completed review or final endorsement. The precise takeaway is that Masullo’s draft was still being developed, and people familiar with it had not announced that the problem was solved.
The Review Has Not Reached the Finish Line
Conrad was cautious when asked whether he expected Masullo’s work to hold up. He explained that judging the approach would require the paper to reach a more stable form, adding that it was undergoing substantial changes. That is an important point: when a paper is being revised extensively, its arguments and presentation may change, and the review may identify issues that require further work.
In mathematics, the strength of a proof is not measured by the appeal of its general idea alone. Each step must be justified, the assumptions must be clear, and the conclusion cannot depend on an unsupported leap. A small flaw in a single line can be enough to invalidate an entire argument, even if the rest of the paper is carefully written. That is why evaluating a major claim takes detailed, patient reading—not a quick impression.
Likewise, calling an approach promising or noting that an expert has followed it is not the same as confirming that it is correct. Conrad did not say the proof had succeeded; he tied his assessment to the paper reaching a more complete stage of review. Since the work is still undergoing substantial revisions, the information made public does not allow readers to conclude that either the Riemann Hypothesis or its generalized version has been proved.
Why Announce Work That Is Not Finished?
Conrad said that deciding whether to share news of Masullo’s work meant choosing between two uncomfortable options. One was to speak about it at that point; the other was to remain silent and return to the subject later. The decision was to share the information through Vakil’s post, even though the paper was not yet in final form. That explains why the news appeared, but it does not turn the announcement into confirmation of the result.
When competing claims emerge about a major problem, the timing of an announcement becomes part of the story. Staying silent could leave the public with news of an AI-generated proof but no awareness that a human effort was already underway. Speaking too soon, however, could wrongly suggest that a discovery had been established. The result is a tension between sharing relevant information and avoiding the presentation of an unfinished draft as a verified solution.
This dilemma does not justify guessing at motives the people involved have not stated, nor does it prove that one announcement was a response to the other. It does help explain why information about work that has not yet been fully checked may become public. The limits of the report should remain clear: there is a proof attributed to Claude, and there is a human paper under review, but the available information does not provide a final mathematical verdict.
Between the Drama of a Race and the Caution of Mathematics
It is easy for coverage to turn the story into a head-to-head contest: a language model versus a young scientist, with the most famous puzzle in mathematics as the prize. But that framing oversimplifies the problem. Mathematics does not award a result to whoever announces it first, and neither technical sophistication nor the fame of the problem settles the question. In the end, what matters is whether the reasoning is correct and can be checked.
The caution some mathematicians have expressed about the claims is not the same as declaring that both efforts are wrong. Here, skepticism is a temporary response to the lack of verification, not a judgment about an author or a tool. Similarly, the fact that neither claim has been verified so far does not prove that either one cannot be right. It simply means the available evidence has not yet reached the point where the question can be settled.
Major problems take time to assess. Arguments may need to be clarified, and a paper may go through repeated rounds of revision before it is understandable and ready for evaluation. When a result concerns a hypothesis that has remained open for so long, it is especially important to distinguish an initial announcement from a draft and from a proof that has survived scrutiny. These are not formalities: they are what separate an exciting claim from a confirmed mathematical discovery.
What We Know—and What We Don’t
What is known is that Levent Albuge, an Anthropic employee, posted a proof he said Claude had generated, and that the claim covered both the Riemann Hypothesis and the Generalized Riemann Hypothesis. It is also known that Vakil attributed a human effort to Masullo, and that Conrad confirmed the details of Vakil’s post. Conrad said he had followed the work and offered feedback over the past several months. He also explained that the paper was still undergoing substantial changes.
What remains unknown is whether either proof is correct, whether the human work will retain its current form after revision, and whether further scrutiny by specialists will confirm every step. Some mathematicians have expressed doubts about the claims, but doubt alone is not a counterproof. At the same time, the information reported here includes no verification that would justify announcing a solution.
As of the update dated November 20, 2026, this is an evolving story, not the end of the Riemann Hypothesis. What is worth following is not just who might succeed, if anyone does, but how the arguments are presented and how each detail stands up to examination. Until that happens, the hypothesis remains open, and both claims remain works in need of an accepted proof—not just a striking announcement.
The Bottom Line
One report describes a proof generated by Claude, while another describes a human draft by Alessandro Masullo that may have been in development for months. Neither has yet been established as a solution to the Riemann Hypothesis. Brian Conrad’s close attention to Masullo’s work and his confirmation of Ravi Vakil’s account do not mean the review is complete: the paper is still undergoing major revisions, and mathematicians have not settled the claims’ validity. For now, the most reliable conclusion is that the mystery has not been credibly declared solved. Careful scrutiny—not an eye-catching announcement—will determine whether either proof holds up.