An Anthropic mathematician named Levent Alpöge announced on Friday that he used an internal, non-public version of the company's Claude chatbot to produce what he describes as a proof of the Riemann hypothesis, one of the longest-standing unsolved problems in mathematics. The announcement came one day after Anthropic's stock market debut, a collision of events that Alpöge himself acknowledged looked suspicious. "There is an obvious cynical view, to do with the timing, but it is false," he wrote, adding that he had messaged CEO Dario Amodei directly the previous Sunday to ask how to handle the disclosure.
The Riemann hypothesis, posed by Bernhard Riemann in the nineteenth century, concerns the distribution of prime numbers along the number line. It sits at the foundation of large portions of number theory, and a verified proof would rank among the most consequential results in the history of mathematics. The Clay Mathematics Institute lists it among seven "Millennium Prize Problems," each carrying a one-million-dollar award.
Alpöge's proof is written in Lean, a formal verification language that allows computers to check mathematical reasoning step by step. The document runs to over 250,000 pages. Terence Tao, among the most respected mathematicians alive, said on Friday that verification of the proof remains some way off. That gap between a submitted proof and a confirmed one is not unusual in formal mathematics, but it matters enormously here: the claim is extraordinary, and the evidence is currently inaccessible to anyone who cannot read and run a quarter-million pages of Lean code.
The timing creates a specific problem for evaluating the announcement. Anthropic's IPO was a major financial event, and a proof of the Riemann hypothesis attributed to its own AI model, publicized the morning after the company began trading publicly, is exactly the kind of story that could move sentiment around the stock. Alpöge went out of his way to disclaim that motive, but disclaimers are not evidence of their own truth. What is verifiable is that the proof used a version of Claude the public cannot access, so no independent mathematician can yet replicate the conditions under which it was generated.
The broader context is that AI companies have been announcing mathematical results with increasing frequency. OpenAI has publicized solutions to problems that had resisted human effort for decades. These announcements tend to follow a pattern: the company or its employees release a claim, independent experts begin reviewing it, and the final verdict takes months. In some cases, announced results have held; in others, errors have emerged in review.
What is established here is narrower than the headlines suggest. An Anthropic employee, working with an internal AI tool, produced a formal document claiming to prove the Riemann hypothesis. One prominent mathematician has said verification will take time. Anthropic has not, based on the available reporting, issued an official statement endorsing the proof. The result is a claim, not a confirmed discovery, made by someone with professional proximity to a company that had strong incentives for positive coverage in its first days as a public company. Those two facts do not cancel each other out, and neither one settles the mathematical question.
Levent Alpöge, an Anthropic mathematician, claimed to produce a proof of the Riemann hypothesis using an internal non-public version of Claude. The Riemann hypothesis, posed by Bernhard Riemann in the nineteenth century, concerns the distribution of prime numbers along the number line and sits at the foundation of large portions of number theory.
Alpöge's proof is written in Lean, a formal verification language that allows computers to check mathematical reasoning step by step. The document runs to over 250,000 pages, making independent verification currently inaccessible to mathematicians who cannot read and run this volume of Lean code.
Terence Tao, described as among the most respected mathematicians alive, said on Friday that verification of the proof remains some way off. That gap between a submitted proof and a confirmed one is not unusual in formal mathematics, but the extraordinary claim and inaccessible evidence create particular significance here.
Anthropic's stock market debut occurred one day before the announcement, creating suspicion that the proof's publicity could move sentiment around the stock. The proof used a version of Claude the public cannot access, preventing independent mathematicians from replicating the conditions under which it was generated.
Continue reading.
Tribe AI hits $100M in revenue on less than $10M raised
Mansour argues Trump speech contracts serve markets after CFTC halt
Kalshi founders face Utah court date as CFTC defends federal authority
Senate nominees avoid comment on prediction market ads
A few more worlds
worth exploring.
From the FreeMalta ecosystem.
The AI Journey
Explore AI beyond the headlines. Tools, ideas and a place to start.
Explore the journey 02 / THE GARAGEThe Garage
The companies behind the names. Explore their stories and the people building them.
Open the garage 03 / THE DUELThe Duel
Two contenders. A closer look. Put products side by side before choosing.
Choose your duel 04 / MALTAINSIDERMaltainsider
Discover the B2B side of Malta. A different lens on business and opportunity.
Meet Maltainsider