Skip to content

Claude's proof of the Riemann hypothesis in 250,000 lines of code: the race of mathematical models tests regional verification teams

Share
Claude's proof of the Riemann hypothesis in 250,000 lines of code: the race of mathematical models tests regional verification teams

Researcher and mathematician at Anthropic, Levint Albogui, announced that he had produced a detailed proof of the famous Riemann hypothesis using a unpublished version of the Claude chatbot, a step that puts regional research labs and universities in a position to genuinely assess mathematical reasoning capabilities apart from commercial hype. The announcement came on the morning of Anthropic’s initial public offering on the U.S. stock market, raising broad academic questions about the timing of the scientific breakthrough disclosure alongside the financial milestones of the developing company.

The formulation of the proof took about six weeks of joint work between the researcher and the model, whereThe proof was written entirely in the Lean language (Lean)the computer-assisted language for formalizing mathematical proofs, exceeding 250,000 lines of code. The Riemann hypothesis, attributed to mathematician Bernhard Riemann, is among the most complex and celebrated problems in mathematics for roughly 170 years, positing a specific ordering for the distribution of prime numbers along the number line. Because prime numbers are a cornerstone of higher mathematics, any confirmed breakthrough in this area would have a profound impact on the foundations of all computational sciences.

While Albogui described his achievement as a new dawn in the pursuit of greater science, he also acknowledged that the precise timing of the announcement seemed grim due to its link with the company’s stock offering. He explained that he contacted the CEO, Dario Amodei, on Sunday to discuss how to disclose the finding, taking full personal responsibility for not waiting, and stressing that the skeptical view of the timing remains incorrect from his perspective.

Prominent mathematician Terence Tao noted that reviewing the proof and verifying its accuracy will still require a long period. Tao explained that absorbing and checking this volume of lines will take a non-trivial amount of time, pointing out thatthe positive aspect is the presence of a human author available for communicationand collaboration to accomplish this task, something previous announcements from other AI models lacked, leaving researchers puzzled.

OpenAI had claimed last September that its staff, using a unpublished version of its tools, solved part of the Navier-Stokes equations governing fluid motion in just 88 hours, through a training experiment that employed more than 10,000 semi-autonomous AI agents. That proof has not yet been validated, sparking criticism in academic circles and among high-level mathematics societies. OpenAI later announced progress or solutions to roughly 372 other high-level mathematical problems, claims that will require months, perhaps years, of specialist review to confirm.

These successive announcements, together with earlier claims by Anthropic of advances in mathematical and physical problems, have alarmed the mathematics community; a group of senior academics directed sharp criticism at companies that boast about solving major puzzles to showcase commercial progress, deeming this trajectorya danger to intellectual workas a result of technology use deviating from its original epistemic purpose.

What this means for the region

The episode places research centers and computer science and mathematics departments in the Gulf, Egypt and the Levant before an imperative to redirect investment toward formal verification languages such as Lean, rather than merely testing the interfaces of general-purpose generative models; the real stake in complex reasoning requires national talent capable of reading and auditing hundreds of thousands of code lines and examining the robustness of computational and mathematical logic before trusting machine outputs in any precise scientific or technological application.

The coincidence of these announcements with global financial markets also compels decision-makers in regional institutions, which pour large investments into advanced computing infrastructure and sovereign AI models, to build independent audit labs that separate genuine scientific achievements from marketing campaigns tied to listings and commercial competition; digital sovereignty is not built by consuming unchecked results, but by possessing systematic verification tools that can validate every line of code.

Don't miss the next story

Subscribe for updates