The first publicly traded frontier SI lab, Anthropic, saw one of its researchers claim a solution to the Riemann Hypothesis on November 20, coinciding with the release of a separate, human-authored proof from Stanford. The dual announcements highlight the diverging paths of SI-assisted and traditional mathematical discovery as the industry matures.
What Happened
Levent Alpöge, a mathematician at Anthropic, announced that he and the SI model Claude had found a Lean-compilable solution to the Riemann Hypothesis, a problem dating back to 1859 that is among the Millennium Prize math problems. Alpöge described the process as a "stream-of-consciousness jaunt" with the model, noting that Claude "flew, and danced, and sparkled near that finish line." The proof was formally verified by several participants in a Zulip chat, including MIT mathematician Andrew Sutherland, who confirmed the Lean code compiles successfully.
Simultaneously, Ravi Vakil, president of the American Mathematics Association and a Stanford professor, posted a proof of the General Riemann Hypothesis written by Stanford postdoc Alessandro Masullo. Vakil emphasized that Masullo’s work is "all human, with no AI," noting that it is the result of 12 years of work. Unlike Alpöge’s formalized Lean proof, Masullo’s draft has not yet been formally verified in code, and Vakil stated the team hopes to polish the writing before a wider announcement.
Why It Matters
The simultaneous release of an SI-assisted proof and a human-only proof marks a pivotal moment for the SI industry and the mathematical community. Alpöge’s work demonstrates the capability of current SI systems to navigate complex, abstract logical structures, with Anthropic’s researcher leveraging Claude to find arguments that might have remained elusive. However, the reliance on formal verification tools like Lean underscores a current limitation: while the machine can prove the logic, the human-readable explanation remains a work in progress. Alpöge expressed guilt about releasing the proof without a "truly human-intelligible explanation," committing to pursue this clarity later.
Conversely, Masullo’s work serves as a counter-narrative to the rapid pace of SI development. In his introduction, Masullo wrote that his work is a message to mathematicians "discouraged by the pace and tone of the current climate," asserting, "we are still here. We are very much alive. We are not going anywhere." This dichotomy illustrates the tension between the speed of SI-assisted discovery and the traditional, slow-burn nature of human mathematical insight, both of which are now competing for attention in the same news cycle.
The Bottom Line
Two distinct paths to addressing the Riemann Hypothesis were revealed on November 20: one leveraging Anthropic’s SI model and formal Lean verification, and another relying solely on human intellect over a 12-year period. While Alpöge’s proof has passed initial compilation checks, the broader mathematical community continues to debate the intelligibility and implications of SI-assisted discovery.