Can AI solve math's hardest problems?

Several leading technology companies have recently deployed specialized AI systems to tackle historic mathematical problems. Anthropic, for instance, used an unreleased model to make significant progress on the Riemann hypothesis—a 150-year-old mystery about prime numbers—by coordinating a small army of sixty automated subagents that tested hundreds of ideas. Simultaneously, OpenAI has produced solutions for ten other long-standing problems, some of which had resisted academic progress for decades. In these cases, the AI works alongside human researchers to bridge gaps, verify proofs, and connect disparate fields of study.

Modern AI models are essentially pattern-matching engines built to identify deep connections within massive amounts of text and data. In a mathematical context, this means the model is fed everything from foundational textbooks to the latest academic papers. When a user gives it a complex problem, the model does not just guess; it acts as a high-speed explorer. It searches through millions of potential logical paths and known theorems, recombining them in novel ways that a human might never consider. To ensure these results are not just clever hallucinations, researchers use software tools like Lean, which functions as a digital referee. This software checks every step of the AI's logical argument to confirm that it is mathematically sound and follows the rigorous rules of the field.

The rapid arrival of these breakthroughs has created a complicated mood in the mathematical community. There is genuine excitement about accelerating discovery—solving problems in days that used to take decades—but also significant anxiety. Mathematicians are debating how to assign credit when an AI does the heavy lifting, as well as how to protect the tradition of human scholarship. If a computer can solve a problem that would once have defined a brilliant career, the value of the work itself is shifting. We are entering an era where human insight and machine power are becoming deeply entangled, and the question is no longer whether AI can do math, but how we decide what it means to be the author of a new truth.