Artem Voloskovets
⬤ Artificial intelligence just cleared a major hurdle in mathematics—it's now capable of working through original problems without anyone showing it the way first. AI systems aren't just crunching numbers anymore; they're actually discovering hidden patterns and building complete proofs that hold up under formal review. What used to be purely human territory is becoming a genuine collaboration between mathematicians and machines.
⬤ The big news? GPT-5.2 Pro solved Erdős Problem #397, a question about whether certain equations involving binomial coefficients have only a limited number of solutions when all parameters are different. Turns out, the answer is no—there are infinitely many solutions. The AI didn't just claim this; it built a formal proof that passed verification through a proof assistant, then got the green light from experts. The problem was officially marked "DISPROVED (LEAN)," confirming the AI's work was solid.
⬤ As one researcher put it, we're seeing "a new loop in mathematical discovery" where AI proposes arguments, proof assistants check them mechanically, and human experts review for any hidden issues. What makes this particularly significant is that Terence Tao—one of the world's leading mathematicians—reviewed and accepted the result, showing that AI-generated proofs are now being taken seriously at the highest levels of mathematics.
⬤ Why does this matter beyond math departments? Because mathematics underpins nearly everything in science—from physics and engineering to computer science and biology. When AI can help clear roadblocks in fundamental math problems, it has ripple effects across all these fields. If machines can accelerate core mathematical work, we might see faster breakthroughs in multiple scientific disciplines at once. This isn't just a cool AI trick—it's potentially a shift in how human knowledge advances.
Artem Voloskovets
