OpenAI says its internal models have solved three problems that humans had been trying to close for DECADES: one in mathematics, one in computer science, and one in physics. In mathematics, it’s the local smoothing conjecture. It concerns how waves concentrate, with the idea that the most extreme peak of concentration lasts only an instant. The problem had remained open since 1991, and humans had only managed to solve the simplest case. If OpenAI’s proof holds, it also resolves another classic puzzle: the Kakeya problem, which asks how small a shape can be while still containing a line pointing in every direction. It has no direct practical application, but it closes a question the field has pursued for decades. In computer science, it’s the Exponential Time Hypothesis. OpenAI claims to have proven it false. The hypothesis states that some difficult problems have no shortcuts—that solving them requires testing nearly all possible combinations. Hundreds of papers have relied on this hypothesis to argue that certain tasks cannot be done much faster. If it is indeed false, there exists a shortcut almost no one expected, and those conclusions lose their foundational basis. In physics, it’s the mathematical existence of QCD—quantum chromodynamics. QCD is the theory explaining how quarks and gluons hold atomic nuclei together. It has been used for about 50 years and matches experimental results, but no one had mathematically proven that it holds as a rigorous theory. OpenAI claims to have proven it. A component of this, Yang–Mills theory, is part of one of the seven Millennium Prize Problems—those offering a $1 million reward—but the prize requires more than what OpenAI claims to have proven. No external parties have confirmed anything yet, and the physics proof has not even been published. According to OpenAI, its models arrived at the physics proof between September 9 and 18, and at the other two results between October 16 and 18. Now it’s up to humans to verify.