Today, I ran into two things—one that made this era feel a little unreal, and one that made it feel very real. I. A 23-Year-Old Outsider Cracked a 60-Year-Old Conjecture Liam Price, 23, with no advanced mathematical training. On a Monday afternoon a week ago, he casually tossed an unsolved Erdős problem to GPT-5.4 Pro. Problem #1196, a conjecture about "primitive sets," had stumped humans for 60 years. In a single session of about 80 minutes, the model produced a proof. Cambridge undergraduate Kevin Barreto helped organize it. Later, Jared Lichtman and Terence Tao participated in simplifying it, distilling the key insight from the LLM's originally rough output. Terence Tao's comment was rather cutting: "Humans had looked at this problem. Everyone who looked at it collectively went down the wrong path at step one." It wasn't that no one had thought about it; it was that everyone who had thought about it made the same mistake.…