tl;dr: A former mathematician argues that AI's rapid progress on theorem-proving benchmarks (like the recent "First Proof" project, where labs solved 6-8 of 10 research-level problems) threatens mathematics as a profession, but exposes a deeper misunderstanding: math's real value lies in human comprehension and concept-building, not proofs themselves. He warns that AI-generated proofs are often unintelligible "Mathslop" that doesn't accrete to the corpus, and that the mathematical community's honor code—which rewards only theorem-proving—leaves it dangerously vulnerable to being declared "solved" by AI while its actual cognitive contributions go unrecognized.