“The most notable mathematical results proved by LLMs have tended to be ones that we would classify as examples or counterexamples.”
A Fields medalist looked at the actual results and found the pattern. The wins are counterexamples, where a wide search plus enormous recall does the job. The unsolved part is knowing which branch of the tree to cut, which is what mathematicians actually do all day. That distinction gets lost every time a lab announces a model solved something.