Probably thought they were safe until chatbots started making progress towards the Riemann hypothesis and the openai claim they’ve resolved navier-stokes. As well as a number of other open questions solved by ai. They all kind of sucked at high level math until recently.