DeepMind and OpenAI models solve maths problems at level of top students

After the IMO gold by 2 labs, also gold by independent researchers with public gemini, silver by a team from china, and so on I wonder where the goal posts are heading. I distinctly remember a sit down with LeCun (I think it was after GPT4 launched, there was also a young guy from MS there) where he confidently said that "yes, yes, they work on poetry but they don't work on math. Call me when they solve hard math problems". Well, someone call him, dunno.