Mathematician Daniel Litt on what he learned from designing a problem for the FrontierMath benchmark and the ability of reasoning models like o3-mini-high to solve it:
https://x.com/littmath/status/1898461323391815820
https://x.com/littmath/status/1898461323391815820