JF Puget: AI's real edge shows in math research, not IOI-style contests

JFPuget · x · 2026-08-14

In this exchange about AI capability boundaries, @tak3sh8 argued it would be no surprise if AI holds an edge in highly algorithmic tasks with clear metrics — e.g. solving hard IOI-style problems, or climbing a benchmark by stacking well-known tricks.\n\nJF Puget pushed back: the notable achievements are not IOI-style tasks, but domains where significant work has already been done — like math.

Related event: Debate: AI May Excel in Algorithmic Optimization(3 posts)→

Original post →

More from Models

Models channel →