Ofir Press: Math is far more RL-able than SWE, other fields won't move as fast

OfirPress · x · 2026-10-29

Princeton researcher and SWE-bench creator Ofir Press argued that math is much more 'concrete', 'definable', and 'RL-able' than other domains including software engineering, so he doesn't expect progress elsewhere to be as fast as in math. In a follow-up he said he finds it hard to imagine the alternative scenario being plausible.

Related event: Princeton's Ofir Press: Millennium Problems Are the Best AI Benchmark, 70% of Open Math Problems Solved Within a Year(7 posts)→

Original post →

More from AGI Musings

AGI Musings channel →