OpenAI has published model outputs addressing three open problems in mathematics and physics: the Kakeya conjecture, the Exponential Time Hypothesis, and quantum chromodynamics. These are not toy benchmarks. They are problems that have resisted professional mathematicians and physicists for decades.
The release includes verification pathways, meaning outside researchers can check the work and build on it. That detail matters more than the results themselves. A model producing unverifiable claims about hard science is noise. A model producing checkable intermediate steps is a research tool.
Read the full post for the specific outputs and the methodology OpenAI used to engage academic communities in review. The process they describe, not just the conclusions, is what determines whether this becomes a reproducible workflow or a one-time demonstration.
[READ ORIGINAL →]