OpenAI’s GPT-5.2 has helped uncover a surprising breakthrough in theoretical physics:

A class of gluon scattering amplitudes long assumed to be zero actually becomes nonzero under a precisely defined “half-collinear” momentum regime.

This result challenges textbook assumptions in quantum field theory, opens new research paths (including extensions to gravitons), and showcases a powerful new model for AI-assisted discovery, where LLMs spot elegant patterns hidden in superexponentially complex math.

AI accelerating research yet again.

The paper: arxiv.org/abs/2602.12176

aipost 🏴