OpenAI’s GPT-5.2 theoretical physics result centers on a claim that single-minus gluon tree amplitudes do not always vanish. The preprint frames nonzero behavior in half-collinear configurations in Klein space or with complexified momenta, shifting a long-held simplifying assumption in amplitude calculations.
It also presents a closed-form, piecewise-constant formula for minus-helicity gluon decay into plus-helicity gluons. In the half-collinear regime, stripped amplitudes map to integer values across kinematic chambers, with assignments produced through Berends-Giele recursion and checked against soft and symmetry constraints.
The research process matters as much as the formula:
- low-n cases were derived and verified by hand up to n=6
- GPT-5.2 Pro proposed a general arbitrary-n conjecture
- a separate scaffolded internal model ran for more than 12 hours to produce a proof, then researchers hand-checked recursion and consistency identities
This GPT-5.2 theoretical physics result is strategically important because it demonstrates a repeatable AI research pattern: conjecture generation, proof construction, then conventional validation. The same research pattern appears more clearly when viewed alongside the GPT-5.3 Codex launch. Taken together, these updates highlight a consistent methodology: generate a conjecture, build a proof path, and verify the result with standard validation checks. Early extension from gluon amplitudes to graviton amplitudes indicates the method can transfer quickly to adjacent high-energy theory problems.
