AI Scientist Spots What Physicists Missed in Gluon Scattering

An artificial intelligence system first conjectured a general formula for a class of gluon scattering amplitudes long assumed to vanish at tree level. Researchers from the Institute for Advanced Study, OpenAI, Vanderbilt University, Cambridge University and Harvard University later proved the result in a preprint posted to arXiv. The work focuses on single-minus configurations, in which one gluon carries negative helicity and the remaining gluons carry positive helicity. Standard textbook arguments had indicated that these amplitudes disappear under generic momentum conditions. The study shows the amplitudes exist on a defined slice of momentum space when the particles satisfy a half-collinear alignment. In that regime, certain spinor products vanish in a coordinated way, allowing the interaction to be constructed systematically. Human researchers first computed explicit examples for small numbers of gluons by hand. GPT-5.2 Pro simplified those expressions and identified a pattern that led to a conjectured closed-form formula valid for any number of particles. An internal, scaffolded version of the system then spent roughly 12 hours reasoning through the conjecture and produced a formal proof. The formula was checked against the Berends-Giele recursion relation and several consistency conditions in quantum field theory. These checks included cyclic symmetry, reflection symmetry and Weinberg soft theorem. In one special momentum arrangement known as region R1, the expression reduces to a compact set of plus, minus or zero values.