OpenAI: GPT‑5.2 Derives And Proves New Formula In AI’s First Physics Breakthrough

by
Alisa Davidson

Revealed: February 16, 2026 at 8:26 am Up to date: February 16, 2026 at 8:26 am

by Ana

Edited and fact-checked:
February 16, 2026 at 8:26 am

To enhance your local-language expertise, generally we make use of an auto-translation plugin. Please notice auto-translation might not be correct, so learn unique article for exact data.

OpenAI introduced that it has launched a brand new analysis preprint detailing work during which its GPT‑5.2 mannequin independently recognized a mathematical sample and produced a proper proof, an final result the group describes as the primary unique theoretical physics contribution generated by one in every of its methods.

The research examines an extended‑standing assumption in particle physics relating to gluon interactions, concluding {that a} class of scattering amplitudes beforehand believed to fade can, beneath particular momentum circumstances, in actual fact be nonzero.

The preprint, titled “Single-minus gluon tree amplitudes are nonzero,” is authored by researchers from the Institute for Superior Research, Vanderbilt College, the College of Cambridge, Harvard College, and OpenAI. It focuses on scattering amplitudes, the portions used to calculate the chance of particle interactions.

Whereas many gluon amplitudes simplify at tree degree, configurations involving one adverse‑helicity gluon and a number of constructive‑helicity gluons have historically been handled as yielding zero amplitude primarily based on normal arguments.

The authors report that this conclusion doesn’t maintain in a exactly outlined area of momentum area often called the half‑collinear regime, the place particle momenta align in a particular however mathematically constant means. Inside this regime, the amplitude doesn’t vanish, and the crew gives an specific calculation. The discovering opens new avenues for additional work, together with extensions to graviton amplitudes.

A notable side of the analysis considerations methodology. GPT‑5.2 Professional first proposed the overall method that seems as Eq. (39) within the preprint, after simplifying complicated expressions derived manually for decrease‑order instances. A scaffolded inside model of GPT‑5.2 then spent roughly 12 hours reasoning by means of the issue, independently arriving on the identical method and producing a proper proof. The consequence was subsequently verified utilizing established strategies such because the Berends–Giele recursion relation and delicate‑restrict checks.

Based on the authors, the method has already been utilized to increase the evaluation from gluons to gravitons, with extra generalizations underway. OpenAI states that additional AI‑assisted findings will probably be detailed in future publications.

Rising Proof Of AI‑Led Discovery Fuels Debate Over Whether or not Machines Can Generate New Science

Within the wake of OpenAI’s newest analysis milestone, debate is predicted to proceed over whether or not synthetic intelligence can genuinely originate new scientific concepts. Skeptics are more likely to query whether or not fashions are uncovering novel insights or just recombining current data in refined methods.

Nonetheless, the rising physique of outcomes rising from superior methods is making that distinction more and more troublesome to attract. As AI begins to probe and problem assumptions which have formed main scientific disciplines for many years, the notion of machine‑generated discovery is shifting from speculative fiction to a improvement that seems more and more imminent.

Alisa, a devoted journalist on the MPost, makes a speciality of cryptocurrency, zero-knowledge proofs, investments, and the expansive realm of Web3. With a eager eye for rising developments and applied sciences, she delivers complete protection to tell and interact readers within the ever-evolving panorama of digital finance.

Extra articles