February 13, 2026
Tiny particles, huge drama
GPT-5.2 derives a new result in theoretical physics
AI drops a particle-physics plot twist; comments split between “prove it” and “it’s happening”
TLDR: OpenAI says GPT‑5.2 helped spot and prove a new particle-interaction formula in a special setup, now posted as a preprint. Commenters are split between “show me external validation” and “AI just did science,” with memes about the singularity and cautious excitement that this could mark a real shift in research.
OpenAI says GPT‑5.2 helped spark a surprise in particle physics: a once-dismissed kind of gluon interaction might actually happen if the particles line up just so. Translation: the AI guessed a neat formula, an internal model spent 12 hours proving it, and human authors checked it with standard tests. The preprint is here. But the internet isn’t exactly holding hands.
The loudest chorus: skeptics who remember past hype. One top comment basically shrugs, “You claimed wild wins before—so we’ll wait for outside verification.” On the other side, AI optimists are popping popcorn: one user laughs that critics swore AI could never discover anything, while another jokes about gravitons and cries, “singularity when?!” In the middle are pragmatists pointing to the 12‑hour “reasoning run” and saying this feels like the first time an AI can grind through tough math without melting down.
Drama-wise, it’s classic internet: Team Prove-It vs Team History Is Being Made. Memes are flying (“AI did physics” speedruns), the link is getting spammed, and the phrase “half‑collinear” becomes “that weird lineup where the ‘impossible’ thing happens.” Whether it’s a revolution or a really fancy calculator, the community agrees on one thing: if this holds up, it’s a big moment for AI doing real science—not just regurgitating it.
Key Points
- •The preprint claims single-minus gluon tree amplitudes are nonzero in a precisely defined half-collinear momentum regime.
- •Authors computed base cases up to n=6, with GPT‑5.2 Pro simplifying them and conjecturing a general all‑n formula (Eq. 39).
- •An internal scaffolded GPT‑5.2 independently derived the same formula over ~12 hours and produced a formal proof.
- •The formula was verified via the Berends–Giele recursion relation and checked against the soft theorem.
- •Extensions from gluon amplitudes to graviton amplitudes have already been made with GPT‑5.2, with further generalizations planned.