Can GPT-5.2 solve a complex physics problem? AI achieves a path-breaking scientific breakthrough after solving a decade-long mystery

Have you ever imagined an AI solving a complex problem that has challenged scientists for more than a decade? Whether we accept it or not, that moment has arrived. In a recent path-breaking breakthrough, researchers revealed that an advanced AI system successfully cracked a long-standing theoretical physics puzzle. By processing massive datasets, testing multiple theories, and spotting patterns that humans often miss, the model delivered results that matched experimental evidence. Experts say this milestone shows how AI can accelerate scientific discovery, assist researchers, and examine ideas on an unprecedented scale. While human guidance remains crucial, this achievement marks a significant step forward in using artificial intelligence to solve real-world scientific problems.


GPT-5.2 Pro has made a major scientific advance by proposing a new formula that shows gluon interactions - long believed to disappear - can actually occur under certain conditions. According to Adwaitx, the internal OpenAI model spent nearly 12 hours developing a detailed mathematical proof to support the formula’s validity. The results were later reviewed and verified by researchers from Harvard University, University of Cambridge, Institute for Advanced Study, and Vanderbilt University, reported Adwaitx. The study was published on February 13 in arXiv preprint titled 'Single-minus gluon tree amplitudes are nonzero'.


WHAT AI REVEALED ABOUT GLUON AMPLITUDES

Particle physicists rely on scattering amplitudes to estimate how likely particles are to interact. For gluons, these calculations usually become simple at the 'tree level', where complex quantum effects are ignored. As per study, as quoted by Adwaitx, for decades, scientists believed that one specific setup - where one gluon has negative helicity and all others have positive helicity - always produced a zero result, so it was largely ignored.

GPT-5.2 uncovered a flaw in that long-standing assumption. According to the study, the usual argument only works when particle momenta are generic and not specially aligned. In a narrowly defined region known as the half-collinear regime, that logic breaks down. In this case, the interaction does occur. The AI went further by proposing Equation 39, a formula that precisely describes how the amplitude behaves in this overlooked scenario.

HOW AI CRACKED THE CODE

Human physicists had previously calculated scattering amplitudes by hand only for small cases, up to n = 6, but the results were extremely complex and difficult to generalize. GPT-5.2 Pro simplified these messy expressions into much cleaner forms and spotted a repeating pattern, pointing toward a single formula that could apply to all values of n.

A scaffolded version of GPT-5.2 then spent about 12 hours reasoning through the problem on its own, reaching the same formula and building a full mathematical proof. According to the study as stated by Adwaitx, researchers later checked the result using the Berends-Giele recursion method and confirmed it obeyed the soft theorem. The formula also passed five strict consistency tests, strengthening confidence in its validity.

Nima Arkani-Hamed, a professor of physics at the Institute for Advanced Study, called the newly derived formulas 'strikingly simple', as quoted by Adwaitx, noting that he had been curious about these scattering processes for nearly 15 years. He explained that uncovering such clean expressions has traditionally been a slow and delicate task, often involving trial and error. According to him, this kind of work now appears far more automatable, a shift made possible by modern AI tools that can handle complexity and pattern-finding at an unprecedented level.