Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
GPT-5.2 Pro helped physicists conjecture a compact formula showing that a class of single-minus gluon tree amplitudes can be nonzero under special half-collinear kinematic conditions. That is the precise result behind headlines saying the AI “solved a 15-year physics mystery.” It is a genuine theoretical-physics result, but not an experimental discovery, a new particle interaction, or proof that the Standard Model is wrong.
The work was announced by OpenAI on February 13, 2026, in connection with the preprint “Single-minus gluon tree amplitudes are nonzero”. The preprint was described as being submitted for publication, so its peer-review status should be treated cautiously.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
Basic Physics: A Self-Teaching Guide, 3rd Edition | $11.47 | Buy on Amazon |
| 2 |
|
Fundamentals of Physics I: Mechanics, Relativity, and Thermodynamics (Open Yale Courses) | $35.00 | Buy on Amazon |
| 3 |
|
Must Know High School Physics | $14.57 | Buy on Amazon |
| 4 |
|
Physics | $61.98 | Buy on Amazon |
| 5 |
|
Physics for Scientists and Engineers With Modern Physics | $91.43 | Buy on Amazon |
What the result actually says
The researchers studied a special class of mathematical quantities used to describe gluon scattering. In ordinary, generic momentum configurations, a tree-level amplitude involving one negative-helicity gluon and the remaining positive-helicity gluons is generally expected to vanish.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →The new result identifies a precisely defined half-collinear region in momentum space where that amplitude is instead nonzero. In the relevant kinematic chambers, the stripped amplitude takes the simple values +1, −1, or 0, depending on the ordering and conditions imposed on the momenta.
#1 Best Overall
That distinction matters. The result does not say that all single-minus gluon interactions are possible, common, large, or experimentally observable. It says that the usual zero conclusion does not apply everywhere: a special, mathematically consistent slice of the allowed kinematic space behaves differently.
Why gluons and scattering amplitudes matter
Gluons are gauge bosons associated with the strong nuclear force. They transmit the force that binds quarks inside protons, neutrons, and other hadrons. Unlike photons, gluons also carry the charge associated with the force they mediate, which makes the mathematics of their interactions particularly rich.
A scattering amplitude is a mathematical building block used to calculate the structure and probability of a particle process. Physicists usually do not calculate an observable directly from one diagram; they combine amplitudes and then apply the appropriate physical rules to obtain measurable quantities.
A tree amplitude is the lowest-order contribution in perturbation theory. It is built from diagrams without quantum loops. Tree-level expressions are often the first place where hidden mathematical patterns can be found, although they do not automatically describe every effect present in a real experiment.
The calculation also uses helicity, a description of how a massless particle’s spin is oriented relative to its motion. “Single-minus” refers to an amplitude with one negative-helicity gluon and n−1 positive-helicity gluons.
The old zero result—and its exception
For generic momenta, standard amplitude arguments make the single-minus tree amplitude vanish. It is easy to turn that statement into an inaccurate headline saying the interaction was thought to be impossible. The more precise statement is that the amplitude vanishes under the usual generic-kinematics assumptions.
The preprint examines a special configuration called half-collinear kinematics. In this regime, the participating momenta obey additional alignment or orthogonality conditions. Those constraints place the calculation on a special slice of momentum space rather than in the generic case.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchGeneric case: the single-minus tree amplitude is expected to be zero.
Half-collinear case: the same class of amplitude can be nonzero, with a simple piecewise result determined by the kinematic chamber and ordering.
A useful analogy is a geometric one. A rule that holds across most of a landscape may fail on a carefully defined boundary or surface. Finding that exception does not invalidate the rule everywhere; it reveals that the original assumptions were narrower than they first appeared.
What GPT-5.2 contributed
The result came from a human-AI research workflow, not from an unsupervised chatbot producing a finished discovery on its own.
- Human researchers calculated initial cases. They worked through examples for integer values of n up to 6. The expressions became increasingly complicated as the number of particles grew.
- GPT-5.2 Pro simplified the expressions. Once the formulas were put into a more manageable form, the model recognized a pattern across the calculated cases.
- The model proposed a general conjecture. GPT-5.2 Pro suggested a compact formula intended to work for arbitrary n, rather than only for the examples that had already been calculated.
- A separate internal OpenAI model worked on a proof. OpenAI says a scaffolded internal model spent approximately 12 hours reasoning through the conjecture and produced a formal proof.
- Physicists checked the result independently. The researchers tested the proposed expression using established amplitude techniques, including Berends–Giele recursion and known consistency relations.
- The work was written up as a preprint. The named authors include researchers affiliated with the Institute for Advanced Study, Vanderbilt University, OpenAI, the University of Cambridge, and Harvard University.
The most defensible description is therefore that GPT-5.2 Pro helped simplify examples and conjecture the all-n formula. The proof and scientific validation involved additional model-assisted reasoning and human physicists.
Rank #3
What the new formula does
The preprint’s key all-n expression is identified as Equation 39 in the cited version. It gives the stripped single-minus gluon tree amplitude in the special half-collinear region.
For a general reader, the important achievement is not the notation itself but what the formula replaces. Direct diagrammatic calculations produce a rapidly expanding collection of terms. The compact expression supplies a general rule for the whole class of amplitudes, with the outcome in the relevant region reduced to a simple sign-or-zero structure.
Readers familiar with spinor-helicity methods and the conventions used in the preprint can consult the full paper and its Equation 39. Reproducing the equation without those conventions would make the result look more mysterious rather than more understandable: its symbols encode the momentum, helicity, ordering, and kinematic assumptions that make the formula valid.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →How researchers checked it
The announcement and preprint describe several checks designed to test whether the proposed formula fits the established structure of gauge-theory amplitudes.
- Berends–Giele recursion: a recursive method for building multiparticle tree amplitudes from lower-point building blocks.
- Soft-theorem behavior: constraints that apply when one particle’s momentum becomes very small.
- Cyclicity: the expected invariance under appropriate cyclic reorderings of external particles.
- Kleiss–Kuijf relations: identities that relate color-ordered amplitudes and reduce the number of independent quantities.
- U(1)-decoupling identities: additional consistency relations inherited from the organization of gauge-theory amplitudes.
- Direct hand checks: calculations by the authors for specific configurations and low-point cases.
These tests are meaningful because a candidate formula must satisfy more than a few numerical examples. It must also behave correctly under recursive constructions, limiting procedures, and algebraic identities already known to govern scattering amplitudes.
At the same time, consistency checks do not turn the result into an experimental observation. They show that the expression is compatible with the mathematical framework being used; they do not establish that the special process will be measurable in an ordinary collider environment.
Rank #4
Where the “15-year mystery” headline comes from
The “15-year mystery” wording comes from coverage of the announcement and refers to a problem that Institute for Advanced Study physicist Nima Arkani-Hamed had reportedly been interested in for roughly 15 years. It is useful context, but it should not be read as the name of a formally defined problem with a universally agreed 15-year deadline.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
The headline is also incomplete if it suggests that physicists had simply declared the interaction impossible and an AI overturned them. The conventional zero result applied to generic kinematics. The new work identifies an exceptional half-collinear regime in which the amplitude is nonzero.
For that reason, “GPT-5.2 solved a 15-year physics mystery” is acceptable only as shorthand when immediately qualified. A more accurate formulation is: GPT-5.2 helped researchers find and verify a compact formula for a previously overlooked special kinematic regime.
Who worked on the preprint?
OpenAI lists the authors as:
- Alfredo Guevara — Institute for Advanced Study
- Alex Lupsasca — Vanderbilt University and OpenAI
- David Skinner — University of Cambridge
- Andrew Strominger — Harvard University
- Kevin Weil — OpenAI, on behalf of OpenAI
The Institute for Advanced Study’s account separately confirmed the participation of its scholars Alfredo Guevara, Andrew Strominger, and David Skinner.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Does this change the laws of physics?
No. The result is important within scattering-amplitude theory, but it does not show that the Standard Model is incorrect, reveal a new fundamental force, or announce a new particle.
It clarifies how a special class of gluon amplitudes behaves when the momenta satisfy unusual conditions. The work may help physicists identify structures useful in future calculations, and the authors report related extensions from gluons to gravitons. Those extensions are theoretical follow-on results, not confirmed experimental predictions.
Best Value
There is also no evidence in the announcement for an immediate application to computing, energy production, consumer technology, or accelerator design. “Nonzero” describes the mathematical value of an amplitude; it does not mean that the corresponding process is frequent, energetic, or easy to detect.
What this says about AI-assisted science
The most significant AI lesson is narrower—and more useful—than the claim that AI can now do theoretical physics independently.
Advanced models can be valuable when researchers provide a well-defined problem, exact symbolic inputs, and a way to check proposed answers. In this case, the workflow combined:
- human selection of the physics problem and kinematic regime;
- human derivation of initial examples;
- AI-assisted symbolic simplification and pattern recognition;
- AI-assisted proof construction;
- human review using established physical identities and recursion methods.
That workflow is different from asking a model a broad physics question and accepting its first answer. It also means that success here should not be generalized into a claim of broad autonomous scientific competence. OpenAI’s separate science and mathematics overview discusses model evaluations, but benchmark scores measure particular tasks and formats rather than replacing expert validation in research.
What readers should take away
The durable result is a new theoretical formula for a narrowly defined but mathematically interesting class of gluon tree amplitudes. The amplitude is generally zero for generic momenta, yet becomes nonzero in a special half-collinear regime. GPT-5.2 Pro helped researchers see the pattern; another internal model helped develop a proof; and human physicists checked the result with known amplitude methods.
That is a noteworthy example of AI assisting frontier mathematical physics. It is not evidence that a chatbot independently discovered a new force, made an experimentally observed prediction, or replaced the physicists responsible for framing and verifying the work.
Because the source announcement described the work as a preprint being submitted for publication, readers should not treat it as peer-reviewed until a publication record independently confirms that status.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsQuick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
