Physicist says OpenAI's GPT-6 Astra wrote the proof for a lattice definition of the Standard Model
The proof sits in a preprint that is not yet peer reviewed, and the physicist calls it an answer in principle, not yet in practice.
The Standard Model is the theory of particles and forces behind most of physics. It has no definition that avoids treating interactions as small corrections, which physicists call a nonperturbative definition. The usual route is to put the theory on a grid of points, a lattice, but the weak force treats left- and right-handed particles differently, and that breaks the usual lattice recipe. A fix for the extra copies of particles it produces has existed since the 1990s, yet OpenAI’s post says the remaining step for the full Standard Model stayed open for more than 25 years.
Nathaniel Craig, who leads the particle theory group at UC Santa Barbara and is a visiting scientist at OpenAI, had a strategy: compare the particle states on finer and finer grids. He could not prove it worked in every case. His paper, posted on arXiv on 6 October 2026, says GPT-6 Astra “was essential”: it turned the strategy into a construction and led the detailed proof, including how to carry force fields onto finer grids and why the comparisons converge. Craig wrote parts of the manuscript and rewrote much of Astra’s text.
The paper says Astra also carried out a Lean check (a computer-verified proof) of the mathematical statements, and that AI agents reviewed whether those statements match the paper. OpenAI’s post of 15 November, the first on its new science blog, describes Astra’s role. It says the lattice theorists Maarten Golterman, Andreas Kronfeld and Yigal Shamir looked over the proof and asked questions.
What it means
If experts accept the proof, the Standard Model gets a candidate precise definition on a lattice, with an AI system doing most of the proof. OpenAI links this to its Yang–Mills existence proof as the other piece a continuum version would need.
Caveats
- This is Craig’s and OpenAI’s own account. The paper is a preprint and not peer reviewed, and the post says scientific dialogue among experts is still required.
- The post calls it “far more an answer in principle than an answer in practice”. It holds at a fixed grid spacing under a bound on field strengths, and does not define the theory in continuous spacetime.
- Whether the construction satisfies reflection positivity, a standard consistency condition, is open.
- Craig is a visiting scientist at OpenAI, which published the post. Golterman, Kronfeld and Shamir reviewed the proof informally and are not named as having endorsed it.
- The Lean check covers the formal statements under their recorded assumptions. The paper says that it is not the final word in physics.