Physicist Proposes Candidate Lattice Definition of the Standard Model; Credits GPT-6 Astra

·

Physicist Nathaniel Craig has posted a paper and an accompanying essay on OpenAI’s website describing what he calls a candidate nonperturbative lattice definition of the Standard Model, one of the central theories of particle physics. Craig also said OpenAI’s GPT-6 Astra played a leading role in developing the proof behind the result.

If the claim holds up, it would mark progress on a long-standing foundational problem in theoretical physics: how to define the Standard Model on a spacetime lattice in a way that works beyond approximation methods. But the work is still at a preprint stage and, by Craig’s own description, should be understood as a candidate construction under review by the broader research community, not as a settled field-wide result.

The technical claim appears in Craig’s arXiv preprint, “A gauge-invariant measure for lattice chiral fermions” (arXiv:2610.09093), submitted Oct. 6, 2026, and prepared for submission to JHEP. Craig is listed there as affiliated with the Kavli Institute for Theoretical Physics at the University of California, Santa Barbara, the UCSB physics department and OpenAI. In the paper’s abstract, Craig writes: “We nonperturbatively construct a gauge-invariant measure for overlap fermions on four-dimensional Euclidean lattices in a class of nonabelian chiral gauge theories that includes the Standard Model. … This provides a candidate nonperturbative lattice definition of the Standard Model.”

That wording is narrower than a headline suggesting the Standard Model has simply been “put on a lattice.” The paper describes a construction in a class of chiral gauge theories that includes the Standard Model, and it repeatedly frames the result as a candidate definition rather than a final, universally accepted one.

In a Nov. 15 post on OpenAI’s website titled “A step toward defining the Standard Model on a lattice,” Craig explicitly credited GPT-6 Astra with a major research role. “GPT-6 Astra was essential to turning the refinement idea into a working construction,” he wrote. “It suggested the right way to carry a given set of force fields onto finer grids, showed that the comparisons on finer grids would converge to a definite answer, and took the lead in developing the detailed proof.” That account comes from Craig’s post and OpenAI’s presentation of the work; the source materials do not provide an independent assessment of Astra’s contribution.

The arXiv submission includes supplementary materials, including a mathematical companion and a Lean formalization and verification report. The paper itself contains a verification section describing three complementary verification avenues. The Lean report says, “All sixteen modules compiled with zero errors or warnings, and the exported proofs passed the separate axiom check and Lean kernel replay.” That means the mathematical statements, as formalized in Lean, were machine-checked for logical consistency. It does not by itself establish that physicists broadly agree the construction captures everything a full nonperturbative definition of the Standard Model should require, or that its physical significance has been settled.

The problem Craig is addressing has been open for decades. Physicists know how to put many quantum field theories on a lattice, which replaces continuous spacetime with a grid and can make hard calculations more tractable. The difficulty is doing that for chiral gauge theories, including the Standard Model, while preserving the right symmetries. A major obstacle is that straightforward lattice treatments of chiral fermions tend to produce extra unwanted particle copies, known as doublers. Craig’s paper uses overlap fermions, a well-known approach designed to address that issue.

Even if the construction is borne out, the source materials indicate that more work would be needed before it becomes a practical framework for numerical simulations. And while OpenAI said in an Oct. 26 post that it was organizing outside “shepherding” groups to verify and contextualize some AI-generated physics results, the materials provided with Craig’s paper do not show an independent verdict on this specific claim.

For now, the result is best understood exactly as Craig’s paper describes it: a candidate nonperturbative lattice definition of the Standard Model, posted on arXiv and accompanied by formal checks, but still awaiting the fuller scrutiny that would determine whether the field accepts it.

Tags: #physics, #standardmodel, #lattice, #openai, #gpt6