█

AI/TLDR

OpenAI · 2026-11-15 · major

GPT-6 Astra and the Standard Model — a candidate lattice definition

In short

UCSB physicist Nathaniel Craig says GPT-6 Astra led the detailed proof in his paper proposing a candidate lattice definition of the Standard Model. OpenAI published his account on its new science blog on 15 November.

OpenAI · 2026-11-15 · 4 sources checked

White OpenAI logo on a black background

A physicist credits OpenAI's GPT-6 Astra with leading the proof behind a proposed lattice version of the Standard Model.

Quick facts

ResearcherNathaniel Craig, UC Santa Barbara + OpenAI
AI modelGPT-6 Astra
PaperarXiv 2610.09093, submitted 6 Oct 2026
Length60 + 63 pages, 12 figures
Formal checkLean 4 formalization on Zenodo
StatusCandidate definition, not yet accepted

What is it?

Nathaniel Craig's paper 'A gauge-invariant measure for lattice chiral fermions' proposes a way to put the full Standard Model of particle physics on a spacetime grid. On 15 November OpenAI published Craig's guest post about the work on its new science blog, where he says GPT-6 Astra led the detailed proof.

How does it work?

Lattice methods handle the strong force well, but the weak force treats left- and right-handed particles differently, and simple lattice rules add unwanted mirror partners. Overlap fermions remove those partners but leave a phase problem that Martin Lüscher solved only for simpler theories. The paper builds a gauge-invariant fermion measure for a class of theories that includes the Standard Model, and backs it with a Lean 4 formalization.

Why does it matter?

A non-perturbative definition of the Standard Model has been an open problem for more than 25 years, Craig says. The case shows a frontier model doing the long technical proof work in theoretical physics, with a machine-checked companion so others can test the claim instead of trusting it.

Who is it for?

physicists, mathematicians and people tracking AI for science

Frequently asked questions

What exactly did GPT-6 Astra do in the Standard Model lattice paper?
According to Nathaniel Craig's account, he already had a strategy but had not proven it worked for every required case. Craig says GPT-6 Astra helped turn that strategy into a construction and took the lead in developing the detailed proof. The arXiv listing itself names only Craig as author and does not mention AI.
Is the lattice Standard Model problem now solved?
Not yet. Craig's paper calls the result a candidate nonperturbative lattice definition of the Standard Model. It covers four-dimensional Euclidean lattices and gauge fields the paper calls admissible. Whether it counts as a solution depends on whether other physicists find the construction holds up under review.
Can other researchers check the GPT-6 Astra proof?
Yes. Craig published a mathematical companion and a Lean 4 formalization on Zenodo, with a Lean verification report, the Lean source project, scripts and logs. Project code is Apache-2.0 and the written material is CC BY 4.0. A two-dimensional supplement with calculation records is a separate Zenodo record.

Sources · 4 outlets

Tags

  • gpt-6-astra
  • openai
  • physics
  • standard-model
  • lattice-gauge-theory
  • chiral-fermions
  • lean
  • formal-verification
  • ai-for-science
  • theorem-proving

← All releases · Learn AI