The Pulse
Physicist says GPT-6 Astra helped develop Standard Model proposal
Nathaniel Craig says GPT-6 Astra helped develop a candidate lattice definition of the Standard Model and led work on its detailed proof. The proposal appeared in an October preprint and leaves further mathematical checks and simulations for future work.

Physicist Nathaniel Craig says GPT-6 Astra helped turn his strategy for defining the Standard Model on a spacetime grid into a candidate construction and led work on its detailed proof. In a Nov. 15 account published by OpenAI, Craig said he had not been able to prove that his strategy worked in every case it needed to cover.
The proposal appeared in a preprint submitted to arXiv on Oct. 6.
Why put the theory on a grid?
The Standard Model describes known elementary particles and three of the four fundamental forces: the strong force, the weak force and electromagnetism. Its calculations often use perturbation theory, in which physicists begin with interactions switched off and add them back as small corrections.
That approach does not work for every problem. Craig explains that the strong force holds together the quarks inside a proton; without it, the proton falls apart. A nonperturbative definition instead keeps interactions present from the start rather than treating them as small corrections.
One way to construct such a definition is to represent spacetime as a lattice, or grid of points. The grid spacing sets the shortest distance the theory can resolve, and a finite grid allows calculations to be put on a computer.
The handedness problem
The weak force treats left- and right-handed particles differently. A straightforward attempt to place a left-handed particle on a lattice can introduce an unwanted right-handed partner, producing the wrong particle content for the weak force.
Earlier research produced a way to define particles known as overlap fermions on a lattice and distinguish their handedness. But, Craig writes, a further issue remained: choosing a phase consistently as the force fields change.
The preprint proposes a way to construct a measure for overlap fermions on four-dimensional Euclidean lattices, within a class of theories that includes the Standard Model. Its abstract describes the result as a candidate nonperturbative lattice definition, rather than a definition of the full theory in continuous spacetime.
What the model contributed
Craig's strategy was to represent the same force fields on successively finer grids and compare the allowed particle states. In his account, Astra suggested the right way to carry a given set of force fields onto finer grids and showed that the comparisons on finer grids would converge to a definite answer.
The arXiv entry lists a Lean verification report among its ancillary files.
Work still ahead
Craig writes that much work lies ahead to make the construction useful for realistic lattice simulations, and that scientific dialogue among experts is needed to decide whether it has all the properties physicists want. It also leaves open how to define the interacting theory in continuous spacetime.
Sources
Companies in this story
Related stories
- GPT-6 Astra Edits Qwen Weights, Raising Math Score to 48.1%
- OpenAI Launches LifeVal to Measure AI on Everyday Tasks
- Cornell researchers use one smartwatch to estimate full-body motion
- OpenAI Details Monitors Trained on Models’ Internal Activity
- ChatGPT Wallet lets US Pro subscribers approve agent purchases