OpenAI published a guest post by physicist Nathaniel Craig on November 15, 2026, describing how GPT-6 Astra helped develop a candidate lattice definition of the Standard Model. The underlying research appeared earlier, in an October 6 arXiv preprint carrying Craig’s name.
Craig, who leads the particle theory group at the University of California, Santa Barbara, is also a visiting scientist at OpenAI. His account gives the model a substantial role: he supplied the strategy, while Astra helped turn it into a construction and led development of the detailed proof.
The scientific claim is narrower than solving particle physics, but still consequential. The preprint proposes a way to define a difficult part of the Standard Model on a spacetime grid without relying on interactions being small. Questions about physical consistency, the limit of vanishing grid spacing and the proof’s acceptance remain open.
The useful distinction is between constructing a candidate mathematical definition and establishing a complete, physically satisfactory theory. Craig’s paper advances the first task.
Why the Standard Model Needs a Lattice Definition
The Standard Model describes elementary particles and three fundamental forces: electromagnetism, the strong force and the weak force. Its extraordinary experimental success can make the remaining mathematical problem sound surprising.
Many predictions use perturbation theory. Physicists begin with a simpler system whose interactions are switched off, then restore those interactions through successive corrections. When the corrections behave well, this produces exceptionally accurate answers.
Some important phenomena depend on interactions being present from the beginning. Craig uses the proton’s mass as an example: much of it comes from the motion and interactions of quarks and gluons. Switch off the strong force, and the proton itself falls apart. Expanding around that starting point cannot adequately describe the bound object. His OpenAI account explains this motivation for a nonperturbative approach.
A lattice replaces continuous spacetime with a grid. The distance between neighboring points, called the lattice spacing, sets the shortest distance the formulation can resolve. At a fixed spacing, physicists can seek a precise definition with interactions fully included.
Lattice methods have already enabled major advances in calculations involving the strong force, including understanding the proton’s mass from first principles. Extending this success to the entire Standard Model runs into an additional obstacle: the weak force distinguishes between left-handed and right-handed fermions.