Nathaniel Craig used GPT-6 Astra to build a candidate lattice definition of the Standard Model
OpenAI published the first guest post on its new science blog on 15 November. Its author, Nathaniel Craig, leads the particle theory group at the University of California, Santa Barbara, is a permanent member of its Kavli Institute for Theoretical Physics and is a visiting scientist at OpenAI. He describes an October preprint, prepared with GPT-6 Astra, that proposes a candidate definition of the Standard Model on a lattice, a grid that stands in for spacetime.
The problem, as Craig puts it, is that the Standard Model has no nonperturbative definition, one that does not treat interactions as small corrections. Lattice field theory has given such definitions for the strong force, but the weak force treats left-handed and right-handed fermions differently, which makes the Standard Model a chiral theory. Putting a left-handed fermion on a lattice in the most natural way produces an unwanted right-handed partner, a problem known as doubling.
Earlier work removed much of that obstacle. Craig cites Paul Ginsparg and Ken Wilson in 1982, the overlap fermions of Rajamani Narayanan and Herbert Neuberger in 1994, Martin Lüscher's way of separating left- and right-handed overlap fermions, and Yoshio Kikukawa's work on the weak interactions. What remained open was how to fix a phase that depends on the force fields at every grid point. Lüscher solved it for simpler theories such as a chiral version of electromagnetism; for the full Standard Model it stayed open for more than 25 years, according to the post.
Craig's strategy was to represent the same force fields on a sequence of ever finer grids, compare the allowed fermion states on neighbouring grids and add up the pieces to get the phase rule on the original grid. He writes that he could not prove it worked in every case he needed. Astra suggested how to carry the fields onto finer grids, showed that the comparisons converge and took the lead in the detailed proof. The proof shows the rule gives the right correction around every allowed loop, depends mainly on nearby fields, and holds under a bound on field strengths at the grid scale. Lattice theorists Maarten Golterman, Andreas Kronfeld and Yigal Shamir reviewed the proof and asked questions.
Craig calls the result an answer in principle rather than in practice. Experts still have to judge whether it has the required properties; he is examining whether it satisfies reflection positivity at finite lattice spacing, and much work remains before it is useful in realistic simulations.

Why it matters
The post is unusually specific about who did what: the strategy came from the physicist, the construction and most of the proof from the model, and the confidence from outside experts who questioned it. It also states plainly that the result is a candidate that the field has not yet accepted.