Edward Witten of the Institute for Advanced Study has credited Anthropic's Claude Opus 5.5 in a new theoretical physics paper, naming the AI model three times in a 36-page manuscript posted to arXiv on November 3, 2026. The paper, "A Note On Lattice Definition of Gauge Theories With Chiral Fermions" (arXiv:2611.05051), is Witten's sole-authored attempt at one of quantum field theory's oldest unsolved problems: how to define four-dimensional gauge theories that include chiral fermions using lattice-based methods.
The credits are unusually specific. In Section 3, "Continuum Interpolations of Admissible Lattice Gauge Fields", Witten writes that "Anthropic's Claude Opus 5.5 suggested a likely variant" of a mathematical construction, and a few sentences later that "Opus 5.5 also described a variant in which the inductive step is made by a sort of linear interpolation." In the acknowledgements, he adds: "Assistance on many details was provided by Claude Opus 5.5."
What Claude actually suggested
The construction at the heart of the credit is a way to smooth a lattice gauge field into a continuous field so that its link holonomies, the quantities that encode how the field twists around the lattice, equal the lattice link variables exactly, while the field inside each small cell of the lattice depends only on the links on that cell's boundary. Witten writes that the resulting construction is gauge-invariant, preserves all lattice symmetries, and "reproduces the link variables exactly".
The problem matters because physicists compute the strong force by putting spacetime on a grid, the lattice gauge theory Kenneth Wilson introduced in 1974. Chiral fermions, particles that, like those in the Standard Model, couple differently to left- and right-handed forces, have resisted a clean lattice definition for decades. A rigorous construction would let physicists study the Standard Model's foundations without relying on approximations.
Three papers, one old problem
Witten's paper is the third recent attempt at the same problem, and AI fingerprints are on more than one of them. The first was by particle theorist Nathaniel Craig, whose 123-page preprint (arXiv:2610.09093) was built with heavy use of OpenAI's GPT-6 Astra, which his acknowledgement calls "essential to this result". The second came from Stephen Hsu (arXiv:2610.29019), who returned to an approach he first proposed in 1995. Witten thanks physicist T. Hartman for flagging Craig's paper, and K. Yonekura for comments on the manuscript.
The reaction in the physics community was swift. Cosmologist Will Kinney posted a screenshot of Witten's acknowledgement with the one-word caption "Witten", and within three hours it had passed 12,000 views, a sign of how closely physicists are watching AI's move from chatbot to collaborator.
A collaborator, not yet a co-author
The credits come with caveats. The paper is a preprint and has not been peer-reviewed, and it does not explain exactly how Witten used Claude beyond the three credited passages, so the scale of the model's contribution beyond those passages is not documented. Witten's own contribution is unmistakably his: he reaches the same destination by a slightly different road, coupling the interpolated field to continuum chiral fermions and widening the construction as far as it will go.
Even so, the moment is notable. A year ago, an AI credit in a paper by the field's most celebrated theorist would have been unthinkable; now it sits in the text of a paper tackling a problem that has resisted solution for half a century. Whether the construction survives peer review is a question for physicists, but the way the paper was written is already part of the story.
