Witten Credits Claude Opus 5.5 for Key Construction in 33-Page Solo Paper

A first in history!

Witten, the "pope" of physics and the only physicist to have won the Fields Medal, the highest honor in mathematics, has publicly thanked AI.

In the acknowledgments of his recent preprint paper, Witten wrote:

Many details were assisted by Claude Opus 5.5.

Many details were assisted by Claude Opus 5.5.

Even more striking is that in the body of the paper, Witten admits: Claude Opus 5.5 proposed a key approach.

Who is Witten?

Witten is a legend of contemporary physics.

For decades, he almost single-handedly unified the five different formulations of string theory, founding M-theory.

In 1990, he won the jewel in the crown of mathematics—the Fields Medal—for his pioneering work in topological quantum field theory and supersymmetry.

He was a tenured professor at the Institute for Advanced Study in Princeton, a position once held by Einstein and Gödel.

A fellow physicist once joked:

If Witten says a formula is correct, even if no one understands it, everyone will assume it is correct and then spend five years figuring out why.

His academic style is classical and pure:

Paper, pen, extremely abstract differential geometry, deep physical intuition.

This titan of physics, now in his eighties, wrote a 33-page solo paper, yet stated plainly in the body: this mathematical construction was not my idea—it was given by AI.

Such is the bearing of a grand master, lofty as mountains and long as rivers.

Five months: AI shifted its position in Witten's work

In June, Witten was studying "Entanglement Wedge Reconstruction" in holographic theory, which involves a large number of higher-order perturbative matrix calculations.

Witten encountered a complicated algebraic generalization. He turned to Claude Opus 4.8.

But Witten put this matter in a footnote.

Perhaps this is what most people felt at the time: AI, though extremely helpful, was a "handyman assistant."

By November, a new version of the paper went online, and this time Claude squeezed directly from the footnote into the core of the argument.

More critical is the change in verb.

Witten did not write "Claude calculated" or "Claude generated."

What he used was:

Claude Opus 5.5 suggested.

When a physicist writes "X suggested that..." in a paper, X is usually the name of a collaborator, or a peer who gave him inspiration.

In five months, Claude's role went from "helped" to "suggested," from "calculating for me" to "proposing an approach."

On humanity's purest and most profound high ground of theoretical physics, AI's identity quietly completed a qualitative transformation.

What is Witten researching? A ghost that has haunted physics for forty years

The subject Witten's paper tackles is one of the hardest bones in high-energy physics: the lattice formulation of chiral fermions.

Let us break it down into extremely plain logic.

Our universe is made of microscopic particles. These particles obey quantum field theory.

Particles have "chirality," like a person's left and right hands: mirror-symmetric, yet unable to coincide.

The fundamental laws of our real world—such as the "weak interaction" responsible for nuclear decay—have a fatal quirk: it only plays with left-handed particles and ignores right-handed ones.

Physics calls such a theory a "Chiral Gauge Theory." Our Standard Model of particle physics is essentially such a theory.

But when studying quantum field theory in continuous spacetime, calculations often produce "infinities."

To understand the underlying properties of quantum field theory, physicists came up with an idea: cut spacetime into a grid.

Like the pixels on a computer screen. Particles live on the lattice sites, and forces live on the links connecting the sites.

This method is called Lattice Gauge Theory.

As long as spacetime is turned into a discrete grid, all physical quantities can be calculated directly on a supercomputer.

But disaster followed.

In 1981, two physicists proposed the famous Nielsen-Ninomiya Theorem:

On a discrete lattice, you cannot simultaneously preserve locality, preserve spacetime symmetry, and have only a single chirality of fermions.

Whenever you try to place a left-handed particle on the lattice, the grid automatically "copies" an extra right-handed particle deep within.

This phenomenon is called "Fermion Doubling."

It completely destroys chiral theory.

For over forty years, generation after generation of top physicists have pressed forward, attempting to perfectly define chiral gauge theory on the lattice.

People tried every approach imaginable: Domain-Wall Fermions, Overlap Fermions... each scheme was extremely intricate and came with various mathematical compromises and costs.

Witten's new 33-page work is precisely an attempt to find a completely new, more thorough non-perturbative definition in this extremely difficult field.

Claude: The Proposer of the Core Mathematical Construction

This time, Witten completely shattered people's cognitive inertia.

In the process of establishing a mapping between discrete lattices and continuous spacetime, there exists an extremely demanding mathematical barrier: Gauge Field Interpolation.

Simply put, on a spacetime grid there are only discrete line segments (Link Variables).

But physical calculations require a smooth continuous space (Continuous Gauge Field).

You must find a mathematical method to smoothly "stitch" discrete jumps into a continuous field.

This stitching process must simultaneously satisfy two nearly self-contradictory conditions:

1. Strict symmetry: The gauge symmetry of the lattice must be precisely preserved, without producing any mathematical defects;
2. Holonomy Matching: The result of path integration along lattice paths must strictly equal the link variables on the discrete grid.

With the slightest mishandling, the mathematical self-consistency of the entire theory would collapse on the spot.

It is precisely here that Witten explicitly documented the role of AI in the main text.

Faced with the contradiction of "being smooth everywhere while strictly locking down holonomy" when stitching interpolations across boundaries of different dimensions, Claude Opus 5.5 proposed a key variant algorithm to Witten: introducing a Boundary-Collapsing Map under coordinate pullback mapping.

Claude's specific construction is as follows:

First, establish lattice coordinates with integer points as vertices in four-dimensional Euclidean space. Take a sufficiently small constant δ>0.

Claude proposed constructing a special smooth monotonically non-decreasing scalar function s(t).

This function satisfies two properties:

1. Translational periodicity: s(t+1)=s(t)+1;
2. Within the δ-neighborhood around each integer lattice point n, the function becomes a completely flat constant: when ∣t−n∣<δ, s(t)=n. Outside the neighborhood, it smoothly maps the interval [n,n+1] to itself.

Based on this scalar function, Claude constructed a four-dimensional space mapping:

This mapping produces a striking geometric effect:

It maps each cell in the lattice to itself, but completely and smoothly "flattens and collapses" the elongated neighborhood of thickness δ around the cell boundary onto the geometric boundary of the cell!

The reason traditional direct interpolation shatters at boundary seams is that when adjacent hypercubes transition inward, the normal derivatives cannot match.

Claude's algorithm flow completely restructures the interpolation order:

1. Recursion from lower to higher dimensions: Physicists first define the gauge field A on 1-dimensional edges, making its holonomy precisely match the lattice variables;
2. Execute "pullback surgery" before dimension lifting: Before pushing the field to 2-dimensional faces, 3-dimensional volumes, and 4-dimensional hypercubes, Witten adopted Claude's core suggestion—first use the mapping Φ to perform differential-form pullback replacement on the field over lower-dimensional cells, safely solving the Dirichlet problem: since Φ flattens the entire δ-neighborhood near the boundary into a constant projection, all normal derivatives of the pulled-back field near the boundary smoothly vanish!

This single modification instantly resolves all contradictions: because all derivatives are zero within the boundary buffer zone, when physicists subsequently solve the Dirichlet boundary value problem under Landau gauge inside the hypercube, the solution naturally acquires infinite-order smoothness when crossing the seams between different hypercubes.

More critically, along any edge, the pullback mapping Φ merely amounts to reparameterizing the line segment, without changing the value of the path integral along the line in the slightest.

Holonomy is precisely conserved, causal locality is locked within boundaries, and infinite-order smoothness holds everywhere in the entire space.

Witten examined and adopted this variant scheme in detail in the paper, and in Section 3 of the main text used it as a key cornerstone for constructing the non-perturbative chiral fermion partition function.

The underlying paradigm of scientific research is being completely rewritten. This fact is plainly evident.

We once believed that in theoretical physics—the jewel in the crown of human civilization, requiring extreme abstract thinking and profound physical insight—humans could maintain dominance for at least several more decades.

But reality has proven that the speed of evolution of cognitive tools surpasses all predictions.

References:

https://arxiv.org/abs/2611.05051

https://x.com/jdlichtman/status/2118055176972259335

https://x.com/dmvaldman/status/2118081768125550842

Second-Chasing ASI

⭐Like, repost, and watch—one-click triple connection⭐

Turn on the star to lock in instant New AI Era push notifications!

Open in WeChat