An AI has completed a physics calculation no human has finished. Here is what “nine loops” actually means and why physicists care.

In quantum field theory, predictions are built from sums of possible interactions. Physicists expand these as “loops”, rounds of quantum corrections. Each extra loop adds complexity that multiplies, like approximating a curve with ever more polynomial terms.

The nine-loop result was the six-particle hexagon amplitude in planar N=4 super Yang-Mills, a clean model physicists use as a sandbox. The human record sat at eight loops, set in 2023 by Lance Dixon’s team.

The AI, running inside Anthropic’s Claude Science platform, got the answer two independent ways: a bootstrap and a form-factor route. Dixon himself verified it was correct, Big Think reported.

The setup is famously fragile. One wrong step in the recipe and the whole thing collapses. The AI built all its code from scratch over about 25 days for a few thousand dollars of compute.

Why it matters: machines can now push past limits in pure theory on their own. But the AI found no new physics, used existing methods, and worked a problem others had nearly solved. A tool milestone, not a theory revolution.