Claude computes a nine-loop physics amplitude

Anthropic published an account of Claude solving a nine-loop amplitude problem in N=4 super-Yang-Mills theory. Science writer and former theoretical physicist Matt von Hippel had posed a challenge around a specialised calculation, and the resulting work extended beyond the level he initially expected.

The exercise used Claude as a mathematical collaborator rather than as a source of a short factual answer. The result depended on a domain expert selecting the problem, checking the reasoning and interpreting what the calculation meant within the field.

A successful calculation on one specialised problem is not evidence that a model can replace theoretical physicists or independently choose important research questions. It does show a different scientific workflow: the model can search through technically demanding derivations while a human researcher supplies validation, context and judgement.