Claude solves a physics equation that had never been solved before.

I presented Claude Fable 5.1 with a theoretical physics challenge: Compute a ninth-order amplitude for N=4 super Yang-Mills theory. Typically these calculations go up to second, third, fourth, or fifth orders. The model took several days to write some Python code using the SymPy library. There were two solutions. Lance Dixon, a physicist at SLAC, verified the solution and agreed. This took somewhere between one and two thousand dollars in computation time, which corresponds to approximately a week’s worth of computing power from 96 CPUs.

When I need to do something very difficult, I do it myself with Fable 5.1. In this case, it solved a new calculation, but that’s not the main point. The important thing is that a physicist verified the answer before I accepted it. If there were no verification, then I wouldn’t believe any brilliant-looking answer.