[Veuxdo]

They're computing things they've precisely been programmed to compute. Number crunching, in the end, with extremely low error rates caused by hardware faults. LLM error rates are comparatively astronomical.