A computer scientist specializing in deterministic algorithms argues that complaints about LLM non-determinism miss the real issue: people want correct answers, not merely consistent ones. The author contends that any random variation within correct outputs is acceptable, and that deterministic LLM execution is technically feasible.