Treating LLM Outputs as Actors Rather Than Deterministic Code

An analysis argues that non-determinism in LLMs is not a limitation if the model is treated as an actor within a system rather than a replacement for code. While the means of obtaining a result can be unpredictable, the final output remains the verifiable component.