What you fail to appreciate is the operation of an LLM is driven by the input data far more than is the case with most programs. Typical programs have a lot of

What you fail to appreciate is the operation of an LLM is driven by the input data far more than is the case with most programs. Typical programs have a lot of business logic that determines their behavior--rules, as you say. E.g., an optimizing compiler has a large number of hand-crafted optimizations that are invoked when code fits the pattern they are intended for. But LLMs don't have programmed cases or rule like that--the core algorithms are input-agnostic. All of the variability of the output is purely a reflection of patterns in the input; the programmers never made any sort of decision like "if this word is seen do this".