However there's another aspect that irks me, and it's the idea that the prompt was much shorter than the document itself. Well if this is the case, then the problem isn't much the use of LLMs, but rather that you consider obvious that your documents are mostly fluff that can be compressed to a bullet list. If the final document is much harder to verify than the information it contains then it means that you're wasting time and resources to ofbuscate rather than clarify.
