You use LLMs to _discover_ how to approach important problems. You don't necessarily need to use the output verbatim. Same as StackOverflow and Google.