Writing is fundamentally the transfer of information from your brain to my brain. If you have 1000 bits of information you want to transfer, you can't give 300 bits of information to an LLM and have it fill in the remaining 700, because it doesn't know what those 700 bits are. If it's able to guess those 700 bits correctly, then they aren't true information, and you really only have 300 bits you want to transfer. You might as well transfer those bits to me directly, rather than having the LLM add on an extra superfluous 700 bits that I then have to filter out.

It's not possible to directly transfer 300 bits of information as that would require a form of telepathy. If it takes 1000 bits of English to distribute something that represents your idea properly it is useful to have a tool to create those 1000s bits instead of doing it manually character by character.

Remember when LLMs were entering the mainstream, and everyone shared tips on how to superprompt, and one of the hot tips was to tell the AI to write a prompt itself?

Like:

BAD: "Write the specs for a system that does X Y and Z"

GOOD: "Write me a prompt to write the specs for a system that does X Y and Z."

As if the LLM magically knew all about itself and the best tips and tricks to prompt itself even if it had just came out and there was no scraped information on how to use them yet.