One of the biggest problems with hands off LLM writing (for long horizon stuff like novels) is that you can't really give them any details of your story because they get absolutely neurotic with it.
Imagine for instance you give the LLM the profile of the love interest for your epic fantasy, it will almost always have the main character meeting them within 3 pages (usually page 1) which is of course absolutely nonsensical pacing. No attempt to tell it otherwise changes anything.
This is the first model that after 19 pages generated so far resembles anything like normal pacing even with a TON of details. I've never felt the need to generate anywhere near this much. Extremely impressed.
Imagine for instance you give the LLM the profile of the love interest for your epic fantasy, it will almost always have the main character meeting them within 3 pages (usually page 1) which is of course absolutely nonsensical pacing. No attempt to tell it otherwise changes anything.
This is the first model that after 19 pages generated so far resembles anything like normal pacing even with a TON of details. I've never felt the need to generate anywhere near this much. Extremely impressed.
Edit: Sharing it - https://aistudio.google.com/app/prompts?state=%7B%22ids%22:%...
with pastebin - https://pastebin.com/aiWuYcrF