I’ve built on this a little. Currently, I have in my database large collections of notes on certain general topics, and I’m having a hard time determining how to reorganize them. (I’m getting a little lost in the “organization emerging from the bottom-up” process because there are so many.) And in a couple cases, I’m trying to write books from the content, but, again, struggling with knowing how best to organize them. But I discovered you can build up prompts successively, and I’m using that to get ideas about ways to organize the information.
So I start by using your prompt, and then I feed it all the notes I have on the topic. (I don’t do it as just one note per submission; I send them in batches, but I have to do it in batches because it can only accept so many words.) So the first time it will spit out atomic note ideas, I’ll enter a new batch and it will provide more atomic note ideas, and so on. (The last one I did gave me a total of 33 note ideas.)
THEN, I follow that with this prompt:
Based on the entire series of inputs and generated atomic note ideas in this series, can you identify the top five most important ideas? And can you also summarize the topic in about three paragraphs? And can you summarize it further in a single paragraph?
It gives me the biggest ideas. I’ve only done it a couple times so far, but it seems to give me what is essentially an annotated bulleted list. And two levels of summary. (Sometimes it struggles a little with this and needs to be reprompted to create a summary that’s longer or shorter if it didn’t offer much variation in the length of the first pair.)
THEN, I give it this:
Again, based on the entire series of inputs and generated atomic note ideas, please provide an outline for a book based on this content.
And it outlines the whole book for me. Sometimes it knocks it out of the park; sometimes its suggested organization is a little…weird. But it’s pretty easy to either use the regenerate button, or to ask it “Can you provide an alternative outline for a book based on this content, still based on the entire series of inputs and generated atomic note ideas?”
So when I’m done with all this, if I only generate one outline, I’ve had it give me FIVE different ways of looking at a pretty large collection of notes:
- A batch of atomic note skeletons, pulling out significant ideas I may or may not have realized were there
- A set of the top 5 main ideas from the whole body of content
- A longer summary
- A shorter summary
- An outline of the whole body of content