Hi HN, Levi (poster) here. I wrote this post detailing the “plot promise” system I’m developing at Varu AI. I found traditional outlining methods limiting for the flexible, very long-form AI stories I wanted to generate. This system uses dynamic narrative threads instead.

Curious to hear thoughts on this approach vs outlines, and especially interested in discussion around the challenges mentioned, like maintaining long-range coherence with current LLM context limits. Happy to answer any questions!

Do you have any examples of stories that an LLM has generated using your techniques?

Yes, on the home page, you can see them. They're unedited:

https://www.varu.us/