Honestly, parsing through version history is actually something an LLM could handle. It might even make more sense of it than without. For example, if someone replies to a comment and then the parent is edited to say something different. No one will have to waste their time filtering anything.
lvxferre@mander.xyz 8 months ago
They could use an LLM to parse through the version history of all those posts/comments, to use it to train another LLM with it. It sounds like a bad (and expensive, processing time-wise) idea, but it could be done.