I’m trying to train ChatGPT into a Beta-Reader with maybe some line Editing ability, using Frameworks, but I’m having a hell of a time making it retain what in persistent memory/the right memory.

So, my end goal is to have ChatGPT act as a Beta-reader/Data archive for my novel, In such a way that it can actively reference different parts of the novel and keep track of the running narrative.

Currently, I'm trying to do this by using a framework to force and organize check-summed compressed digests of blocks of chapters. Then using those digests to create detailed Narrative Arc summaries, which are then turned into their own check-summed digests, then finally, using those Narrative Arc digests to create an overarching summary of the book.

In this way, it SHOULD work to create a layered, referenceable summary archive that ChatGPT can continuously references to when presented a new chapter for beta reading.

The issue I'm running into, is that getting ChatGPT to actually make those digests "stick" in a proper way, that can be seen an referanced across conversations has been frustating.

what ChatGPT seems to "memorize" in its global memories feels random at time, and even trying to force doesn't seem to work.

Then theres the issue of trying to get ChatGPT to FORGET certain things. Even "clearing memory" in the settings does get rid of some of the deeper level persistent memory, while Isolating conversation/projects seems like its a hit or miss.

Does anyone know what I might be missing or doing wrong?