I find the memory feature stores a lot of short term relevance information, like ideas I’m bouncing around, that have to be reviewed frequently to conserve space.
I’d like to have a way to more automatically separate this kind of memory for easier management of storage. Short term relevance could be deleted automatically.
Also, the relationship between memories and custom instructions isn’t clear to me. The kind of memories I want to keep long term may be better kept there, but there isn’t enough space.
Anyone have any insight on working with these features more effectively?
I’m interested to hear how others think about the context your memories provide to the way you work with ChatGPT.