What is difference between Memory from Custom Instructions

I assume that Custom Instructions are meant to be manually added by users when their requests are very clear and explicit. Each round of questioning merges the Custom Instructions into the Prompt.

Memory, on the other hand, is acquired through conversation, as requested by the user to remember. The use of Memory is decided by GPT in conjunction with the context, and is an agent behavior.

I’ve drawn a flowchart, but I’m not sure if it’s correct.

In my sense, Memory acts as a checkpoint store of information over time with interactions. Instructions on the other hand are immediate directives tuned for response.

It would be really cool if there was a way to get draft memories from all past interactions, which you can audit to add/remove chatgpt memories.

1 Like

you mean like memory summary and key point selection would be better?