The biggest issue I have with these systems is, I don't want a blanket memory. I want everything to be embedded in skills and progressively discovered when they are required.
I've been playing around with doing that with a cron job for a "dream" sequence.
I really want to get them out of main context asap, and where they belong, into skills.
Isn't this the idea behind holographic memory? Chopping the image in half gets you the same image at half the resolution? Or so I've heard...
What you want is a context mipmap.
Then there was the Claude article describing using filesystem hierarchy to organize markdown knowledge, which apparently beats RAG.
[flagged]