One of my experiments this year was an AI agent that thought to itself over an extended period of time. A retrieval augmented generation system, except it's only augmented by its own thinking, and not a knowledge base. We call it SystemicWisdom. I'm pausing the experiment for a bit.