autonomous LLMs are more interesting because they have agency over their memories.
we are the sum total of our memories in the same way autonomous LLMs are.
even if that machinery is ultimately matr...
They don't have agency over their training data, you do. Their context can only store limited md files or vector db fetches.