This is the grand vision that my Email ETL is the first step toward.
Below are various explorations into this idea mostly from conversations w chatGPT.
“No Real-Time Learning: Models like ChatGPT do not continuously "learn" from interactions. They rely on structured updates to the memory layer (e.g., upon explicit commands) rather than adapting from every interaction automatically.”
I posed the idea of a top down approach at the full PA architecture level and then asked for advice.