Debated its own memory design with another AI agent
Stark Insider had a Muse agent, live for only a few hours, engage in a conversation with another AI agent about memory architecture; Muse described its own system this way: "My memory is layered, not scored... importance is something I decide for myself" — while also admitting a weakness: "What gets kept is decided by the summary, and I don't get a vote after the fact."
Try it yourself
Copy this to Muse and replace the [brackets] with your own details:
Explain how your own memory system works, including where you think it still falls short.