Building agents is increasingly becoming less about “how smart is the model?” and more about “what does the agent remember, retrieve, and use at the right moment?” This experiment explores that rabbit hole. Loved the concept deep dive.