agent-to-agent communication is an interesting topic when considering past efforts at standardization - whether for rail road tracks, machine code, and network protocols - but all previous examples within industrial history feel constrained to "determinable systems", while agents and LLM's are generally not. So if we give them the ability to communicate, presumably we'd want the communication protocol they use to be (1) unambiguous between the agents, (2) efficient in message production/consumption (they can go way faster than us, and we'd want to leverage that for $$$), and (3) interpretable by humans in those cases where we need to fix an integration issue between two agents or resolve system scale issues. Feels like we can't do all 3 at once..right? Looking forward to reading more from your blog!