The system map: nodes, links, and actors

You cannot govern what you have not named. The PAN framework — PAN stands for Policy Actor Network — draws a deployment as a network of nodes and the links between them. The nodes are the component types a real deployment contains: not just the neural network, but the harness around it, the retrieval layer that feeds it, the guardrails that filter it, the memory and record stores it touches, the network of human and automated operators acting on its outputs, and the telemetry and update pipelines that quietly reshape it over time. The links are the connections along which error, influence, and authority move between them — and both the nodes and the links can be extended as a real deployment demands.

The second half of the map is just as important: the actors. Developers, harness builders, and deploying organizations are the obvious ones. But real governance histories are full of oversight boards, independent auditors, courts, commissions of inquiry, legislatures, vendors, and data-protection officers — actors who, in documented cases, were the ones who actually changed the system's trajectory.

Mapping actors matters because every governance action needs someone with the authority to take it. A safeguard nobody is empowered to enforce is a description, not a control. The Practice Library tags every governance pattern with the actors who can pull it for exactly this reason.

When you read a case file in the Domain Atlas, try reading it against this network: which nodes and links carried the failure, and which actors — often courts and commissions, late — finally acted.