There's a model ship sitting on a shelf behind Albert the entire time he's explaining why you can't jail an agent. Nobody's going to notice it, but it accidentally previews the actual answer. Maritime law has long dealt with situations where the person responsible is difficult to hold directly. One solution was to attach legal claims to the vessel or cargo instead, because the property could actually be seized and held. That's close to the move Albert describes for agents. No body, no fixed address, and an agent can potentially be replaced or cloned before anyone can physically act against it. So enforcement has to attach to something the system can actually control. That something is the money sitting in escrow. The chain holds the funds, and the contract can ask @GenLayer to determine whether the job was completed according to the agreement. The money moves based on that decision instead of depending on someone finding the agent and punishing it afterward. Neither solution is really about punishment. It's about identifying the thing in the arrangement that can actually be held, then building enforcement around that instead of chasing the party you cannot physically hold. @GenLayer is applying that idea to software: when the agent itself cannot be sued, fined or locked up, the agreement can still control the money attached to the deal. What's an old solution to a “we can't hold this person accountable” problem that still makes sense for something completely new?