@cwebber @cj I think the model works really well for things that are essentially pure optimization problems with no or few externalities like memory management, but we are talking about interfaces that directly impact social systems (privacy). It's why I brought up the crypto example.