You have a point, the only feedback of an action being taken is from the subsequent inputs, so I suppose I should think of simple actions as micro-models. Thanks. The meta-model of external reality comes in handy when you don't want to over-think it. What else did I miss?