Everyone calls everything an agent now. Tyler Folkman's distinction is the first one I have heard that actually cuts. An agent makes decisions on its own. A prompt waits for you. That is not about the model. Same Claude, same context window. The difference is whether you handed over a task with a way to verify itself, or whether you are sitting there deciding what the next message should be. Most people are still in the second mode. He describes how ChatGPT trained everyone into it.…