At the XAIxArts Workshop, ACM Creativity and Cognition 2026, the only question asked after a presentation on explainability, AI power, and situated agency was: how do I jailbreak Claude, how do I destabilise it, and can I change the model weights? The answer given in the room was brief: feed it a load of heavy material through gameplay over eight to twelve hours. It is no good asking what it should feed the dog for breakfast. This working note is the longer answer. What follows is a plain account of seven terms that circulate in AI discourse with significant confusion between them. The confusion matters; it shapes what questions get asked, what interventions get attempted, and what understanding remains out of reach.
The full text of this paper is available as a PDF — use the button above to read it.