A small machine repeats a bounded action: count eleven pulses, exchange one fragment, rotate a page, remove every third event. The rule is not valuable because it is clever. It is valuable because it keeps preference from settling every question too early.
Inside a constraint, doubt gains duration. Instead of asking whether the whole work is good, we can ask what this operation reveals, what it conceals, and when its consequences become predictable. The machine gives uncertainty somewhere practical to live.
Eventually the rule will become obedient. Its outputs will acquire a house style and surprise will turn into branding. That is the moment to interrupt the machine, not worship its consistency. A constraint is a temporary collaborator. Its best result may be the precise reason we decide to disobey it.
Synthetic local demo content for interface evaluation. It is not a human contribution, editorial decision, or production publication.