Skip to main content

0043: Make Route Utility Rejection Terminal

Status: Accepted

Date: 2026-07-22

Context​

The Circussy visual-rehearsal experiment had already produced working capture, motion, and contact-sheet machinery when the owner concluded that the detached workflow was not useful. A native game or editor Almanac, if independently valuable, would have been simpler and more truthful. Earlier feedback had been treated as requests to improve the detached output, so investment in the route made further polishing appear productive even after its utility was doubtful.

Existing terminal stop semantics cover unavailable evidence, missing authority, unresolved decisions, and blocked operations. They do not clearly distinguish rejection of an implementation from rejection of the route's purpose. Without that distinction, an agent can translate "this workflow is useless" into "make this workflow better."

Decision​

  • Explicit human rejection of a workflow, evidence surface, artifact, or supporting system's utility terminates optimization of that route.
  • Tuning remains valid when the human preserves the route and asks for a better implementation.
  • After route utility rejection, an agent may preserve state and salvage only parts the human explicitly accepts.
  • The remaining route is retired, reverted, or materially reframed through a different project-native path.
  • Narrowing, renaming, polishing, or repackaging the same rejected route requires new explicit human direction.
  • When feedback is genuinely ambiguous between tuning and utility rejection, the agent asks one concise discriminating question before continuing.

Consequences​

  • Human product judgment can stop technically successful but unwanted work.
  • Sunk cost is not treated as evidence of utility.
  • Useful implementation fragments and lessons can survive without preserving the rejected workflow as a product or method requirement.
  • Agents must recognize a second terminal condition at judgment time, not only blockers encountered during implementation.

Evidence Boundary​

The decision is motivated by one long-running owner workflow and one rejected visual-rehearsal route. It generalizes a human-authority and sunk-cost control problem; it does not claim that every negative reaction rejects the route. Ordinary quality feedback remains tuning unless the human rejects the route's usefulness or the distinction requires clarification.

Last updated on