A governed research surface: discrete continuation-value pricing, bounded action, and executable evidence.
A scalar shadow price multiplied by total consumption can misprice a discrete action that crosses several marginal-value regions.
Maximizing immediate utility minus this action charge is Bellman-Q selection up to an action-independent constant.
Two editions derived from the published claims and certified evidence: an accessible offline-Piper narration and the locked silent source film. The bounded FabricPC predictor did not clear the gate; the FabricPC intervention recovered Compitum’s missing action-pricing mechanism.
Scene-timed lay narration synthesized locally with Piper 1.4.2. English captions are available in the player.
The locked silent film remains separately available and unchanged.
A certified action change. The exact transition point is not asserted here.
A compact bounded-flow demonstration adapted from Sketched: explicit allowance, forbidden moves, stop, rewind, and replayable trace.
The agent is paused outside the surface.
The presentation reads certificate JSON directly. Narrative and machine evidence remain distinct.
evidence/evidence-index.json…