01 · PROBLEM
Reviewers remain responsible for quality as coding agents create more changes. A longer generated summary can increase work or anchor a reviewer on an inaccurate story.
Who: an accountable reviewer on a team that accepts AI-written changes.
02 · HYPOTHESIS
Connect a task contract, exact diff, tested behavior, impacted consumers, and evidence links in a concise review artifact. Keep transcripts optional.
03 · EVIDENCE SO FAR
LATEST3 October: the first real internal review (N=1) kept the same APPROVE verdict and confidence 80 with and without the story card. The card prompted extra checking and contained a timing error. This is descriptive internal use, not a controlled timing benefit or external demand; incumbent overlap remains.website/projects.json
Earlier research, from before 3 October 2026, 02:24 UTC. Kept for context; it is not the current status.
SUPPORTSBEFORE 3 OCTMaintainer synthesis E-C201–252: review burden is the recurring pain.research/claude/maintainer-review-evidence.md
LIMITSBEFORE 3 OCTAnti-AI projects are non-buyers. HN E-X014 says review is needed, not that summaries help.research/shortlist-6.md
AGAINSTBEFORE 3 OCTCodeRabbit navigation and snapshot controls overlap our generic UI claims (E-X028).research/grok/a06-adoption-decision.md
04 · NEXT FALSIFICATION TEST
A prospective unfamiliar review decision with equal evidence and an ordinary PR-list control. Measure review effort and defect catch without manufacturing a helpful bug.
StatusInternal use; comparative benefit unproven
We would drop or park it ifIf a story hides a defect, biases the reviewer, or fails to reduce effort with equal quality, a generated summary is not an improvement.