
What happened
A one-person contractor wired Claude Code's /code-review into a pre-delivery script, and review execution went from "under half the time, zero in busy weeks" to 100%, with each review logged.
Why it matters
The gain he credits is structural, not detection quality — going from zero reviews to always one is treated as mattering more to accident rates than cutting an hour to five minutes.
What to watch
/code-review flags code that is broken, not code that contradicts the spec, so passing it is not proof of correctness. He tested three past production bugs and it caught two.
WHO IT HITSSolo freelancers and small contract-dev shops who ship code without a second reviewer are the clearest fit, since the gate forces a review record before delivery. Teams already using AI coding assistants may also read it as a caution against shipping auto-fixed diffs they did not personally inspect.
Summaries like this, in your inbox every morning.
The author runs contract projects entirely alone — writing the code and reviewing it himself — and describes the failure mode he feared as shipping code he only believed he had reviewed. His earlier countermeasure was simply "being careful", which he says failed because care is not a system, and it broke first during busy periods.
What he changed was the shape of the workflow rather than the tool's raw capability. Reviews run non-interactively with write permission withheld, results are written to a file so a record exists even when nothing is flagged, and a high-severity finding stops delivery instead of merely sending a Slack notice. He separates the two roles explicitly: the AI review holds the authority to block a delivery but not the authority to approve one, and spec conformance, client explanations and the decision to apply fixes stay with him.
The stakes he identifies hinge on that division. He flags a dangerous side effect — starting to assume that a passing review means the code is fine — and counters it by putting "spec check is separate" into every notification. Whether the setup holds up appears to depend less on the tool's accuracy than on his keeping the human side of the split, since the one bug among three that it missed was a spec misunderstanding.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Chunghwa Telecom president Rong-shy Lin said the shift from conversational AI to autonomous AI agents is resha…

Apple and Meta said they would jointly release a new open-weight AI model designed to run on high-end Macs, ai…

A writer and entrepreneur opened his daily to-do list and found 156 items, and concluded that an AI assistant…

A Zenn contributor sorted AI use into three modes — AI as an assistant, AI running alongside people, and AI th…

A non-engineer operator says one person ships four smartphone apps on Google Play, three on the App Store, wit…

A developer released agent-guardrails, a one-script Python standard-library plugin that guards Bash, Write, Ed…
