Disclosure: this article contains affiliate links. If you buy through them we may earn a commission at no additional cost to you. Our recommendations are based on the evidence recorded alongside this article, and prices were verified on the date shown.

A Hacker News poster going by stikit asked a question a lot of small dev teams are quietly asking themselves: does a human still need to review your code, or has that job moved to AI? The verdict from the thread isn't universal — it's one team's account — but it's concrete enough to be useful if you're deciding whether to change your own review process.

What changed and why it matters

The thread, titled "Ask HN: Does a human still review your code?", was posted by user stikit to ask whether human code review remains necessary now that AI tools can do a version of it. The poster's own answer is the interesting part: their team moved from a human-driven review process to one that's now largely automated.

What pushed them off human review wasn't a belief that humans are bad at it. It was the cost of the old process — code got rewritten multiple times, often for what felt like arbitrary adherence to one reviewer's idea of best practice, and the poster describes that cycle as painful. The replacement is built around Claude's /review and /code-review commands, which the poster calls invaluable. Worth noting, per the poster's own description: this covers one small company's internal tooling plus personal projects.

Hands-on findings

The thread itself doesn't include benchmarks or before/after metrics — it's a first-person account, not a controlled test, so treat it as one practitioner's experience rather than a verified performance claim. What it does describe clearly is a shift in kind: reviews that used to require a person to read a diff and argue about style now run through Claude's review commands as a standing step in the workflow, and the poster frames that as a meaningful improvement over the rewrite cycle they had before.

The honest caveat is that "largely automated" isn't "fully automated," and the poster doesn't claim to have removed human judgment from the loop entirely — just the repeated, subjective rewrite pain that made the old process slow. If your team is evaluating a similar move, the practical question isn't whether an AI reviewer can catch bugs; it's whether it removes the specific friction you actually have, which for this poster was style-driven rework, not correctness.

If part of your evaluation includes where the reviewed code actually runs and how much headroom your hosting plan gives a growing team, it's worth doing that homework alongside the workflow change. You can Check current price for Vercel Enterprise to see what's included at the tier built for teams operating past a single small project, which is useful context if your automated-review setup is scaling alongside your deployment needs.

Pricing and plans

Vercel lists Pro and Enterprise as its team-oriented tiers on its pricing page. Both pages exist and were checked directly as part of preparing this piece, but the specific dollar figures and included limits are not reproduced here — pricing pages change, and quoting a number today risks being wrong by the time you read this.

That's a deliberate choice, not an omission: rather than repeat a figure that might already be stale, the responsible move is to point you at the source and let you read the current terms yourself.

For teams past the free tier, Check current price for Vercel Pro shows exactly what's included at that level today, which is the only way to know if it matches what your team actually needs versus what a stale screenshot might suggest.

Who it suits and who it does not

The HN thread is a useful data point specifically because it's narrow. The poster's setup — a small company's internal tooling plus personal projects — is not the same population as a regulated enterprise codebase with external audit requirements, and the thread doesn't claim otherwise.

Best for:

• Small teams currently stuck in a human-review cycle defined by repeated rewrites over subjective style preferences, matching the exact pain point the original poster described.
• Solo developers and personal-project maintainers who don't have a second engineer available to review every change.
• Internal tooling at small companies, where the poster's own automated setup is actually running.

Skip it if:

• Your review process exists primarily for compliance or audit sign-off, which is outside what this thread's account addresses.
• You're at enterprise scale with a codebase and team size the original poster explicitly wasn't describing.
• Your current review pain isn't about rewrite cycles or subjective style disputes — the poster's account is specific to that friction, not a general claim that automation fixes every review complaint.

Verdict

The strongest thing this thread supports is narrow but real: for a small team drowning in style-driven rewrite cycles, moving to a largely automated review process built around Claude's review commands solved the specific problem the poster had, in the specific context of internal tooling and personal projects — it is not evidence that human review is obsolete everywhere, and the account itself doesn't claim that. If you're weighing a similar change, match your own pain point against theirs before assuming the fix transfers, and if hosting cost is part of that budget conversation, See current pricing at the merchant so you're deciding against current numbers rather than a number someone quoted months ago.