Commenting and Approving Pull Requests

· ai-agents · Source ↗

TLDR

  • Approve PRs while simultaneously leaving non-blocking comments, using Conventional Comments labels to signal intent and keep code moving.

Key Takeaways

  • Conventional Comments labels – nitpick:, suggestion:, question:, issue (non-blocking): – clarify intent at review time without requiring a follow-up round.
  • Approval-reset-on-commit and auto-merge repo configs can undermine this workflow; verify settings before adopting.
  • Linters, auto-formatters, type checkers, and security scanners in CI reduce trivial comments, keeping approvals signal-dense.
  • Finding a blocking issue at code review often signals upstream misalignment on startups and SMBs, not just a code problem.
  • The goal is a highly aligned team where most feedback is non-blocking, making comment-and-approve the default rather than the exception.

Hacker News Comment Review

  • The sharpest pushback is about review latency: the difference between a 2-hour and 23-hour turnaround is significant, and a second back-and-forth round compounds the cost.
  • Azure DevOps has a native “Approve with suggestions” state that mirrors this workflow, though whether comments get read post-approval is an open question.
  • Commenters agree the approach works well with experienced engineers and strong test coverage, but scales less reliably with junior engineers who may not act on unblocking comments.

Notable Comments

  • @singron: “there is a world of difference between getting a review within 2 hours and 23 hours” – frames review latency as the true hidden cost, not approval state.

Original | Discuss on HN