AI tools eroding code review: team reads only 5% of PRs
Upset-Day9099 · reddit · 2026-08-30
The author's team merged 214 PRs last month but only properly read about 12. The rest were approved after skimming summaries and glancing at tests. Review capacity failed to scale with output.
Tooling made it weirder: Coderabbit catches enough real bugs to make trusting it feel rational, but that trust is exactly why they read less every month. The author wonders where this lands—reviewing only auth/money paths? Nothing? They draw a parallel to compilers replacing hand-checked assembly and ask where other teams draw the line today.
More from coding & agent
- Perplexity open-sources SPACE, the sandbox platform powering its Computer agents — AravSrinivas · 2026-10-16
- Vercel makes Python generally available: FastAPI, Django, Fluid Compute support — steren · 2026-10-16
- Full course: build a playable Unity fishing game with Claude Opus 5.5 in six steps — Roger_M_Taylor · 2026-10-16
- Armin Ronacher's kid vibe-coded multiplayer with AI, teaching dad about PeerJS — mitsuhiko · 2026-10-16
- Marc Lou praises PostShiba, a cute indie alternative to AWS SES for email infra — marclou · 2026-10-16
- RExBench: best coding agent completes only 33% of autonomous research extensions — najoungkim · 2026-10-16