Team of 8 finds AI-written PR reviews reduced to rubber-stamping after 3 bugs slip through
stfufeedmenuggets · reddit · 2026-11-02
A Reddit dev explains that with most of his 8-person team's code now written by Claude Code or Cursor, PR approvals have devolved into scrolling and clicking approve. CodeRabbit catches surface issues, but 3 bugs still hit production last month — a retry wrapper around a non-idempotent payment call, a correct migration that locked a big table for 4 minutes, and an inverted feature-flag check — each approved by two humans who said the code "looked fine." He asks what actually works: author self-explanations before review, smaller PR limits for agent code, or a single accountable reviewer.
More from coding & agent
- Open-sourced: the 6,809-part robotic lobster Claude built, in a single HTML file — atullchaurasia · 2026-11-02
- Claude turns lobster photos into a 6,809-part robotic lobster in one 82 KB HTML file — atullchaurasia · 2026-11-02
- Aaron Levie: everyone becomes a manager of agents as work decouples from hours — inductionheads · 2026-11-02
- Dev's hands-on with Kiro Crew: spec-driven agent workflows yielding hundreds of automated tests — DigitalColmer · 2026-11-02
- ripgrep searches 30k Obsidian notes in 0.15s vs 4s with built-in search — dSebastien · 2026-11-02
- TAWA GUARD Launches Pre-Model Input Checks to Block Hidden Prompt Injections in Agent Inputs — TawaGuard · 2026-11-02