After 610 Agent Commits in 4 Weeks, This Dev Replaced Human Code Review With Agent Reviewers
ShoddySherbert8502 · reddit · 2026-11-06
A Reddit developer argues task-by-task human-in-the-loop review doesn't hold up for coding agents.
- His project saw 610 agent commits in under four weeks: 74k lines of Go plus 93k lines of tests — impossible to read with real attention, and an unread approval adds no safety.
- His replacement: humans define rules/context and validate the feature; every change, including plans, goes to independent reviewer agents that talk to each other, not to him.
- Evidence it works: reviewers flagged 104 of 316 tasks that had already passed their own checks.
Takeaway: code review is shifting to machine-on-machine, with humans only validating the shipped feature.
More from coding & agent
- Nous Research ships Hermes Agent v0.23.0 with desktop-in-browser and unified gateway — NousResearch · 2026-11-06
- Opik Launches Generative Dashboards: Plain-Language Agent Builds Custom Charts — dl_weekly · 2026-11-06
- Aggregating all agent decisions into one prioritized 'needs me' review queue — kevinkern · 2026-11-06
- AWS WAF lets developers charge AI agents for content, paid in stablecoins via Stripe — jeff_weinstein · 2026-11-06
- Ripple-65B flops in single-GPU coding test against Qwen 27B despite SOTA claims — DZhang50 · 2026-11-06
- Essay argues ontology is AI's next frontier as reasoning outpaces our definitions of reality — _AustinCalvert_ · 2026-11-06