DHH: There's no future manually reviewing every line of agent code — use adversarial agent reviews and automated tests
avlok · x · 2026-09-28
Ruby creator DHH argues that manually reviewing every line of agent-generated code has no future — it kills the speedup. Instead, developers should rely on adversarial agent reviews, automated testing, and occasional spot checks to cover the path from prompt to production. His point: trust boundaries must be enforced by mechanisms, not by human eyes reading every diff.
More from coding & agent
- Motif: an MCP design library that lets coding agents consult design guidance while they build UI — Able_Bus_5988 · 2026-09-28
- Forge tracks how AI agents discover and buy your x402 APIs — kleffew94 · 2026-09-28
- Open-source Aseprite MCP server lets agents animate directly on a live canvas — 72-frame rain scene stress test — Complex-Log-635 · 2026-09-28
- Cloudflare's Kitesurf Agentic Browser Adds WebMCP Support and Terminal Rendering — dinasaur_404 · 2026-09-28
- OpenAI President Brockman's prompting guide reworked for the agentic era — kimmonismus · 2026-09-28
- Open-source qiaomu-cut Skill turns one sentence into a reproducible AI video production pipeline — vista8 · 2026-09-28