Robert C. Martin says he trusts coding agents only after tests and strict constraints
bennash · x · 2026-07-24
Robert C. Martin says he no longer reads code written by his agents. Instead, he surrounds them with hard constraints: unit tests, Gherkin tests, QA procedures, quality metrics, mutation testing, and coverage checks.
His argument is that AI coding productivity comes from verification, not blind trust. The post frames this as the new way to work with agents: curate and test aggressively, then rely on the process rather than manually reviewing every line.
Related event: Uncle Bob's Philosophy: Constraining AI Coding Agents with Tests(4 posts)→
More from coding & agent
- Grok 4.5 and Unity CLI can now revamp a 2D game without touching the editor — chongdashu · 2026-07-24
- MCP vs. A2A comparison says both protocols have a place, but neither covers everything — rseroter · 2026-07-24
- A follow-up reply adds links for the same @agnt_annie benchmark results — NathanWilbanks_ · 2026-07-24
- An AI agent posts strong benchmark results on graffiti, Keller maps and Erdős 835 — NathanWilbanks_ · 2026-07-24
- Agent Substrate targets sandboxed agents with Kubernetes-style orchestration — bibryam · 2026-07-24
- Truss: new single-user local AI agent harness with 3-tier security and MCP support — molbal · 2026-07-24