AI Coding Isn't Fully Autonomous: Complex Projects Require Rigorous Human Review
georgemillo · x · 2026-08-13
Developer George Millo shared his engineering practices regarding the reliability of AI coding. While he often builds with AI without closely inspecting the code, he insists on rigorous human review for huge, complex, and mission-critical applications at his day job.
Key Points
- Fault Tolerance Dictates Workflow: A bug in core systems could cost millions. Even though AI generates most of the code and assists in reviewing it, human eyes are still required on almost all PRs.
- Understanding is Key to Collaboration: Reviewing isn't just about catching bugs; it's about understanding the app. Developers must read the code to evaluate the AI's plan and steer it effectively.
- Models Have Limits: Current models still miss things, make mistakes, and suggest flawed plans on large codebases. Those who believe AI can solve any coding problem haven't used it on a sufficiently large project.
More from coding & agent
- Reverse Engineering Context Compaction in Open-Source Codex CLI — Kangwook_Lee · 2026-08-13
- Omnigent v0.9.0 Released: Introduces Smart Routing and Refreshed UI — matei_zaharia · 2026-08-13
- Fully AI-Coded: Claude Builds Git Hosting Service on Cloudflare — samgoodwin89 · 2026-08-13
- Vibe Coding Backfires: Devs Reintroduce Old Bugs When AI-Building Frameworks — xeophon · 2026-08-13
- Pure Prompt to 3D: Claude Builds a Walkable City Street in Browser — prasenx · 2026-08-13
- Senior Engineer's 6-Step Claude Code Workflow: From Idea to Production — MaryamMiradi · 2026-08-13