Developer Postmortem: Fully Autonomous AI Coding for Complex Projects Fails
MarcJSchmidt · x · 2026-07-25
A developer shared lessons learned from six months of using AI tools like Claude, GPT, and Kimi for fully autonomous coding, which ultimately failed. For non-standard, complex projects, letting AI take the wheel leads to a decayed codebase filled with inconsistent files and drifted assumptions.
He has since shifted to a human-in-command strategy: he now writes the core code and pseudo-code himself, using AI only to fill in the blanks. He also intentionally writes documentation in his own language—being purposefully vague or specific—to maintain 100% understanding of the project. He also noted the struggle of finding the perfect claude.md prompt to stop the AI from overstepping and wrecking the codebase.
More from coding & agent
- Codex tip: use Sol with Astra and Luna sub-agents to save usage — pvncher · 2026-09-11
- agents-best-practices: a provider-neutral Agent Skill for designing and auditing agentic harnesses — tom_doerr · 2026-09-11
- Cognition's SWE-2 uses a KKT duality argument in RL to shift the effort Pareto curve — YouJiacheng · 2026-09-11
- First-ever Three.js Conference lands in Paris, with a panel on AI-shortened design workflows — OdinLovis · 2026-09-11
- Data engineering, not agent frameworks, is the real bottleneck for enterprise AI agents — dhruv2038 · 2026-09-11
- RTK Terminal Compression Cuts Tokens but Leaves Your AI Coding Bill Unchanged — Bartaseth · 2026-09-11