Veteran dev: agentic coding loops 8 times, burns 100k tokens, and wrecks mature codebases
cgouguen · reddit · 2026-09-10
A developer with 30 years of experience argues agentic coding workflows fail on mature codebases: agents loop 8+ times, burn 100k+ tokens talking to themselves, hallucinate dependencies, and produce spaghetti that passes basic tests but violates project architecture. His alternative: as the codebase owner, hand-pick the few relevant files and send them to a chatbot or aider — files that are hard for AI to locate but easy for you. He concedes agents shine on greenfield projects (built a Rust project in minutes) but finds them expensive and unmaintainable at scale.
More from coding & agent
- Claude Code + Monid MCP builds a $12K/year B2B prospecting engine on live data — aryanXmahajan · 2026-10-26
- Karpathy shares CLAUDE.md rulebook to stop AI coding agents from piling up tech debt — mdancho84 · 2026-10-26
- Rethinking dependencies in the agent era: drop libs that save 20 lines of code — gethackteam · 2026-10-26
- RustGIS hits 2,000 GitHub stars in two weeks with agent-first architecture — giswqs · 2026-10-26
- Vercel Ships 12 New Products at Its 2026 Conference as It Races Rivals for the Agent Entry Point — FinanceYF5 · 2026-10-26
- Build an AI agent from scratch in under 50 lines of pure Python: ReAct loop tutorial — adnan_hashmi · 2026-10-26