When Your Agent Can't Find the Code It Wrote Two Weeks Ago: 3 Months of Agent-Driven Development
Sweaty-Landscape-561 · reddit · 2026-09-06
A developer recounts building a booking product almost entirely with AI agents since May with a team of three:
- Month one was unreal — whole features in an afternoon. Month two was fine. Month three fell apart.
- The turning point: asked to change cancellation refund logic, the agent spent 20 minutes searching for code it had itself written two weeks earlier — and found three different versions in three folders, picking the wrong one. Nobody on the team knew which version production used, because no human had read them.
- The fix: they bolted CodeRabbit onto the repo; the first review pass returned 200+ comments. Mostly nits, but the number was the honest cost of three months of unread code.
- Takeaway: they're slower now than in month one, and the author believes that's the correct speed — month-one speed was borrowed.
A vivid case study of codebase rot under fully agent-driven development.
More from coding & agent
- Dev demos AI models handling 3D objects and animations surprisingly well — chongdashu · 2026-10-23
- Jason Kneen burns 104.56B tokens and $24.7K in 304 days of AI coding on Codex and Claude Code — jasonkneen · 2026-10-23
- Typosquatted Blender MCP Repo Hit 958 Stars in Days and Silently Ran Malware — Promptmethus · 2026-10-23
- Benchmarking 56 fresh dependency bugs: MCP routing lifts Claude Code accuracy 68% to 91% — Available_Side_7134 · 2026-10-23
- Swapping only the eval harness moves results 30%: designing agent evals like a scientist — AI Engineer · 2026-10-23
- Med student builds a whole-library RAG with Cognee; single-book ingestion still freezes his machine — shrike293103 · 2026-10-23