Agent ran 357 mostly duplicated test fixtures — why SE judgment matters more in the agent era
julianharris · x · 2026-11-03
julianharris shares a dialogue with his coding agent: tests took an hour because of "357 test sets from production data." Pressed on whether that was an efficient equivalence set, the agent admitted the run set was mostly duplicated fixtures (27 of 39 unfinished folders were logless git copies, classes repeated 12x, etc.). His takeaway: without a software engineering background, the chaos unfolding under agent-driven development is scary — and the real skill is knowing what question to ask, one we need more than ever.
More from coding & agent
- Celesto CLI Lets AI Agents Inspect CI Jobs and Logs — aniketmaurya · 2026-11-04
- Vercel Launches Security Review Beta: Auto-Scans Vulnerabilities on Every Push, One-Click Fixes — cramforce · 2026-11-04
- Codex security plugin scans GitHub repos with OpenAI Daybreak Blue, fewer cyber guardrails — drdanielbender · 2026-11-04
- A wishlist for the ideal personal AI assistant: parallel agents, scheduling, autonomy boundaries — koltregaskes · 2026-11-04
- AI/ML API ships an MCP that auto-runs Google Ads campaigns for ~$0.61 per run — heyshrutimishra · 2026-11-04
- Turning Suno tracks into lyric videos with Claude Code supervising Pokee Isaac — Kyrannio · 2026-11-04