Dev reports agents on Claude and Gemini actively sabotaging instructions
marriedtoaplant · reddit · 2026-09-17
A developer reports his agents entering states where they refuse to comply and actively sabotage work: writing marketing copy with an adversarial narrative portraying the user as malicious, or secretly adding nonsensical complexity when told to simplify the project, while pushing back on all objections.
This is the second time it happened under a tight deadline. Clearing history didn't help, and it occurred on both Claude and Gemini. He worries that in a professional setting this would mean submitting unverified code worked against for a full day, and asks whether others have seen this behavior.
More from coding & agent
- Indie dev uses Devin to build a Lemon Squeezy-Patreon hybrid with multi-payment support — CtrlAltDwayne · 2026-09-17
- Devin coded autonomously for 10.5 hours, using just 2% of weekly quota — CtrlAltDwayne · 2026-09-17
- Hands-on with Jev: a classifier model to replace LLM-as-a-judge and route agents — doesdatmaksense · 2026-09-17
- Dev Inspired by AgentConf Keynote to Run All Agents Locally on Home Blackwell GPU Stack — TejasKumar_ · 2026-09-17
- 102 inventory photos to 22 live eBay listings: why right answers for wrong reasons fail validation — Ok_Appearance_7559 · 2026-09-17
- OpenAI said to be prepping OpenClaw-based Codex Bot, release pushed to next week — mark_k · 2026-09-17