WRONG, NOT BROKEN

Summary from the TLDR DevOps newsletter

Traditional observability can show that an AI agent is fast, available, and error-free while missing whether its output is actually correct.