THE READING ROOM
THE READING ROOM
An agent can explain why it wants to perform an action without establishing that it may perform it. It can produce a well-formed tool call without establishing that the arguments are safe. It can report success without proving that the external effect occurred exactly once.
An expert blog asks readers to trust more than its sentences. It asks them to trust the author's selection of evidence, interpretation of that evidence and willingness to correct a mistake.
A retrieval-augmented answer can look excellent in a demonstration and fail for a reader who asks the same question differently. It can cite a real document while relying on the wrong version.
An AI interface can be technically responsive and still be dishonest. It can animate a progress bar while nothing measurable is progressing. It can display “cancelled” while a server continues working. It can replace a current answer with a late result from a previous request.
A migration can compile successfully, pass a demonstration and still change the behavior that made an agent useful. The new implementation may choose different tools, forget state at a different boundary or retry an operation that the old implementation treated as terminal.
Cost per successful task can move differently from token price when retrieval, retries and failed answers are included.
An agent can select a function, assemble arguments and explain why it wants to call it. None of those abilities grants permission to perform the operation.
A fluent answer can conceal a broken search system. It can also make a good search system look worse than it is. If we evaluate retrieval-augmented generation only by reading the final response, we lose the ability to distinguish those failures.
