As more debug notes, docs and code come from machines, engineer Tom Bedor has named the friction: when is it acceptable to forward AI output for another human to read? His answer, in an 11 June post, is a rule. If you are requesting human attention, demonstrate human effort.
The sharp moment is a small story. A teammate ran Bedor's design through an AI, then sent him the critique with the note, "I didn't read this, so it might not be entirely accurate." His reaction: if reading it wasn't worth your time, why is it worth mine? That asymmetry is the whole argument. If I can have a robot say something, so can you, so unlabelled AI text quietly shifts the cost of verification onto the reader.
His practice is mundane and probably right: label AI content, add your own commentary, and review your own AI-generated code before asking a colleague to. Attention was scarce before AI, and treating it as free is how teams drown in plausible text.