tl;dr: Relaying verbatim AI output in Slack, PRs, or chats wastes others' time—they can prompt the LLM themselves with better context, and AI text is verbose, jargon-heavy, and often subtly wrong. Instead, read and validate the output, then respond in your own words. Otherwise, in scenarios like code review, the reviewer ends up doing the actual implementation via their own LLM, with you serving only as a "meat proxy."