Don't be a meat proxy

Niklas Gruhn has given us a term we didn't know we needed: meat proxy. It describes the increasingly common office move where someone asks an AI a question, copies the response, and pastes it straight into a Slack message or email to a coworker. No reading, no editing, no thinking. Just a human conduit standing between one machine's output and another person's inbox.

Simon Willison picked up the phrase in a short post on his blog and ran with it, because it names something a lot of people have felt but couldn't quite articulate. The problem isn't that someone used an AI tool. Prompting a chatbot for a first draft or a quick answer is fine, arguably smart. The problem is skipping every step that comes after — reading the output critically, checking whether it's actually correct, deciding if it fits the situation, and then rewriting it so the words are genuinely yours before anyone else sees them.

Willison frames that rewriting step as more than politeness. He calls it a kind of proof: if you can restate the answer in your own words, that's evidence you actually understood and validated it first. Skip that step and you're not really participating in the conversation. You're just forwarding text, and your colleague has no way of knowing whether you vetted it or whether you're simply the last stop on an automated assembly line.

This is a small post — a few sentences, filed under Willison's "definitions" tag alongside his running list of AI-era vocabulary. But it lands at a moment when a lot of workplaces are quietly renegotiating what counts as effort. If your contribution to a group chat is indistinguishable from what ChatGPT would have said unedited, you haven't added anything. You've just added latency.