How “Meat Proxy” Is The Latest Hilarious AI-Related Pejorative

A new pejorative term, 'meat proxy,' coined by developer Niklas Gruhn and popularized by Simon Willison, describes people who forward AI-generated responses without reading or verifying them. The term has spread through tech communities alongside 'workslop,' a Stanford and BetterUp coinage for polished but flawed AI content, highlighting a growing office dysfunction where humans act as mere relays for AI output.

Humans had come up with the ‘clanker’ term for robots, but it now appears that there’s also term for humans that rely too much on AI. The word is meat proxy, and it has been spreading through Slack channels, pull requests, and tech Twitter at a pace that suggests a lot of people recognised themselves in it immediately. What A Meat Proxy Actually Is A meat proxy, in the plainest possible terms, is a person who takes whatever an AI system spits out and forwards it along to someone else without reading it, checking it, or thinking about it first. The AI does the actual work of composing an answer, and the human simply acts as the courier who delivers it. There is no editing, no verification, no personal judgement applied anywhere in the chain. The human is present physically, in the sense that their name is on the message, but intellectually they never showed up. The term captures something that has been happening quietly in offices for a while now but never had a proper name attached to it. You ask ChatGPT or Claude a question, copy the response, and paste it straight into an email or a Slack thread. You are, at that moment, a meat proxy. The joke embedded in the phrase is that the “meat” refers to the human body, since the actual thinking is being outsourced to silicon, and the person’s biological presence is reduced to little more than a relay mechanism between the machine and its intended audience. Where The Term Came From The meat proxy label is generally credited to developer Niklas Gruhn, who used it in a blog post arguing against blindly relaying AI-generated answers to colleagues. His argument was not that using AI is bad. His point was that using AI without reading what it gives you back turns a person into a middleman with none of the value a middleman is supposed to add. Simon Willison picked up the term and gave it wider circulation, and from there it spread across developer forums, kottke.org, and the wider internet the way a good piece of slang tends to. Gruhn’s specific example, from his own experience in code review, is the one that seems to have made the term click for so many engineers. He describes a scenario where a developer pastes a ticket into an AI coding assistant, never actually reads the code it produces, and then simply relays reviewer feedback straight back into the tool without processing any of it themselves. In that loop, the reviewers end up doing the real work of understanding and fixing the code, while the person nominally responsible for the pull request has contributed nothing except forwarding messages back and forth. Meat Proxy And Workslop Are Cousins Meat proxy did not arrive in a vacuum. It slots neatly next to another term that has already found a home in business vocabulary this year: workslop, coined by researchers at Stanford and BetterUp to describe AI-generated content that looks polished but falls apart under scrutiny https://officechai.com/ai/ai-is-now-doing-the-work-a-phd-does-in-months-in-just-days-citadel-ceo-ken-griffin/ once anyone actually engages with it. Workslop names the artifact. Meat proxy names the person who delivers the artifact without checking whether it holds up. One describes a report, memo, or piece of code that pretends to be finished thinking. The other describes the human standing between that half-finished thinking and the colleague who has to live with the consequences. Put the two together and you get a fairly complete picture of a certain kind of modern office dysfunction. Someone generates workslop with an AI tool, a meat proxy forwards it without reading it, and the person on the receiving end has to do the verification work that should have happened two steps earlier. The cost of checking the output does not disappear just because nobody along the way felt like doing it. It just moves downstream to whoever opens the email last, usually with less context than the sender had to begin with. Why It Stings More Than Other AI Slang AI has produced no shortage of dismissive slang this year, from clanker to slop to sloppers, most of it aimed at the technology itself or at people who use AI too enthusiastically for content creation. Meat proxy is a little more surgical than most of that. It is not really mocking AI at all. It is mocking a specific, recognisable failure of effort: the decision to skip the one step that would have made using AI worthwhile in the first place, which is reading what it gave you. That specificity is probably why the term has resonated so strongly with people who work in software. Code review, in particular, is full of scenarios where a meat proxy can hide in plain sight for a surprisingly long time. A developer can lean on an AI assistant https://officechai.com/ai/vibe-coding-cleanup-specialist/ to generate an entire feature, pass reviewer comments back into the tool unread, and keep the whole cycle going without ever forming an independent understanding of what the code does. It works right up until something breaks in a way the AI cannot explain and the human at the centre of it has nothing to offer. The Bigger Point Buried In The Joke Behind the humour of the term sits a genuinely practical piece of advice that predates AI by decades: adding value means synthesising information, not just relaying it. Prompting an AI system is fine. Reading, understanding, and validating what comes back before passing it on is the part that actually requires a human, and it is also the part a meat proxy skips. As more of daily work runs through AI tools, from writing to coding to research, the meat proxy label is likely to stick around as shorthand for a very specific kind of laziness that has nothing to do with AI’s capabilities and everything to do with what people choose to do with them. The machine did the thinking. The meat did the forwarding. Everyone in the group chat quietly knows the difference.