Meat Proxy

A meat proxy is a person who forwards text, code or other output from an AI system without reading, understanding or checking it, and so acts only as a relay between the machine and the recipient. The term is a pejorative that spread among software developers in August 2026. It targets the missing human judgment, not the use of AI as such.

Where the term comes from

An earlier form of the phrase appeared in a blog post titled “meat-based llm proxies”, published on March 31, 2026, on the site not-an-llm and attributed by Simon Willison to Frank Allenby. It described people who route every message they receive through a language model and pass the reply back, so that talking to them means talking to the model. The post also noted that the model tends to back its user’s existing position, which makes disagreeing with such a person futile.

The short form took hold through the developer Niklas Gruhn. In “Don’t be a meat proxy” (August 3, 2026) he complained about chatbot answers pasted verbatim into Slack threads, code reviews and group chats, and admitted that he had done the same. Such output, he wrote, is long, often plausible but wrong and dense with jargon, and the person asking could have queried the model directly. His advice: read, understand and validate the output, then reply in your own words. Simon Willison shared the post the same day, calling it an excellent new term, and later added a note on Allenby’s earlier use. Gruhn’s post drew more than 700 comments on Hacker News (Allwork.Space), and by late August the term had reached Kottke, Futurism and CNET.

How it is used

The term names a person; its counterpart workslop names the artifact, AI-generated work that looks finished but lacks substance. As the tech site OfficeChai put it, the cost of checking does not disappear but moves to the recipient, who often has less context than the sender. Gruhn’s sharpest example comes from code review: a developer pastes a ticket into a coding assistant, submits the result unread and feeds the reviewers’ comments back into the tool, so that the reviewers end up doing the work. Similar labels include “human ChatGPT wrapper”. The writer Cory Doctorow’s “reverse centaur”, a person serving as an appendage to a machine, describes a related relationship (Futurism).

Contested points

Hardly anyone defends unread relaying; the dispute is about where the line runs and who is to blame. Gruhn encourages people to prompt AI and objects only to forwarding without review. The engineer Vivek Haldar argued in September 2026 that knowledge work should go the way software engineering went, from uncritical “vibe coding” to disciplined use of agents in which the human still reads and checks the results. On X, the tech founder Nicholas Charriere warned of “a very fine line” between using AI to be more productive and being a meat proxy, as Futurism reported.

Some of the pressure comes from above. In the survey behind the workslop research, 41 percent of employees said leadership had encouraged AI use without explaining how or why (BetterUp), and the researchers trace low-quality AI output partly to such mandates (Harvard Business Review). The term meat proxy, by contrast, places the responsibility on the individual who presses send.

Why it matters now

The AIciety classification describes Level 7, delegated action, as AI carrying out tasks on people’s behalf. Delegation presupposes someone who sets the task and checks the result; a meat proxy reverses the roles and carries the machine’s output instead of directing it. If the recipient then pastes the message into their own assistant, the exchange comes close to Level 8, communication that no human writes or reads. The term thus questions what a human in the loop contributes if they only pass things on. In October 2026 the song Meat Proxy, made with Suno and Claude, set the complaint to nu-disco, and the video Escape Velocity nods to the term.

Related

Workslop · Sycophancy · Human in the Loop · Slopcore · Meat Proxy — when the human adds nothing (on the song) · Escape Velocity — the video that made Musk feel the AGI · The AIciety classification

Sources: not-an-llm: meat-based llm proxies (March 31, 2026) · Niklas Gruhn: Don’t be a meat proxy (August 3, 2026) · Simon Willison: Don’t be a meat proxy (August 3, 2026) · Kottke: Meat proxy (August 24, 2026) · SFisms: Meat proxy · OfficeChai: How “Meat Proxy” Is The Latest Hilarious AI-Related Pejorative (August 25, 2026) · Victor Tangermann: Your Coworkers Are Becoming “Meat Proxies” for AI (Futurism, August 26, 2026) · Allwork.Space: “Meat Proxy” Is The New Insult For Coworkers Who Let AI Do The Thinking (August 26, 2026) · Tyler Lacoma: Are You a ‘Meat Proxy’? (CNET via Yahoo Tech, August 27, 2026) · Vivek Haldar: How (and why) to be an effective meat proxy (September 27, 2026) · BetterUp: Workslop infographic · Niederhoffer et al.: AI-Generated “Workslop” Is Destroying Productivity (Harvard Business Review, September 22, 2025) · AIciety classification · Meat Proxy — when the human adds nothing · Escape Velocity — the video that made Musk feel the AGI

Leave a Reply

Your email address will not be published. Required fields are marked *