A $6,000 San Francisco billboard advertises a chatbot that is one man typing

San Francisco’s newest AI product has a $6,000 a month billboard, thousands of users, and no AI. Every answer is typed by one man.

The bit, executed at billboard scale

The 14 by 48 foot board at Sixth and Folsom reads ChatTJB, the leading chat interface powered by AI. The AI stands for Average Individual. Behind it is Tucker Bryant, 32, a former Google employee turned poet and conceptual artist, who built the site over a weekend and answers the prompts himself.

The numbers arrived fast. In its first week ChatTJB took about 15,000 prompts from roughly 11,000 people, peaking at 5,000 prompts an hour over one weekend. Bryant’s own review of the output he produced: not high quality responses, very very mid, objectively mid.

The site carries a disclaimer that it is satire and a public art project.

What the joke is actually about

Bryant calls the project a high effort shitpost about cognitive surrender, the speed at which people hand over their thinking to anything that looks like a chat box.

The experiment worked because nobody checked. Fifteen thousand prompts went into a text field, and the interface was the entire proof. A billboard claiming AI, a familiar chat layout, and a blinking cursor were enough to establish the premise; the actual mechanism behind the box never came up.

That is a finding worth taking seriously beyond the joke. The trust people extend to AI products attaches to the interface pattern, not to the system behind it. If a chat box with a marketing budget can pass as AI while a poet types the answers, the same surface can just as easily front a much worse actor than a conceptual artist with a disclaimer.

The mid quality of the answers, by the creator’s own assessment, did not slow adoption either. Eleven thousand people kept typing. The bar an interface has to clear, it turns out, is lower than the industry’s benchmark charts suggest.