Ex-Googler's ChatTJB billboard hides a one-man chatbot gag

A billboard in San Francisco is advertising what it calls the leading chatbot interface powered by AI, and the twist, reported by Futurism, is that the AI in the small print stands for average individual. The service is called ChatTJB, and the average individual doing the work is Tucker Bryant, an artist and former Google employee who reads each message and writes back by hand.

Bryant calls it a single-operator large language experience, and the website describes itself as artisanal intelligence, handcrafted by a single human being. Image requests come back as hand-drawn pictures. A disclaimer on the site notes the project is a work of satire and a public art project, and users can pay for a Pro subscription that, per Bryant, helps him pay off the billboard.

The piece worth sitting with is the argument underneath the gag. Bryant told the outlet that we're now so used to confident, fast, and smart-sounding answers from AI that we've largely stopped questioning them, and that some of us are so far gone that the second we drop into that soothing dark interface, see the little thinking animation, and the text coming up letter by letter, critical thinking skills go out the window. ChatTJB is engineered as the opposite, he warns his answers will be slow, probably not correct, and cryptic, and promises users a bad experience on purpose. The friction is the point.

The honest caveat is that this is one person's art project, not a study. Futurism does not put a number on how many people have messaged ChatTJB, how much the Pro tier is bringing in, or how the billboard was funded, so take the scale as anecdotal. What the reporting does not give you is any sense of whether the people typing into ChatTJB know they are chatting with a human or genuinely mistake Bryant for a model.

The forward-looking thing, if you build assistants for a living, is the design lesson hiding inside the joke. If a deliberately slow, deliberately cryptic responder can pull enough attention to earn a San Francisco billboard, the interesting question is which kinds of intentional friction, uncertainty markers, deliberate pauses, visible uncertainty, would push users of real models back into thinking for themselves.