Tucker Bryant Turns ChatTJB Into Human-Powered Chatbot

Bryant launched the website in April, according to the San Francisco Chronicle. He told the Chronicle that the billboard was designed to look serious while giving viewers "a little wink that it's actually not serious at all." The billboard was scheduled to remain up for one month, SFGATE reported.

The service's language extends the joke. USA Today reported that Bryant has called it "artisanal intelligence, handcrafted by a single human being" and described it as a "single-operator large language experience," rather than an LLM. In a video cited by USA Today, Bryant summarized the premise as: "Picture this: ChatGPT but, on the other end, it's just me giving bad advice."

Bryant told SFGATE that he wanted the interface to be "really mid, really average" so that people might take responses from AI products less seriously. He told USA Today that ChatTJB's replies would be slow, potentially incorrect, and cryptic. Those properties are the inverse of the instant-response expectation established by mainstream conversational systems.

Public reporting frames ChatTJB as an art project about reliance on chatbot outputs rather than a competitive AI product. Bryant told USA Today that the project addresses "cognitive surrender," a term the outlet linked to University of Pennsylvania Wharton research on users trusting chatbot advice. Futurism reported that Bryant encountered the term through his partner and connected it to his own experiences over-trusting LLM outputs.

Bryant told Futurism that the project was not intended as anti-AI. Instead, he described the work as an attempt to replace the polished, futuristic experience of token-streaming chat interfaces with "a really average or bad experience." The San Francisco Chronicle similarly reported that he used AI to create the website itself, an irony Bryant acknowledged in discussing the project.

The experiment offers a compact illustration of how interface cues can affect perceived authority independently of model capability. A familiar dark chat window, apparent response generation, and AI-oriented marketing can establish expectations before a user evaluates factual quality. Research and product teams working on evaluation, provenance, or calibrated trust commonly examine this gap between presentation and reliability, especially where users seek advice rather than simply generate text.

Its relevance is cultural and interactional: the project uses a one-person workflow to make the labor, latency, and fallibility normally abstracted by AI interfaces visible to users.