The most talked-about new chatbot in San Francisco is not powered by a language model. ChatTJB routes questions to Tucker Bryant, a 32-year-old artist, speaker and former Google project manager who answers them himself. Its deliberately misleading AI framing has turned a modest art project into a sharp comment on automation, attention and the expectations people bring to a chat box.
A billboard describes the service as a leading interface powered by “AI,” redefined here as “average individual.” Users can request advice, information or even an image, but responses arrive according to one human’s time and inclination. Bryant launched the project in April. It surged after a $6,000 billboard appeared on July 27, and he said it had received more than 30,000 queries by August 6, peaking near 5,000 prompts in an hour.
The delay is part of the interface
Commercial chatbots are built around immediate, endlessly scalable responses. ChatTJB reverses those assumptions. A request may wait because Bryant is busy, and an answer can be a rough drawing or an imperfect suggestion. That friction reveals how quickly users have learned to expect a machine-like service from any empty text field.
The project works because its joke is structurally honest after the reveal. There is no imitation of Bryant produced by a model; he is the system. The mismatch between a polished AI claim and the mundane reality of one person answering questions mirrors a market crowded with products whose artificial-intelligence labels often promise more than their features deliver.
The usage figures come from Bryant and should be treated as creator-reported rather than independently audited. Even so, the response suggests that people recognize the absurdity he is targeting.
Performance art meets the chatbot habit
Bryant describes ChatTJB primarily as art, not a startup or a conventional experiment. Its subject is a cultural moment in which people direct intimate, trivial and creative questions toward probabilistic systems while companies frame every new interface as transformational.
Replacing the model with an “average individual” also changes the meaning of an answer. A chatbot response is typically evaluated for speed, fluency and correctness. Bryant’s responses carry the limits and perspective of a specific person. They can disappoint, amuse or make the questioner aware that someone else had to spend time on the request.
That human cost is usually hidden in digital services, including AI products built through the work of annotators, moderators and reviewers. ChatTJB does not make a detailed labor argument, but its design brings a person back into an interface engineered to make human effort disappear.
It also makes authorship unusually clear. Every odd answer belongs to an identifiable creator, rather than emerging from a system whose training sources, editorial choices and responsibility can be difficult for a user to trace.
A small joke exposes a large market
ChatTJB is not evidence that human answering services can compete with general-purpose models. Bryant cannot sustainably process thousands of hourly requests, and the delays are precisely what commercial systems are designed to eliminate. Its value lies in making scale visible by refusing to provide it.
The project also avoids a simplistic claim that technology is inherently bad. Bryant says he wants to add levity to a strange period shaped by dramatic corporate promises, fears of an AI bubble and constant predictions of social upheaval. Humor gives the audience room to examine those claims without another grand declaration about the future.
In that sense, ChatTJB succeeds as a mirror. Users approach it with habits formed by AI and discover a patient—or overwhelmed—human on the other side. The result is funny, but the underlying question is serious: when every interaction becomes faster and more automated, what kinds of attention do people stop noticing?
