Human-powered AI chatbot exposes our need to be heard
The billboard went up in San Francisco in late July, and the first thing you notice is the confidence. A man’s face, a simple promise: “the leading chat interface powered by AI.” The joke only lands if you read the fine print. Here, AI stands for “average individual.” Tucker Bryant, a 32-year-old former Google project manager, is not running a language model. He is sitting at his desk, reading your prompts, and typing back his own thoughts. He is the product. He is also the entire infrastructure.
The strange part is not that he built it. The strange part is that it works. People are not asking for jokes or memes, at least not at first. They ask about their birthdays, about dinner, about the ingredients in their fridge. They ask for advice on becoming a morning person. They ask a stranger to draw a frog eating spaghetti. The requests arrive at a rate of 5,000 per hour at peak, and Bryant answers them one by one. He is not fast. He is not an expert. He is just there, and that turns out to be enough.
The project began in April as a quiet experiment. It took off on July 27, when Bryant spent $6,000 on billboard space. By August 6, he had received over 30,000 queries, a figure he shared in an interview with. The San Francisco Standard. The math is not the point. The point is what the queries reveal about a moment in which we have outsourced not just our tasks but our attention. We have built machines that answer everything, and we have forgotten that the answer was never the valuable part. The valuable part was the person on the other end, the one who had to think.
The Quiet Surrender of Judgment
Bryant calls it “cognitive surrender,” and he uses the term carefully. He is not anti-AI. He built his website with an AI coding tool and says so openly. But he has noticed something in his own behavior that unsettles him. He lives in San Francisco, has for seven years. One day, he asked a language model whether he should wear short or long sleeves on a 65-degree day, a detail he recounted in a profile published by. The San Francisco Standard He knew the answer. He has been dressing himself for decades. And still, he deferred.
That moment is the hinge. It is not that the machine gave bad advice. It is that the question should never have been asked. The act of asking transferred a trivial decision to a system that has no stake in the outcome. The machine does not feel cold or warm. It does not know what it is like to walk down Market Street in a light jacket. It only knows patterns, and patterns are not experience. Bryant recognized the absurdity, and that recognition became the seed of ChatTJB.
The deeper issue is not the weather. It is the slow erosion of the instinct to decide. We are training ourselves to consult before we consider, to check before we choose. Each query is a small surrender, a repetition of the habit that makes the next surrender easier. The machine does not need to be right. It needs to be available. Availability is its own kind of authority, and we are increasingly unable to distinguish between the two.
Bryant’s project inverts this dynamic. When you ask ChatTJB a question, you are not getting a database. You are getting a person who has to think, who might be distracted, who might not know the answer. That uncertainty is the point. It forces you to re-engage with the question yourself, to hold it in your mind while you wait for a reply that may never come. The delay is not a bug. It is the cure.
The Human Role in the Loop
The professional world has spent the last two years asking what AI will replace. The answer, so far, is not entire jobs but entire judgments. The radiologist who reads scans, the lawyer who reviews contracts, the editor who catches errors: each of these roles contains a core of discernment that cannot be automated, yet each is being reshaped by tools that promise to handle the routine parts. The routine parts are where the skill was built. Remove them, and the discernment has nothing to practice on.
Bryant’s volunteers understand this instinctively. He recruited ten helpers from over 3,000 applicants, according to his own account in the same interview. The selection process was not about credentials. It was about willingness to sit with the discomfort of being asked. The volunteers answer prompts the same way Bryant does: honestly, imperfectly, with their own biases and blind spots. They are not pretending to be machines. They are offering the one thing machines cannot simulate: the experience of being a person with a body, a history, and a point of view.
The irony is that this human labor is now framed as a luxury. ChatTJB Pro costs $5 a month, and Bryant is explicit that it is worthless. Users get nothing but the knowledge that they are supporting art. Four people have paid so far. The billboard cost $6,000. The math does not work, and that is precisely the point. We have reached a moment where human attention is so scarce that it must be rationed, and yet we give it away freely to systems that do not need it.
The historical parallel is the telephone operator. For decades, human operators connected every call, and their work required memory, patience, and judgment. When automation arrived, the operators were not replaced all at once. They were gradually shifted to exception handling, to the calls the machines could not route. Then the machines got better, and the operators disappeared. The judgment they had exercised became irrelevant because the system no longer required it. The same process is now unfolding across every field that involves answering questions.
The Value of Being Wrong
Bryant is quick to admit his limitations. He is not a good artist, and he says so. His advice is no better than anyone else’s, and he knows it. On YouTube, he described ChatTJB as “the worst LLM of all time,” a phrase he used in a video posted to his channel This is not false modesty. It is a deliberate design choice. The value of the project is not in the quality of the answers but in the fact that they come from a person who can be wrong.
The wrongness matters because it creates friction. When a machine answers, the response is smooth, confident, and often wrong in ways that are hard to detect. When a person answers, the response carries the texture of doubt. You can sense when someone is guessing, when they are tired, when they are trying to be kind. That texture is information, and it shapes how you receive the answer. You do not take a human’s guess as gospel. You weigh it, you question it, you push back. That process is the thinking we are losing.
People have responded to ChatTJB with unexpected sincerity. They share personal details, ask for advice on real problems, treat the project as something more than a joke. Bryant finds this “very tender to see.” The tenderness is not about him. It is about the relief of talking to a person who has no agenda, no data to collect, no engagement metrics to optimize. The machine is always optimizing. The human is just there.
The project has a shelf life. The billboard comes down in late August. The volunteers may burn out. The novelty will fade. But the question Bryant raises will not fade, because it is not about the technology. It is about what we are becoming when we stop trusting ourselves. Every time we ask a machine a question we could answer, we are practicing a kind of forgetting. We are rehearsing the moment when we no longer need to know.
The consequence is already visible. We are raising a generation that has never known life without instant answers, that has never experienced the productive discomfort of not knowing. That discomfort is where curiosity lives. It is where questions get sharpened, where hypotheses get tested, where understanding gets built. Remove it, and you do not get efficiency. You get a world where everyone asks and no one thinks, where the answers are always available and always shallow.
Bryant’s project is a mirror, not a solution. It shows us what we have outsourced and what we have lost in the transaction. The machine does not make us superfluous. We make ourselves superfluous, one query at a time. The billboard will come down, but the question remains: when the machine answers for us, who is left to ask why we asked in the first place?
Sources
1. Google
2. YouTube