Why a real person is currently pretending to be an AI

A San Francisco artist named Tucker Bryant recently launched a project called ChatTJB. The website looks and functions like a standard chatbot interface, encouraging people to ask questions or request images. But there is no software, no complex computer program, and no automated engine working behind the scenes. Instead, Bryant and a rotating group of people manually read each message and write or draw a human response. While it started as a joke, it quickly received over 30,000 queries, with people sharing sincere, personal stories once they realized a real person was on the other end of the conversation.

Most of the tools we use today, like ChatGPT or Claude, are powered by what is called a large language model—a complex software engine trained on massive amounts of internet text to predict the next likely word in a sentence. When you send it a request, the model isn't actually thinking; it is performing a high-speed statistical game to guess what a human might say next. Because these models are fast and available, they have become our default consultants. ChatTJB works by completely bypassing this mechanical process. Instead of a statistical calculation, the user gets a social interaction. The value shifts from getting a data-driven answer to the simple, human experience of being acknowledged.

This project is a pointed experiment in what Bryant calls cognitive surrender—the habit of handing our critical thinking over to software. When we default to AI for every small decision, we stop practicing the skill of making those choices ourselves. We start to defer to a machine even when we are perfectly capable of deciding what to wear or how to spend our free time. ChatTJB is a gentle, human-powered nudge to help us slow down. It asks us to consider why we feel the need to use a machine for a human problem, reminding us that sometimes, the best answer isn't an instant output, but a conversation with another person.