Tucker Bryant, a 32-year-old former Google project manager, launched a website called ChatTJB in April that looks like a typical AI chat interface but is powered entirely by people. Users type a question, wait for a response, and receive an answer written by Bryant or one of more than 10,000 volunteers who have signed up to help.
The concept behind ChatTJB
The site mimics the stripped-down design popularised by tools such as ChatGPT: a simple text box, a prompt and a reply. The twist is that there is no large language model behind the scenes, the "AI" actually stands for "average individual". Bryant promoted the service with a billboard in San Francisco that proclaimed it a "leading chat interface powered by AI" with a tiny asterisk explaining the joke.
Since the billboard went up, more than 100,000 prompts have been submitted and the volunteer waiting list is growing by roughly 1,000 people each day. At its busiest, the service received over 5,000 questions an hour, a volume that forced Bryant to step back and invite others to take over the answering.
Why it matters
The project emerged from Bryant's worry that people increasingly outsource their thinking to artificial intelligence without questioning the answers they receive. He described a personal episode of asking an AI whether to wear short or long sleeves on a 65-degree day, a practice researchers call "cognitive surrender". Studies by Wharton professors Steven D. Shaw and Gideon Nave found that reliance on AI can improve accuracy when the system is correct, but can reduce it by 15 percentage points when the system errs.
"When people hear the term, they immediately recognise the phenomenon, both in others and often in themselves," said Shaw.
ChatTJB reintroduces friction into the interaction, users may wait hours for a reply, and the answers come from a human perspective rather than an algorithmic one. This scarcity can make the responses feel more meaningful and reminds users to scrutinise the information they receive, a quality that fast, scalable AI tools often lack.
What happens next
Although Bryant does not intend to commercialise the service, the unexpected popularity has attracted interest from potential collaborators. He has said he would consider partnering with an organisation that could turn the project into a sustainable venture.
For now, the bulk of the answering is being handled by volunteers, and the waiting list continues to expand. The experiment raises broader questions about how society balances the speed of AI with the need for human oversight, and whether similar human-in-the-loop models could be applied to other digital services.