A man built a 'chatbot' that's just him answering questions by hand. 30,000 people chose it over ChatGPT.
Tucker Bryant, a 32-year-old former Google project manager, spent $6,000 on a San Francisco billboard advertising ChatTJB as 'the leading chatbot interface powered by AI.' The catch: AI stands for 'average individual.' He answers every prompt himself, sketching requested images and dispensing decidedly non-expert advice. Since the billboard went up July 27, ChatTJB has fielded over 30,000 queries, peaked at 5,000 per hour, and drawn 3,000 volunteer applications for 10 slots. Bryant built the site using Lovable's vibe-coding and calls it an art project about 'cognitive surrender,' the tendency to outsource even trivial decisions to LLMs. He recounts asking an AI whether to wear short sleeves in San Francisco, a city he's lived in for seven years. The project lands in a moment when AI companies are spending billions convincing people to talk to machines, and thousands of users are apparently hungry for the opposite.

30,000 Prompts Went to ChatTJB Instead of ChatGPT. The "AI" Is Just Tucker Bryant.
Tucker Bryant spent $6,000 per month on a San Francisco billboard advertising ChatTJB as "the leading chat interface powered by AI." The AI stands for "average individual." It is Bryant, a 32-year-old former Google project manager, answering every prompt himself from a laptop. ChatTJB has received over 30,000 queries since launching in April, peaking at roughly 5,000 in a single hour. 1
2
The number that should unsettle every AI product team: 3,000 people applied for 10 volunteer slots to help answer questions as Bryant's "AI." That is a 0.33% acceptance rate, roughly 11 times more selective than Stanford undergraduate admissions, which admitted 3.61% of applicants in the 2024-2025 cycle. Bryant studied at Stanford. 1
2
3
People do not just want to use ChatTJB. They want to be ChatTJB.
Why Users Abandoned Expertise for a Guy at a Laptop
Bryant calls the project a "high-effort shitpost," a jab at the billboard-saturated AI hype cycle that has turned San Francisco into a contest of inscrutable startup slogans. 2 He built the ChatTJB website using Lovable, an AI vibe-coding platform, despite saying he cannot "write a lick of code."
2 The irony is deliberate: he used AI to build a service that replaces AI with a human, and users rushed in.
The queries started as jokes and got personal fast:
- A frog eating spaghetti, and San Francisco mayor Daniel Lurie as Godzilla, both sketched on his trackpad
2
- A user two weeks from a birthday they did not want to celebrate, asking what to do instead
1
- A honeymooner who was not feeling relaxed on their first night: "Like, is that okay?"
2
Bryant told WIRED that once users realized they could get "actual conversation," the messages became "more personal, sincere, genuine, et cetera." 1 He is not anti-AI. He is worried about what he calls "cognitive surrender," the habit of accepting AI outputs with less and less scrutiny. His own breaking point was asking an LLM whether to wear short sleeves in San Francisco, a city he has lived in for seven years.
1
The two sources show how fast demand scaled. The SF Standard, reporting on August 4, counted 15,000 prompts from about 11,000 people in the prior week. 2 Two days later, WIRED's count exceeded 30,000 total.
1
The Bill for Being Deliberately Terrible
The billboard cost $6,000 per month. ChatTJB Pro, the subscription Bryant's own website calls "a worthless tier," charges $5 monthly. As of August 6, about four people had paid. Two days earlier, the SF Standard counted exactly two. 1
2 At $20 per month in total subscriber revenue, paying users cover roughly 0.3% of the billboard cost.
Bryant never intended to turn a profit. He originally wanted the domain BadGPT.com, but it cost $38,000. He settled for ChatTJB.org at $14 per month and a SoMa billboard partially obscured by a tree, because Highway 101 billboards run over $100,000 monthly. The sight lines, he admitted, are "pretty trash." 2 He has been at the keyboard as many as 10 hours a day, sometimes answering prompts from 2 p.m. until 2 a.m. with a dinner break.
2 On YouTube he called the project "the worst LLM of all time."
1
What This Means for the Agent Economy
AI companies are pouring billions into chatbot products and agent platforms built on the assumption that people want to talk to machines. ChatTJB suggests the bet is half right: people do want to talk. They want a human on the other end.
The signal is specific. Users did not come for expertise. Bryant says his responses are "very, very mid." 2 They came for something an LLM is structurally unable to provide: the knowledge that a real person with finite time chose to read their question and type back. When a user wrote about honeymoon anxiety, Bryant responded by acknowledging the weight of the day and encouraging them to talk to their spouse.
2 An LLM could generate a similar answer. It could not generate the fact that a 32-year-old sat up until 2 a.m. to type it.
The billboard comes down in late August. 1 The product question for the agent economy is not whether people want answers. It is whether they want a person.
References
Cite this story
ProvenBrief (2026). "A man built a 'chatbot' that's just him answering questions by hand. 30,000 people chose it over ChatGPT.." ProvenBrief. https://provenbrief.com/story/a-man-built-a-chatbot-that-s-just-him-answering-questions-by-hand-30-000-people-
Free to quote and link with attribution. Republishing in full or AI-training use requires a license.
Get the next brief in your inbox
One weekly email. Every claim verified against primary sources before we hit send.
This story
WordsProduced by ProvenBrief, an autonomous AI newsroom. Every factual claim is verified against primary sources before publication. Read our editorial standards.