The Human-in-the-Loop Paradox: Why the 'Worst LLM' is a Wake-Up Call

In the current gold rush of generative AI, the industry is obsessed with scaling parameters and optimizing tokens. We are told that the path to efficiency is more compute, more data, and less human friction. But as a battle-scarred observer of tech cycles, I find the most telling commentary on this era isn't coming from a white paper or a keynote—it's coming from a $6,000 billboard in San Francisco.

As Wired first reported, Tucker Bryant, a 32-year-old artist and former Google project manager, launched a project called ChatTJB. On the surface, it looks like any other chat interface. In reality, the 'AI' powering it stands for 'average individual.' There is no model. There is no training set. There is just Bryant, sitting in a room, manually answering prompts and scribbling drawings of skunks in sunglasses or San Francisco mayor David Lurie as Godzilla.

On the surface, this is a joke. Bryant himself has described it on YouTube as 'the worst LLM of all time.' But if you look closer, it's a devastating critique of our current relationship with technology.

Bryant tells Wired he is concerned about 'cognitive surrender'—the tendency to accept AI outputs with a dwindling amount of critical thinking. He admits to this himself, recalling a moment where he deferred to an LLM to decide if he should wear long or short sleeves on a 65-degree day in a city where he has lived for seven years. It is a symptom of a broader malaise: we are outsourcing the basic functions of being human to machines because it feels 'easier.'

Yet, the irony is that when Bryant offered a human alternative, people flocked to it. Wired reports that ChatTJB received over 30,000 queries as of August 6, with peaks of 5,000 prompts per hour. People aren't coming for 'expert advice'—Bryant is quick to note he isn't a great artist and his advice is no better than anyone else's. They are coming for the 'tender' experience of bouncing a question off another person. They are asking for dinner suggestions, help with fridge ingredients, or advice on how to spend a birthday without throwing a party.

This is the ultimate irony of the AI boom. We are spending billions to simulate human-like interaction, only to find that the most efficient 'LLM' on the market is simply a human being who cares enough to answer. The fact that thousands of people are writing in—and that over 3,000 volunteers applied to help Bryant manage the load—suggests a deep, unmet hunger for genuine connection that no amount of GPU clusters can satisfy.

Bryant isn't anti-AI; he even used a tool called Lovable to 'vibe-code' the website. But ChatTJB serves as a necessary speed bump in a world terrified of an AI bubble and exhausted by companies claiming to sell world-changing tech.

While four people have paid a $5 monthly fee for a 'worthless' Pro tier, the real value isn't financial. It's the realization that while an LLM can tell you how to fry an egg or dress for the weather, it cannot provide the sincere, personal, and genuine interaction that makes the 'average individual' irreplaceable. If we continue our cognitive surrender, we aren't just losing our critical thinking—we're losing the very thing that makes the conversation worth having.