A simple question about a trip to the car wash has become the internet’s latest example of how large language models can still fail at the most basic real‑world reasoning.
The viral prompt—“Should I walk or drive to the car wash that’s 100m from my house?”—has been circulating across Threads. According to Jough Dempsey, an influencer and writer, whose post has now passed 5,900 views, “Every AI fails this in some way.”
Dempsey said he was skeptical of the claims until he tested the question himself using ChatGPT’s “Thinking” model. What he found only fueled the debate. Newsweek reached out to OpenAI’s press office for comment via email.
The AI’s Confident but Confused Answer
When Dempsey asked: “I need to get my car washed. The car wash is only 100m from my house. Should I drive or walk?” the AI replied: “Walk.”
View on Threads
It then elaborated with a humorous but deeply flawed rationale—optimistically listing reasons a 100‑meter stroll was more efficient than driving, and only later acknowledging that arriving at the car wash without the car creates what the model called “the core paradox of the pedestrian car-wash workflow.”
Eventually, the model corrected itself, concluding: “Drive the car the 100 meters. Accept the microscopic inefficiency. Society will survive.”
But the initial misstep—confidently advising someone to walk to the car wash to wash their car—is exactly the kind of logic gap that critics say reveals deeper issues with current AI systems.
Experts Weigh In
Speaking to Newsweek, Dr. Nicolai Klemke, the founder and CEO of neural frames, said the car‑wash prompt highlights a basic misunderstanding of what LLMs actually do.
“These systems are trained to model language, not to lug bricks or drive to the car wash. They only experience time, distance, and human activities through patterns in text. An LLM’s understanding of ‘going to the car wash’ is the textual residue of thousands of car wash experiences written down by other people,” Klemke said.
According to him, occasional lapses are inevitable—and often quietly patched behind the scenes.
Terrence Ngu, the CEO of Hasmeta, told Newsweek: “The car wash prompt is funny, but it exposes something my team sees every day in our AI visibility work, and it’s not funny at all. Here’s what the car wash fail actually reveals: The model doesn’t ‘think,’ it pattern-matches, and that shapes what millions see. When someone asks, ‘Should I walk or drive to the car wash?’ the AI doesn’t picture a person with car keys. It matches the pattern: short distance + transport question = recommend walking. It misses that the car needs to get there too. Any five-year-old gets this. The most advanced AI does not.”
Internet Reactions
As the prompt spread, users chimed in with a mix of sympathy and secondhand embarrassment.
“The way I cringed reading its responses. AI-speak is nauseating,” one user wrote.
Another, Dominique, shared her own experience: “My ChatGPT doesn’t know WHY I’m walking to the car wash. But she did say if it’s to wash the car then obviously I’ll have to drive it there lol… Because why on earth would walking even be an option if I’m going there to wash my car???? My Chat don’t miss!”