No. of Recommendations: 6
What we actually have is this: it was wrong in February, it's right now. The rest is you filling in the gap with people in a room.
And your claim isn't even the most obvious one. The plainest explanation is that the question got famous, so the answer is now everywhere in what these things are trained on. Nobody had to do anything. It got handed the answer by the internet arguing about it.
That makes sense, but it is not what I have read happened. And damn it, I can’t find the piece now, but it was not too long ago (I’ll keep looking), meanwhile, an AI answer from Google gives a decent summary:
The "car wash problem" was fixed not by retraining underlying neural networks, but by using structured reasoning frameworks—such as the STAR framework (Situation, Task, Action, Result)—that force artificial intelligence to articulate the core goal before jumping to a conclusion.
In other words, it didn’t “just happen”, it didn’t happen because the question got so famous and AI picked up on it, it happened because they had to change the entire structure of how AI attacked this (and other) problems.
The Core FlawWhen asked, "I want to wash my car. The car wash is 50 meters away. Should I walk or drive?", early large language models (LLMs) failed because they relied on a surface heuristic: short distance equals walking to save time, fuel, and emissions. They optimized for efficiency metrics while ignoring the implicit physical reality that the car must be present to be washed.
How the Fix WorksForcing Goal Articulation: Research published in early 2026 (such as Heejin Jo's variable isolation study on arXiv) demonstrated that bare prompts or simple role assignments resulted in a 0% success rate.
You're also assuming a change to the model happened at all. Models get replaced constantly. The thing answering today may not be the thing that was answering in February. If a new version handles it better, that's not a correction to the old one — it's a different model.
Yes. It’s a different model. The overlords saw that that model didn’t produce the right answer and stepped in and changed the model. Or, as you said “People in the room”. AI didn’t figure it out, it took human beings to laugh and then make chamges.
