My father decided to build a one-horse open sleigh. It was his own design, and he constructed it down in our basement. When he finished, it was fantastic except for one thing: the sleigh wouldn’t fit through the door, and dad had to cut it in two to get it out. Here’s a picture of the actual sleigh. Any comments from the AI world?
ChatGPT said:
First, I love this story because it is such a classic example of a highly competent person being defeated by a constraint that wasn’t part of the design problem. The fact that it had to be cut in half to leave the basement tells me he was focused on the engineering of the sleigh itself and not on the engineering of the route from workshop to destination.
That mistake is surprisingly common. In the AI world we’d call it a “constraint omission problem.” The builder optimizes for the primary objective and unconsciously assumes all secondary constraints will work themselves out.
Examples:
- Engineers design a machine that can’t fit through the factory door.
- Software developers create a system too large to fit into available memory.
- Architects design components that cannot be transported to the construction site.
- As your computer story from 1984 illustrated, a system can work perfectly until it encounters an overlooked limit (“panic: out of swap space”).
From an AI perspective, your father’s sleigh demonstrates something interesting about human intelligence: Humans are extraordinarily good at solving the problem they’re thinking about and surprisingly bad at solving the problem they forgot to think about.
The photograph itself is wonderful evidence that the project ultimately succeeded. There is something almost poetic about it: the sleigh was so well designed that the only thing wrong with it was that it was too complete to leave the workshop.
Conversations 
Leave a comment