For this project, I wanted to see what would happen if three different AI models received the exact same creative prompt.

The assignment was to invent an exhibit for a fictional Museum of Completely Ordinary Objects, where mundane household objects are treated with the seriousness normally reserved for priceless historical artifacts.

I used the same prompt with ChatGPT, Claude, and Copilot, including instructions for each model to brainstorm multiple objects, rank its ideas, create the exhibit, critique its own work, and revise the weakest section before presenting the final result.

The results were surprisingly different.

Copilot chose a plastic bread clip and turned it into “The Polyethylene Sentinel: A Study in Domestic Continuity and Quiet Restraint.”

Claude chose the spare button sewn inside a shirt. Its exhibit treated the button as a tiny piece of “manufacturer-issued optimism,” supplied in anticipation of a failure that may never happen.

ChatGPT chose a folded paper towel wedged beneath an uneven table leg and treated it as evidence of the point when a temporary household workaround quietly becomes permanent infrastructure.

What interested me most wasn't which model wrote the funniest exhibit. It was how differently they interpreted the same instructions.

One focused primarily on the object itself. Another found an overlooked contradiction built into an everyday object. Another focused on the human behavior surrounding an ordinary object.

My favorite was Claude's Reserve Button, because the humor didn't have to be added to the object. Looking at something completely familiar with unnecessary seriousness revealed that the object was already a little strange.

The experiment changed how I think about prompting. A detailed prompt can establish the rules, structure, and creative boundaries, but it doesn't determine the idea the model will find inside those boundaries.

Same prompt.

Three models.

Three very different ways of noticing the ordinary.