A machine drew the map. We took the scenic route.

For an imaginary weekend project, a designer asked an AI to invent a city. It produced efficient streets, tidy parks, and a suspicious lack of places to get lost.

So she added an alley that went nowhere, a bookshop under a bridge, and a square with terrible acoustics. The result felt more alive. A useful first draft can give us something to disagree with, and that disagreement is often where the interesting work starts.

The problem with a perfect plan

The designer, whom we'll call Mara, had given the system an unusually tidy brief. Every home should be close to a park. Every shop should be easy to reach. Every road should have a clear purpose. The resulting map satisfied the instructions, but walking through its imaginary streets felt like browsing a well-organized warehouse.

Adding a reason to wander

Mara began introducing details with no obvious efficiency score. A courtyard caught the evening light. Two paths met beside a public piano. The bookshop's front door faced the river instead of the busiest road. These choices made the map harder to summarize, but easier to imagine living in. They gave its residents small reasons to take an unplanned turn.

A conversation with the draft

She then asked the AI to reconsider practical problems created by her changes. Could someone using a wheelchair reach the courtyard? Would deliveries block the river path? The model helped surface questions, while Mara decided which tradeoffs to accept. The draft became a shared object to examine, rather than a final answer to admire.

The useful kind of friction

In this fictional project, the machine's first contribution was a starting point. Its second was a way to test the designer's response. Creativity came from the back-and-forth: proposing, noticing, resisting, and revising. A good tool can make more options visible while leaving the reasons for choosing them firmly in human hands.