The world as a database schema

We asked a language model to design a database schema for three things: a city, a company, and a relationship. The exercise sounds abstract, but the results were revealing.

For the city, it created tables for infrastructure, populations, zones, and flows. But interestingly, it also created a "friction" table — tracking where movement slows, where resources get stuck, where queues form.

For the company, it didn't start with "departments" or "employees." It started with "decisions" and "dependencies." The org chart was a derived view, not a primary table.

For the relationship, it modeled "expectations," "debts," and "repairs" — not just interactions. The model saw relationships as systems of accumulated obligations and resolutions.

The schemas tell a story about how AI sees structure: not as categories, but as flows and tensions.