Effective system design requires insights drawn from serious contexts of use.
It is not enough for a tool for thought, simulation engine, or modeling substrate to be elegant in the abstract. Serious use creates pressure that toy examples cannot:
- the data is messy
- the stakes make false confidence expensive
- the users notice missing affordances immediately
- the model has to survive contact with incentives, time pressure, and partial knowledge
- the abstraction has to preserve what practitioners actually care about
This is why I distrust purely top-down designs for tools that claim to model complex systems. The math matters. The interface matters. But the work has to bootstrap near people who are already trying to act on a real system.