We know that all models are wrong. We say so frequently, nod at Box, and move on. What we are less honest about is what "wrong" actually means when the thing being modelled is genuinely complex, not merely complicated, but irreducibly so. Any model of a complex system must exclude something, and what gets left out does not disappear. It keeps interacting with the rest of the system in ways we cannot predict, as the system and its environment develop over time.

This matters most at a specific seam in our practice: the transition from thinking in systems to designing them. Using systems thinking to understand a domain is a contextual activity, provisional, open, alive to surprise. Building a software system from that understanding is a mechanistic one, requiring commitment, reduction, and exclusion. These are not the same activity operating at different scales. They are epistemologically different, governed by different criteria, and damaged by being run together.

Thinking in systems is not the same as designing them. Conflating these destroys both what we seek to understand and what we wish to create.