[iaw]

> We often fail to realize how little we know about a thing until we attempt to simulate it on a computer.

I'll often write out with pen and paper of how I think the flow of a model will go, it's always fascinating how much that diagram changes once I actually build the model. There are both drastic simplifications and increases in nuance based on experience with the system being modeled and where divergences occur.

[TeMPOraL]

Same here. And this also works for existing code. When I start diagramming some complex bit of flow, I'm often surprised by how little I actually understood it. I think translating to a different medium of expression forces you to actually read the code, instead of just skimming it and thinking you understand it.