That's barely possible because code has to coherently work on a shared goal. But I agree in principle -- reduce duplication of assumptions as much as possible.

That's barely possible because code has to coherently work on a shared goal. But I agree in principle -- reduce duplication of assumptions as much as possible. One way that helps with that is creating abstractions. I don't mean clever grandiose abstractions, those are extremely hard to get right. I mean precisely those abstractions that factor shared code. But even before that, it's important to get the control flow right to minimize doing the same thing multiple times in the codebase. Because even when the implementation is in a central place, calling it multiple times from different locations is what you say, duplicating assumptions.