The reason for this seems self-evident: if you have time to learn the basics of many different fields, you don't have time to learn any of their finer nuances. The process of learning to put those building blocks (e.g. for loops, expressions, variables) together in specific ways with outcomes you can predict (e.g. computing y to the N th) is what I'm calling creating a mental model of computer programming. A good generalist is necessarily good at spotting patterns: even if a model is different—different labels, different ways to clump things together—it might be fully or partly equivalent to other models in different disciplines in subtle ways.