If you're ramping up, you cannot write a significant amount of code without errors. If you try, it's much harder to figure out which change caused the failure o

If you're ramping up, you cannot write a significant amount of code without errors. If you try, it's much harder to figure out which change caused the failure of a previously working system, because you're less aware - by definition, ramping up - of the mapping between edits and failure modes. Once you're ramped up, yes, it's more efficient to produce more code and fix the odd issue that crops up. I rely more heavily on high level, slower running functional tests on code where I've been the primary author or have worked on the system for a long time. But just be aware of what you're optimizing for: tenured employees. I wouldn't kid yourself that it's "deep reasoning", beyond people on a dopamine kick, or beginning programmers working with simple data structures. You can't reason much about a system you don't know, because everything is details - millions of lines of code, you can't mentally evaluate it. It's not until you've built a mental model of the moving parts, and are able to mentally abstract away details not relevant to your task, that you can think things through in the large.