This reminds me strongly of some concepts I've heard of in TDD Ruby circles though I'm certain they're not unique to that. The bit on writing your function names first is like the practice of "write the test with the API you want to use, then make it pass with an implementation that does just that." It's a kind of planning-by-wishful-thinking that just seems to work in certain scenarios. But I think for it to be successful, you need to read enough best-practices code in your target language to understand better approaches. That said, using your phantom code first will give you a better API design. The trick, as in this article, is to keep writing ideal implementations like turtles all the way down. If your logic is complicated enough that you'll need notes or comments, descriptive function names are wonderful. Particularly in languages like Objective-C with named parameters.