he's saying that the various processes, methods, and tools we espouse should follow from a fundamental theory of software development. Right now, the nicest thing that can be said for the various things we espouse is that they are "laws" based on empirical observation.

But we don't have an underlying theory explaining which programming languages are more productive than others, or which processes will work and which will fail. It's a lot like chemistry before the periodical table and especially before quantum mechanics. In orther words, it's alchemy.