His argument is that you first need to know what you are building before you build the factory to build it.

If you know what you're building, the software factory is a foregone conclusion.