Seems like a great argument for clear, modular, well-written code to me. Sometimes you get dead ends, but that's true of evolution too. If your base is sufficiently flexible that it can tolerate significant changes, the amount of refactoring you have to do further down the line is diminished.
He needs to take a look at this one:
"Does Code Decay? Assessing the Evidence from Change Management Data"<p><a href="http://nisla05.niss.org/technicalreports/tr81.pdf" rel="nofollow">http://nisla05.niss.org/technicalreports/tr81.pdf</a><p>The answer is "yes", except if you undertake extraordinarily rare "perfective maintenance". Mostly, managers Just Want It Done Fast, so perfective maintenance never ever gets done, leading to devolving code.
The dead ends of the first design usually inform the design decisions of a major rewrite/refactoring. So, in that way the line of evolution is unbroken. Even moreso, if you expand the metaphor from the evolution of an individual or a species to the evolution of an ecosystem, which is really what a software project is - the code, the libraries, the deployment environments, the containers (hardware, databases, app servers, etc), etc.