Besides the difficulty of the search, one problem with evolving larger software is, how do we specify the behavior that we are searching for? As you point out, if we specify test vectors, then we’re not searching for software that is correct, only software that is mostly correct. But giving a complete description of the software’s behavior could be almost as hard as writing the software.

Maybe there are problem domains where we can take a cue from biology and choose to require not software that is perfect, but software that is mostly right most of the time. Though it’s not clear that humans do better, in any case

On the other hand, if you are in a profession like investment banking or otherwise brokering mergers and acquisitions, as that author is, then a book on a narrow topic like this is unlikely to net you more income than a single additional deal would.

So the goal of the book must be to raise the author’s profile. I tend to agree with Penelope Trunk, that writing a book is no longer the best way to raise your profile: 5 Reasons why you should not write a book.