Whilst I don't doubt that having the knowledge of an actual compsci education is valuable, because you're supposed to be able to build more than one product (and build it correctly), I don't think we should quote "really good things". Homebrew is quite popular and for the short time I used MacOS, I did like it, perhaps not as much as Linux package managers but better than anything Windows had to offer at that time.
Being smart and writing beautifully architectured software is also not something that sells. What sells is software people actually want to use. I often find myself focussing too much on the first part of getting a nice architecture, and in general writing nice code. But the UI/UX is the first thing people interact with and if that part is bad, the rest doesn't even matter anymore.
I realize this is not the case in every industry, but it's not something we can just consider 'worthless'. Homebrew does a fine job (as far as I could tell) of providing a decent user experience. Even if the internals would be pure spaghetti.
That said, I do think that at google you'd look for a balance of both, and they thought he didn't have enough knowledge about the underlying theory, which is a fine reason not to hire someone.