IMHO "root of all evil" sits there with YAGNI. An useful way to think about the world (overoptimizing early is often as bad as overabstracting, and just as common), but a principle which shouldn't be taken to an extreme.

I think a more useful mental model is of one-way versus two-way decisions (local optimizations can be applied later., the architecture has to be right from the start), as well as expected risks and payoffs.