Ask HN: What are problems inappropriate for iterative solutions?

I just watched Gary Bernhardt's talk "A Whole New World" [1]. Two-year old spoiler alert: he introduces an editor and terminal that doesn't actually exist. His thesis: that some pieces of software--specifically developer infrastructure--take long periods of dedicated thought (without user discovery) to improve upon. Case in point: Clojure.

To you, what are other kinds of technology that can be clearly improved upon, but not "iteratively" in the agile and lean software sense?

[1] https://www.destroyallsoftware.com/talks/a-whole-new-world