Why have we discovered so few local optima outside of general-purpose programming? There are important classes of problems that don't require the full complexit

Why have we discovered so few local optima outside of general-purpose programming? There are important classes of problems that don't require the full complexity of a general language. It can be much simpler to solve those in a limited language, and that gives the programmer more power. Limited languages also give implementers more power, since they can exploit invariants that hold for programs written in that language, that are not true of programs written in general languages, to gain efficiency wins. This is why the author is able to say that Mawk "wins on both LOC and performance criteria — a rare feat indeed, transcending the usual competition of slow-but-easy scripting languages versus fast-but-hard compiled languages". The fact that that kind of win-win can't be had in general-purpose languages seems like a strong incentive to pursue more specialized ones. I wonder how many subspaces there are that are small enough to be exploitable in this way yet large enough to address valuable classes of problems. It feels like we should have more of these.