It's often said that programming languages should have "a small set of orthogonal primitives" but I wonder if there's any empirical basis for this?

A highly relevant thread (thx @missingfaktor): twitter.com/slava_pestov/s…