Would they write the same code, if given the same spec? No? How about over multiple trials? No? Even the individuals' own code would differ from one trial to the next?

Well, then, their interpretations of the spec are stochastic. Good enough for me, anyway.