[Morendil]

This is called "an argument", one in which you appeal to some stereotypical descriptions of programmers which may or may not be accurate.

It's not the same thing as what we call "evidence", which consists of actually going out in the world and collecting information from more than a tiny and highly biased sample (which, like it or not, is a fair description of the set of programmers most of us will run into over the course of a career).

For instance, for one "Dudley Code-right" who is just as you describe, how many "Dudley Code-fast" exist, who are just as you describe except that no one else can understand their elegant constructions and it turns out that the "nearly bug-free" part, well... disappoints? Dudley Code-fasts look highly "productive" for a little while, then cause big problems down the road.

What we're dealing with here is the representativeness heuristic, which often makes us misjudge statistical truths. It's akin to the way people are much more afraid of flying than of driving, even though the former is orders of magnitude safer.