Dijkstra's advocating a Sapir–Whorf[1] of programming languages, which I long suspected to be true; but haven't been able to rigorously prove or disprove.
What Erik describes in this paragraph is painfully apparent to me whenever I see someone interact with such a stacked desktop:
"[W]indowing was to emulate the familiar, comforting desktop, a cluttered one at that. But it is extremely difficult to use efficiently a system that displays bits and pieces of documents . . . with just their edges sticking out here and there to identify them."
Using e.g. StumpWM's windowlist, I can unambiguously navigate to some desired window without guessing from partial information.
If it is the case, furthermore, that "windowing was to emulate the familiar, comforting desktop;" I think it's a skeuomorphism that we can profitably abandon.
That sort of thing might be similar to stochastic methods like simulated annealing, where one accepts temporarily suboptimal states in case a more optimal one is around the corner.
Let psychedelics be analogous to the random walk: it may be the case that one finds oneself shaken from some local optimum only to be banished to a desert of suboptimal states and no deterministic way to return.
I decided to abandon bullet-based decks a few years ago for images and narratives (even when dealing with programming), followed by a live-coding demonstration; anecdotally, the response to talks I've given in this format has been great.
Years later, people will spontaneously write and reminisce about such and such a talk on e.g. extending Roxygen.
This is, incidentally, the same format as my daughter's bed-time stories; coincidence?
"Pithatician" is a novel use of πείθειν, "to persuade;" it reminds me of pithiatism, which is supposed to be a form of hysteria curable by persuasive suggestion.
Bizarrely: I, too, had to strain to read the page; which I attributed, however, to the weight of the font: so thin that it won't render without artifacts.
[1] http://en.wikipedia.org/wiki/Linguistic_relativity