[mcguire]

You may be on to something. My first feeling on reading "this hypothetical writing machine..." is, "pffft; call me when you have something tangible."

The difference, I think, between a thought experiment and a hypothetical engineering construction is that the former have solid rules. You can argue for or against it without introducing new hypotheses. A Turing Machine is literally impossible because you cannot get unbounded amounts of memory, but the "rules of the game" are given; a word processor in 1963 is (probably) possible, but that is really all you know.

For example, take your quote. How is that different from a stack of note cards, a technology common in 1963? In that it is different from a stack of note cards, can modern word processors do what he describes ("compile a reordered draft quickly", from a too-complex tangle of thoughts)?