It's not entirely clear if the test he relates (as "the test I did several years ago") in Part 3 is the same test he is referencing in Part 1. Part 1 does mention $50M worth of testing, after all.

But the test in Part 3 is to replace all instances of '|' with instances of 'e', in a chunk of English text. Keyboard users take 99 seconds, mouse users take 50 seconds.

He also implies that his experiment does not involve a cursor scheme that allows word-wise or sentence-wise movement (something present in friggin' Notepad, much less a real editor).

I'm not even gonna _start_ singing vim's praises at this point, because aside from preaching to the choir, it's like shooting fish in a barrel.