I've read some of his stuff on this issue, and noticed his experiment design was atrocious, and he has 0 peer-reviewed research. So, I'm taking anything he says with a grain of salt. (I've been doing user research for over 15 years, in a different context, but still I've read tons of HCI/UX work, so I'm not just talking out of my ass here)

By and large, I think vim evangelism is deeply misguided, in that it asks you to learn too many things, it sees golfing your edits down to minimal keystrokes as something other than a total waste of mental energy, it encourages heavy plugin use, and it turns the vimrc into a whole program of its own. The whole point of using an editor like vim is that you don't have to think about it very much. Which is why I encourage using the basic movement and editing keys with reckless abandon. The only thing that justifies learning a new feature is that the simple way is deeply inefficient. Not just less efficient, but so terrible that it's wearing out your fingertips and leaving you annoyed. Global search and replace with regular expressions is a good example. Sooner or later, hitting the letter n and the period key over and over to repeat a replacement is going to get so tiresome that it's worth learning regex, but a novice user can go a long while before doing that and still be quite productive.