Instead of pointer, you can do keyboard-everything, which some UX enhancements go for, like the browser extension Tridactyl.

If you look at a lot of interfaces that are built for a specialized power user (e.g. cashiers), they avoid pointers and have keys for everything. Also, AutoCAD, the last I used it, looked to be centered around command-line primacy.

Sadly, POS terminals these days seem to be going in the direction of high-latency touch interfaces. Those old curses-based terminals were so fast to get things done in.

Magit[1] is another modern example of a TUI done right: discoverable, good dwim[2] inference that doesn't get in the way of experts, plus an escape hatch for typing out the exact git commands for those 5% usecases.

[1] https://magit.vc/

[2] do what I mean

The research done for the original Macintosh UI showed that keyboard users aren’t actually faster than mouse users, but think they are because they lose track of time while concentrating.

Do you have a link? I suspect there's some asterisks there. I used to operate a photo minilab and could process a roll's worth of photos in 1-2 minutes. That's 4 seconds at the outside to evaluate a photo, make brightness and color corrections, next photo. No way I'd be able to sustain the same rate by having to mouse around and click at least four different targets, bouncing from brightness to magenta/green to blue/yellow back to magenta/green to cyan/red before giving brightness a final tweak. Fitt's Law[1] is death to speed for all sorts of workflows.

I'm not saying that keyboard-based operation is superior in all cases, but a good keyboard-centric interface can eliminate the need to acquire a target (e.g. menu/toolbar item) for the most common operations because there's a hotkey. Well-understood operations can go almost at the speed of thought (either the operator's or the machine's).

[1] https://en.wikipedia.org/wiki/Fitts%27s_law

Thanks for the link. Unfortunately without any details on the experiment design it's hard for us to get anything out of it. What was the task, who were the participants, what was their familiarity with the software they were being tested on?

I can say that this talk of using the keyboard being so fascinating that it takes up significant mental resources to be not represent my experience using and seeing others use keyboard-centric interfaces. When I use magit, or when I was operating the minilab I mentioned above, I don't have to think about what key to press to do the thing I want. I am in fact "so disengaged [with the mechanics of manipulating the interface] that [I] have been able to continue thinking about the task they are trying to accomplish". Competitive StarCraft is another example that illustrates the same point without relying on personal anecdote.

The late Jef Raskin's Humane Interface had some interesting ideas, though it does not seem to have caught on.

It might be possible to do a smoother and more effective version now that we have CSS3 and all these cool transformations available.

A command-based interface is one alternative, either a text box or voice control (Siri/Cortana/Alexa,Google assistant etc.). But it's always something you sacrifice. In that case I think it's overview, especially when multi tasking. Another is full screen applications, pretty much what we have on phones and tablets. Accessable VR/AR might open up new models.

Speech, Neural lace, AR or VR.

I imagine the next revolution in UI will be because our computers change form and present in a completely different way, for example a virtual assistant that lives in the cloud and talks to you via AR visualisations and plain old speech.

Revolutions are not usually a reworking of the dominant mode but a displacement to another medium - e.g. iphone replacing computers for many people.