I think all the "optimization" was a mistake. Instead of optimizing for the most-used keys, I think it's better to optimize to make it easy to remember.
Phone typing was easy to learn because the letters were both predictable (alphabetical) and because you could look at the letters when you forgot, especially while learning.
Since looking at the keys isn't possible here, I think it's even more important that being able to predict the locations of letters is possible and easy.
T9 was amazing. Any upcoming human interface industrial engineer ought to study it. When driving or in class, I used to regularly send SMS messages without looking at my phone’s screen. It was THE biggest thing I missed when I got my first iPhone 3G.
Only recently have Google's and Apple's TTS come close to matching it. Occasionally I still miss T9 like when in a movie theater where I can't speak to type and don't want to look at my screen.
I'm a bit of a keyboard maniac. I don't know why I am this way, but I am. On my computer I have a lovely split mechanical keyboard which I've remapped to Colemak, but I've always wished I could get a good typing experience on my phone.
I like being able to type while walking and without looking, and by far the best phone keyboard for that is 8pen [0]. A masterclass in innovative UX imo, and clearly a ton of work went into it. Unfortunately, they pulled it from the play store (and deleted all the tweets on their twitter) some years ago for reasons unknown to me.
Related is MessageEase [1], which is not as good but still passable. But it's incredibly ugly and takes up a ton of space on my screen, so I don't like to use it.
The old steam gamepad keyboard was also predictable in this way; I don't know if it's still used, I just remember being impressed by the concept. With two sticks and shoulder buttons I think you could really get somewhere with this idea
Steam's keyboard was mentioned in the Gamasutra article linked to in the article[1], which I found more interesting than the original article. I'd recommend reading it; the two-stick entry method detailed by the end of the article makes basically every game-pad entry system I've ever used feel archaic.
That's really interesting! The new steam controller keyboard takes advantage of the dual touchpads to give you the split hunt and peck as described in the study. You can see how there are larger targets for certain letters, and more total inputs that can be picked. However, you have to look at the keyboard to type, it's not a solution for the touch typing problem.
Having experience with precise angles on a gamecube controller in Super Smash Bros Melee, I can't help but feel like I would prefer the penultimate solution in the article you linked, with appropriate zones that can be targeted directly. Locking you to a quadrant and forcing most letters to be a compound input doesn't seem worth it to me. You could also definitely improve the quadrant system by allowing you to break into a new quadrant without returning to the center, maybe by passing the vertical or horizontal center line, depending on the quadrant. But, I obviously haven't tried it myself.
It was the Big Picture "Daisy Wheel" keyboard. They've since removed it to focus on what I think is a much weaker controller keyboard designed primarily for the Steam Controller that is pretty garbage for standard xbox controllers.
That would be helpful in the very beginning but even with just a little practice the gestures become muscle memory and remembering the gestures a non-issue. I believe making the gestures predictable would have greatly decreased efficiency which is my priority here as my goal is to be able to use this system for work not just for the occasional text message.
If you're the only one using it, then it doesn't matter as much. Do whatever you like.
But if you want others to use it, remember that they are different than you and even learning such a system is a big investment for many people.
Qwerty keyboards work because people can see the keys and they're everywhere. When you come across a keyboard of another layout, even just alphabetical, it's a big pain.
If people are going back and forth from your layout to any other kind, it's going to be a huge headache. And if they put down your layout for more than a couple weeks, they're going to have to re-learn it. Having it be predictable is a lot more desirable if there's ever a time you have to re-learn it.
That's not the optimisation used for Qwerty, Dvorak, etc. By optimising for memorisation you're sacrificing ease of use for power users for the sake of beginners who will mostly give up. That makes sense for a product you're trying to sell but not for a niche, hobbyist thing like this where you're trying to supercede keyboards.
Phone typing was easy to learn because the letters were both predictable (alphabetical) and because you could look at the letters when you forgot, especially while learning.
Since looking at the keys isn't possible here, I think it's even more important that being able to predict the locations of letters is possible and easy.