I am currently typing this through gesture typing. That is how I do essentially all text input these days, but despite years of experience, clearly a keyboard is way better.
But the other realities of needing a way to position it relative to the body etc are too much.
Truly, there must be something that comes next. Brain interfaces are not going to be that for a very long time.
I think it is valuable to focus on the options that may be superior in every respect, not a compromise as gesture typing is. A technological advance that will render the old one obsolete.
Voice has some potential. Subvocalizations, the muscular morons that we engage in while reading, might be leveraged.
But what about tracking the fingers directing, accurately and absolutely in space? There are some approaches which use microphones on the fingertips and ultrasonic emitters with signal propagation delay to triangulate the finger position in real time. This would allow the capture of finger motions in a globally superior way, perhaps. The microphones could be adhered to the fingernails, and perhaps powered with the opt-in of the fingers so there are no wires.
If you wanted to, you could have a dummy keyboard, basically just not connected to anything, and typing would still work. This demonstrates that the features set contains all of what keyboards currently have.
We could then go backwards from considerations of how we would like to be able to twitch our fingers associated with different key commands, and build the systems to interpret those motions, instead. Not as simple as that, I know, but...
P.s. I apologize of this is considered opinion, but I googled a lot and there is absurdly little discussion on this subject. Keyboards really just can't be the best way to such a degree as all this. We gotta move on.