We could carve out exceptions for specific niches. Ugly, but could work.

Who says the software has to be open? It would be better for sure, but companies may be interested in selling software for assistive technologies.

Interest may go up if the barrier to entry lowers. Undocumented hardware is quite the turn off.

You can't really sell software for running a braille display on its own, for example. As for interest going up, I doubt that very much, as linux is still lacking good vision accessibility, such as a screen reader that doesn't brake with every update to the gnome desktop. windows and macOS figured that one out years ago, and the only decent open source screen reader is windows only.

Braille displays were precisely on my mind: there is a sizeable community of blind programmers out there. And realistically: how hard could it be to operate such a device? The thing is basically a super-low resolution screen with an abysmal refresh rate. A sufficient design could be a serial interface where you send all the pixels in a single packet. Or perhaps we could go fancy with modifying single characters or shifting the display left or right… done right, writing a driver for this thing is nearly trivial, to the point of being possible in user-space.

One reason for the blight of the blind here is the ludicrous complexity of the software, much of it comes from the ludicrous complexity and diversity of hardware interfaces, such that there is no competition and everything is terrible.

Most people however sharply underestimate the mountain of avoidable complexity we've accreted over the decades. I would situate it between 99% and 99.9% for personal computing. No silver bullet perhaps, but I know almost for a fact that there's enough silver dust around to gather 2 or 3 bullets.