>>Please elaborate. How does this resonate with the average user who doesn't know anything about infosec?
Elaboration, with too much pop culture... ;-)
When you use cash, for example, you're using capabilities. You can hand out exactly $3.50 to the Loch Ness Monster[1], and no matter what, he's not going to be able to leverage that into taking out your entire bank balance, etc.
The current "ambient authority" system is like handing the Loch Ness Monster your wallet, and HOPING he only takes $3.50.
Another metaphor is power outlets, which limit how much of the power from the grid makes it to your device. The current system is much like the electric - i - cal, at the Douglass house in Green Acres.[2]
The point is, you can run any program you want, and give it only the files you want, and nothing else, by default in such a system. For the user, it really doesn't have to seem that different, they already use dialog boxes to select files to open and save things, they could use a "power box"[3] instead, which looks the same, except then the OS enforces their choices.
[1] https://www.quora.com/Why-does-the-Loch-Ness-monster-want-3-...
So we'll probably stick with what we've got until AI is truly empowered to change things, which we are probably a decade away from. At that point, it is far more likely that AI will be taking in full audio, video, and data from your environment, and will know you well enough that the mundane tasks will just happen, without need for any UX at all. Maybe a small device for you to tweak things and control non-standard tasks.
But again, that is a decade off, if not two. We're currently headed into the first downturn of the AI-driven world, when the hype dies, people really spell out the problems, platforms realize that most people don't want generative AI, and all of this quiets down, taking a back burner for 7-10 years while the research advances to move beyond today's problems and evolves into what people might actually want.
One of the main things I've gotten out of the whole OpenClaw/Moltbot/Clawdbot situation is that the general public has a dangerously low grasp on information security. There's usefulness to that type of assistant, but I have yet to see a compelling, general consumer take on it.
If you don't agree, take a step back and tell me how many people prefer navigating a terminal window using a keyboard instead of a graphic interface using a mouse.
The future belongs to a more frictionless, no keyboard, voice activated UI, IMHO.
Most people don't see innovation until it is materialized in front of them.
I guess, maybe because you don't know it any better(systems and device form factors), you're trying to correct an already dumbed down(for mass acceptance) interface paradigm, with one which is even more indirect and imprecise.
We are already seeing traditional coding evaporate overnight, let alone have people memorize commands and type it like we were in the 19th century.
I'm just thinking it's not as clear-cut as you make it to be, as the past shows, multiple times. For whichever, maybe technically unrelated reasons.
Also "use it or lose it" and "learned helplessness" comes to mind.
Plus the average user doesn't care about data sovereignty, what they care about is UX and dopamine.
How many users you know of that are concerned with data collection by big tech? How much does that account for percent wise?
Furthermore I see nothing wrong with the desktop metaphor, it's just that we mostly only had a miserable magnifying glass, giving only a small viewport into a crammed childs toy, instead of real large high-resolution screens as can be had now, or sensible virtual desktops for more common sizes. To be expanded by "Metisse", an early 2.5D extension for FVWM, and later "User Interface Faćades". Maybe with some Zoomable UI sprinkled on top, like in https://eaglemode.sourceforge.net/ or whatever the clandestine weirdos from https://arcan-fe.com/ may come up with. (IF. EVER.)
Voice is natural, it is fluid, it conveys emotion, intent.
You cannot seriously be comparing metaverse immersion BS with voice commanded devices.
But one thing is seeing the content and another thing is interacting with the UI to accomplish an office task, for instance.
Since privacy will be an issue, "Shazam-like" filters will inhibit uncleared capture of voice.
Think about it. Not everyone wants to be recorded as a bystander. Privacy will be an issue.
The technology for audio signature already exists and works fine.
It will be a matter of opt-in/opt-out from users, not an OS decision.