When I hear about "device-based verification" what comes to mind is a device that can be put into some kind of child safety mode, by parents who want to give their children phones or whatever. The device then "knows" whether or not its user is a child without any kind of biometrics or identification.
It has some problems and could case a lot of harm if it's badly designed, but it's the only method that seems close to workable in any conceivable form. Why is it never even talked about in these discussions?
The "you'd have to prove to someone that you're an adult" is where we disagree. I was talking about parents setting a "user is a child" flag on the devices they let their kids use. They already know who their children are, no proof is necessary. The device can then send an http header to websites for example indicating that it's a child user. That part could be mandated and standardized by law. It's 99% of the problems solved (in legal theory; obviously not every website and app in the world will choose to participate in any of these schemes) with 1% of the dangers.
So long as they don't go overboard with misguided efforts to make it impossible for children to defeat the thing, it seems fine. It's dismaying that all the proposals end up with all these ridiculously dysfunctional ideas instead.