Take the (rhetorical) question in context. The claim was that "Everything XML wanted to be, JSON did better." Of course it can be _done_. The point to defend is that doing it like this in JSON is _better_.
It's self-evident that XML is better for the mixed content problem (which makes sense, given that mixed content is something that it was consciously designed for). Anyone arguing otherwise is, ironically, approaching the issue by narrowly considering only the things that JSON is good for and then evaluating XML against that—e.g. without it ever occurring to them that they should consider mixed content, as in the example given—or they're deluding themselves/being dishonest.
Of course we agree that the JSON representation is verbose. The question is: what is harder to use in a program? A verbose JSON representation that is just dictionaries and lists, or a concise XML one that is represented and manipulated using the DOM API?
> The question is: what is harder to use in a program?
In your imagination, maybe, but otherwise, no, it's not. The claim being prosecuted is that "Everything XML wanted to be, JSON did better." (It would be bad enough to lose sight of this once, but I explicitly repeated it for your benefit in the comment that you you are directly replying to. This iteration makes #3.)
It's self-evident that XML is better for the mixed content problem (which makes sense, given that mixed content is something that it was consciously designed for). Anyone arguing otherwise is, ironically, approaching the issue by narrowly considering only the things that JSON is good for and then evaluating XML against that—e.g. without it ever occurring to them that they should consider mixed content, as in the example given—or they're deluding themselves/being dishonest.