A few observations of hypothes.is include
* Despite it being around a while, not a lot of people seem to generally, outside of let's all comment on this article/set of articles, use it. Notably, some seminal papers of social annotation are uncommented, by anybody! I remember more people using ICQ's long forgotten chat with others also browsing this page feature despite that being synchronous vs asynchronous comments.
* It has a 'group' function. You can filter by a group you're in. This is useful given the social constructivism that comments can give - it's constructed by a group after all, and different groups may lead to different results.
Kinda leading on from that, assigning a point is simply a numerical indicator. What does it mean? Aside from different voting types (only up? up/down? % up? what about neutral/not voted vs views? etc.), a vote's kinda positivist. So.. a badge of trust? From whom and why (the qualitative aspect)? Do you want social groupings for votes? Are all votes equal? Why not measure backlinks, or citations and do you trust these?
That's all ignoring stuff like bots, gaming, and there possibly being agreement from the entire internet whether something's good.
I visit HN because I trust the filter that's the community, a social filter self selected by choosing to be part of the HN community. As well as the front page, I also often browse new where a headline may take my interest almost as much as the front page; new is, after all, already community filtered by people that care enough about a link to submit (minus the spam). I'd actually quite like a Slashdot 'friend'/'friend of friend' follow option here - do any of the HN apps that that?
1. Often. It's often from search results.
2. No, I probably wouldn't use such an extension.
3. Despite looking into them, I don't regularly use any such tools that already exist.
Edit: Something I feel may be important that I've not really looked into: Where's the agency of the author? People directly commenting, as it would appear to a viewer, on their work. Would some authors oppose that and argue it's a derivative work [in the case of annotation]?