I get @TomCritchlow’s sentiment, but… the extra “work” it currently entails for the social part dramatically ups the signal to noise ratio for me compared to Twitter.
You’d definitely want the ability to filter by your social circle, especially on popular sites. In fact this sort of discovery mechanism would be cool if it could be more broadly built into either the web or perhaps into IndieWeb social readers which would know your social graph and could surface related details.
Perhaps expanding a browser extension like Crowdwise to include Twitter support might be a potential solution? I would worry that portions wouldn’t add much other than a lot of likes and bookmark-like data. While some,  might consider Twitter as an annotation layer (not always directly linked) on the web, the overall quality isn’t necessarily going to be built in there.
It would be cool if Crowdwise also added Hypothes.is’ API to their list of sources.
I’m also reminded of Peter Hagen’s experiments with Hypothes.is seem very similar but with a different UI. His version flips the discovery question on its head.