Internal representation doesn't matter to me as much as the external view, so I guess I'm actually looking for your planned extensible API - I'll be eagerly awaiting that.
With more and more annotation services, there may come a need for an "annotation manager". Memex could be the manager for Hypothes.is as you suggest, but a separate manager could also use Memex and Hypothes.is as backends. So, for example, the annotation act is done with Memex, but the research stage is done with a separate tool.
Cool product/vision! I like your guys' stance on the user's data and with an API might settle on this one. :)
Maybe it is a question of needing to experience what happens though.
Most people are of the "wait and see" and "you gotta do what you gotta do to survive" mentality. So they "wait" and let events unfold, and survivors in those cascade of events are idealized as morally good.
This is a great format for drilling down! A few recommendations:
* Hovering out instantly resets, which causes you to lose your place. How about a (user-configurable?) timed gradual reset so you can get back to where you were if you accidentally hovered out for a second.
* I am having a hard time wrapping my mind around a flat list as nested - this just might be old habits. Maybe a user setting to indent nested children?
* The colors indicating score could be on more of an obvious scale - I'm having trouble instantly identifying whether green is better than purple, etc. To me, a pale color is less exciting than a 'colorful' color, but pale is used to signify most popular here.
* Maybe you can click+drag to pin multiple comments. The singletons have tiny slices and there's often many of them - it would be nice to view them all at a glance.
* Related to previous, maybe a filter for how deep the subtree is, or by score. Sometimes you don't want to see it all, or want to see only the best (or the worst).
I really like (in the radial presentation) that the newest comments are as accessible as the oldest, and that multiple sort options are available at once without the need for a round trip back to the server (the radial presents age, colors in the radial present relative popularity, the side bar presents global popularity).
This is because of ethical reasons: one of the conditions that Deepmind made with Google in their deal was for an "embargo on using its technology for military and intelligence applications."
With more and more annotation services, there may come a need for an "annotation manager". Memex could be the manager for Hypothes.is as you suggest, but a separate manager could also use Memex and Hypothes.is as backends. So, for example, the annotation act is done with Memex, but the research stage is done with a separate tool.
Cool product/vision! I like your guys' stance on the user's data and with an API might settle on this one. :)