I like seeing these different ways of using the underlying Steem database.
I have a few notes.
It seems that the Screem UI is only paying attention to whether the post has the #screem tag and not whether it has the "screem" value in the "used-by" field of the post's JSON metadata. For example, the post that I am replying to is included in the Screem feed (using its title as the content of the "tweet") even though it doesn't have the "used-by" field in it is JSON metadata. Steemit users may wish to tag their long-form posts with #screem to discuss the Screem platform, but may not wish for their post to show up as if it was a "tweet" (what are you calling it instead, a "screem"?) in the Screem feed. So I think it makes more sense to just filter based on the "used-by" field you already create using your UI.
Using posts for micro-blogging is not the best idea since the change to Steem which put a soft limit on the number of posts in a day before a penalty is added to post rewards. Every tweet/screem people create using Screem will eat up into their quota and lower their potential payouts on the long-form posts they create on Steemit. What I would like to see instead is for Screem to utilize comments for its tweets/screems so that users can micro-blog as much as they want without hurting their post payout rewards. However because there is a time limit of 31 to 44 days after the top-level parent post of a discussion is created before no new comments (or at least the regular type that can earn money) can be posted on that discussion thread, your platform would have to be a little clever with how it manages the comments and discussion threads. It would essentially need to create a new post for the author once a month so that it always has a recent designated discussion thread in which to post new comments by the author that act as their tweets/screems.
Other ideas for future improvements include using the JSON metadata to refer to another comment that is also acting as a tweet/screem, in either a way that acts similar to a retweet on Twitter or in a different way that acts similar to embedding a tweet on Twitter. The Screem UI could then enable similar features (rescreeming and embedding screems) and represent them appropriately by taking advantage of the appropriate JSON metadata.
Finally, I noticed that the time displayed on the tweet/screem in your UI uses the last edit time rather than the creation time. Was that a deliberate decision? I think it would be nicer to show the creation time but also include in parentheses the last time it was edited (if it even was ever edited).
Everything you said is spot on. We are doing it this way from now on.