I came across Reactable a couple weeks ago, and have been playing with ReacTIVision, which is their computer vision engine. It seems pretty promising for quickly bootstrapping a table-based interface. (Although I do wish their calibration docs had a bit more detail.)
Thank you! Keeping things simple, even naive, has made it easy to tear it down and start over, which I've done a few times now.
As for RealTalk, the author is also interested to see what she comes up with! In the past week I've been playing with the notion of "claims" and "wishes", inspired by a description of RealTalk by Tabitha Yong, seen in this fantastic post by Omar Rizwan: https://rsnous.com/posts/notes-from-dynamicland-geokit/
That blog post was meant to document Step #1 in what I hope to be a much longer term project.
Step 1: Magic table: Tangible user interface. Interface is real-world, physical objects.
Step 2: Paper programs: Multiple programs running simultaneously. Code is visible to all. I got a basic implementation working about a week ago. You can see an experiment here: https://twitter.com/emmazsmith/status/1195544461897547776?s=...
Step 3: RealTalk: Still chewing on this one...
Perhaps I should have emphasized the "work-in-progress" aspect of the project. Point taken, lesson learned.