It's a pity I can't read most of the text in this demo when playing the video on my 13" laptop screen. But even without the details, it's very impressive!
Coming from my current Smalltalk-dominated point of view, I recognize the "textual code in-the-small" approach, but with a different structure in-the-large. Whereas a Smalltalk image is a database of code, a Data Rabbit canvas is a graph with code in the nodes and data in the edges.
Makes me wonder: how large is the set of useful large-scale structures for information processing systems? Is it combinations of a small number of primitives, or a large number of non-decomposable structures?