I find this approach weirdly frustrating, despite finding it compelling, and I don't know if I can make the cause for that explicit... I guess I worry that we don't conceive of algorithms this way, we just express them this way to illustrate them. I see the value in the illustration, but I want it built for me from some other source. Otherwise I have to build my algorithm twice, and this way feels harder.

Thanks for the comment! This was partially inspired by systems like pythontutor, where they do exactly that: take a source and visualise it. This project is explicitly about exploring the other direction. I haven't yet convinced myself that this a comfortable way of programming either. The demos are all done in a very straight ahead beginning-to-end way, and I've been thinking a lot about how to better support exploration and iteration.