I think the main advantage might be that you have a palette of operations your program could perform next and clues (from the shape and colors) about which oper

I think the main advantage might be that you have a palette of operations your program could perform next and clues (from the shape and colors) about which operations make sense in a given context, which makes it easier to explore. Text is much more open-ended and doesn't suggest anything (although I guess that intelligent completion in the editor can provide similar benefits).