It seems to me the next "graphical language" interface should be more like working in an IDE, but instead of entering the text directly and having the hints pro

It seems to me the next "graphical language" interface should be more like working in an IDE, but instead of entering the text directly and having the hints provided - we reverse the problem and only allow the user to enter what is allowed by the syntax of the language. That would essentially generalise to "creating visual code", only now it would use the keyboard instead of the mouse.