Correct, that doesn't work in the system as implemented. But my understanding is that there has been a goal from the start to _eventually_ get physical code editing to work. I recall hearing the researchers talking about wanting to eventually get handwriting recognition working, so you could edit code by hand.

Cameras good enough to do that kind of work aren't really available right now. It does still make the code more accessible by making it immediately visible on the object so if you want to know how a thing works you can just pick it up.