Am I the only one who finds the "evaluation as conversation" idea incredibly annoying? I mean, it's not like I can't run the conversations 100x times faster in my head and with less "visual noise". And teaching you to run code inside your head should be the main purpose of such exercises - outsourcing to an interpreter you have a conversation with defeats this purpose!
If I had been asked this question earlier, I might have said "It could be useful". Having tried this page, though, I agree with you. It was painful and distracting to step through the first example, and I can't imagine doing it for non-trivial pieces of code. At least for me, it would devolve into clicking "Next" blindly until the end.
