This is mentioned in the comments on youtube as well. The reply from author is "That's because there was a half-second near the end in my driveway where the thinning of my hair was much more visible than I was comfortable with, so I edited that part out."
haven't tried, but if you train it using only correct Python I bet it can only produce correct Python. Have to check this out to be totally sure of the absence of a weird corner case.
A small followup to Jurgen's post. Python's indentation is meaningful so any change to it would mean we changed the program. In that sense, python is not a good target for this tool. Also, as a simple implementation expedient for this version, I assume that '\n' is not significant.
Identity: there exists one unique element e such that for any a it holds that e⋆a = a⋆e = a. This e is the same no matter what a you pick.
Invertibility: given any a, there is b such that a⋆b = b⋆a = e. For two different a's you get two different b's.
As example let G be a set of all strings and the operation is appending. There is element e == "". For any string a it holds that if you append empty string to any side, you get the original string back. However there is no inverse. You can not append two string to form an empty string. Apparently, strings with appending do not form a group.
Group example: let G be a set of all integers and the operation is addition. Then e is 0, as for all a it holds that a + 0 = 0 + a = a, and for each a you can find -a such that a + (-a) = (-a) + a = 0.
I had serious trouble with the demo, though. In order to figure out the switches at the top I had to dig in the code. Apparently when it displays off, it is actually on. A little color would help here.
