What is fishy? This isn't even state of the art for program synthesis. Here's another simple example trained on the Linux kernel: http://karpathy.github.io/2015/05/21/rnn-effectiveness/
I'm not sure what point you're trying to make. Do you think a neural net is not capable of generating the code in the gist? Because it's pretty easy to do that. The harder part that we're still trying to figure out is getting that code to do something meaningful.