in the Paul Graham example, should it really parse to just 1 document? I feel like it should actually try and split on lines?