At first it seemed like it would be posted automatically. But you read and interpret the data for your little friend, right?
The phrases are random, check the code for them here: https://github.com/jlhonora/iot/blob/master/rpi/twitter_mana... . The "mood" for the phrases depends on the ran distance.
The graph I posted was generated with matplotlip and I tweeted it manually. When I finish struggling with D3 I'll put an on-demand graph with the data.
The activity graph (shown on the twitter feed) is for that same night, which seems to be quite consistent over time. There are no peaks either, so I think it should be accurate. After all, he does run all night.