As you said, big grains of salt... It would be probably more helpful if they were normalized by the percentage any particular language is used overall on Github.