According to some super interesting research by Dr. Holger Hennig, it turns out that when humans play music together they don't keep a perfect beat. Not only that, but the imperfections in their beat obey certain statistical laws as they interact with each other.
This has been modelled in code to produce the attached CSV of "realistic" human deviations which has then been rendered into an audio track I coded in Sonic Pi. For more information on the humanizer, check out the github repo
This is really nice! And the result really sounds more human, wow.