For more information, check out the Travis CI Build.
The initial dataset and performance on that dataset will be for approximately 6-24 billion objects created, cross indexed against another 1000-50000 choices to query against. Imagine a complete list of 6-24 billion people (the theoretical population limit of Earth in some circles is 24 billion). Then imagine each of those people indexed for favorite music, sex, height, age and other demographic information. Needless to add, but I will, it'll be a huge ass amount of data to query against. Like I was saying, get subscribed and I'll be walking through this whole experiment.
Current Status: Issues/Stories on Hubaord Kanban