Would love to see a few basic libraries install ahead of time (NumPy and SciPy, namely) and hear about performance implications of running virtually.

Cool tool though - I've done data science work for a few companies now and the most frustrating thing is always getting set up on their stack.