A new alpha version has just been committed (v 0.20090312a), and with it a bunch of changes:
Current work on the resources leveling algorithm has been made public.
The document is yet in a pre-alpha version. It will be highly modified along with new bugs and features found, maybe even rewritten completely.
However, it gives the current directions the algorithm is taking, and is mature enough to get some feedback.
Alongside of this document, a current implementation exists in Ruby language, with regression-testing usecases. It will soon follow on this site.