He does give pseudocode in the paper for transforming an NP problem M<NP> into a P problem. It should be straightforward to apply this with M<NP>=SAT. (Whether it actually works is another matter...)