Perhaps as a proof of concept they can do a native implementation of GNU grep in Haskell and publish some timing stats.