Joel-on-Software gives google-suggest a thumbs down for the moment.
Big leap, details not quite right. Gotta have both to get somewhere. Power laws again.
Very cool tech though, easy to see there's huge potential there... which breaks my Metropolis-Hastings Algorithm paradigm. Metropolis-Hastings would require both the big leap and the details to be correct simultaneously to make the jump. "Might be good" is more of a branch-and-bound / interval-arithmetic type of reasoning.
Hmm, branch-and-bound is rather more digital and symbolic than the analogue Metropolis-Hastings algorithm. More about abstraction. Very different reasoning style. Not really my cup of tea, but other people seem to get good results from it. Left brain right brain thing perhaps?