Comment by blobbers

What happens when you find the models are terrible? The claimed results don't match? My dev cycle tends to be write a test for blah blah, add feature to satisfy test, make sure tests pass.