In the world of both software and entrepreneurship, people recommend starting with a basic vision and relentlessly and incrementally improving it. So whats wrong with that approach applied to science.