Adding complexity to software has never been easier than it is right now, we really have no idea if the models will progress to the point where they can actually write large systems in a maintainable way. Taking the gamble that the models of the future will dig us out of the gigantic hole we are currently digging is bold.

Models fall prey to Kernighan's Law even more easily than human developers.

Finding bugs does not equal being able to do good architecting.