I finally came around to read Olivier Blanchard's State of Macro paper that was discussed in other blogs when it came out. I am not sure I have the same reading about the state of macro that he has.
While I agree there has been convergence, the real convergence, and it is phenomenal, is between microeconomics and macroeconomics. I consider macro now to be a subfield of micro, so much is the use of microfoundations pervasive. Nowadays, macroeconomists can actually be understood by people outside their field, and vice-versa. The only difference is in the kind of questions they ask.
I do not understand his aversion to reverse-engineering a model. When you observe the data, you try to find a model that fits the data. You may take a general class of models and then narrow it down to the one(s) that correspond best to the data. Heck, isn't exactly what structural estimation is doing, reverse-engineering a model?
Olivier Blanchard is an unabated neo-keynesian. He is still trying to find some way in micro-foundations to justify the Phillips curve, or even IS-LM. He uses Calvo pricing and pretends this is general equilibrium. With Calvo pricing, he also assumes from the start that rigidities matter, while empirical evidence about this is mixed. He insists on analytical solutions so that a model can be reduced to a few equations. As far as I understand it, macro has evolved since those days where you were limited by tractability. With computers, you can solve more realistic models that may actually highlight complex relationships. You can find numerical answers to questions that remain ambiguous in an analytical setting. But this is now old neo-keynesianism.
In fact, V. V. Chari, Patrick Kehoe and Ellen McGrattan seem to have written a response to the Blanchard piece, essentially claiming that the shortcuts that neo-keynesians take do not render their models useful for policy. This trio of authors has written a series of controversial papers, but the current one is one I can agree with the most readily. The new generation of neo-keynesian models are bloated, tend to deviate from microeconomic principles and add lots of free parameters in order to fit the data better. In that vein, one can do better with neural networks where basically anything goes as long as it fits. But you gain no understanding of the economy.
The greatest advance in macroeconomics since Lucas, Sargent and Prescott has been the use of solidly founded theory. Neo-keynesians have consistently fought against this once they realized their favorite equations could not be justified. Blanchard is still fighting this battle, and although he made concessions, he lost that war. The new battleground is on getting these models usable for policy. Neo-keynesians are ahead in policy circles, but only because they have again renounced on the very principle of modern macro, the consistent use of theory and microfoundations.