Questions of convergence under an infinite time horizon will depend so much on epsilontic refinements in the system of assumptions - and on the infinite constancy of these refinements - that we are humanly speaking absolutely certain of getting infinite time horizon results which have no relevance to concrete reality. And in particular we are absolutely certain of getting irrelevant results if such epsilontic exercises are made under the assumption of a constant technology. 'In the long run we are all dead.' These words by Keynes ought to be engraved in marble and put on the desk of all epsilontologists, in growth theory under an infinite horizon.