This essay relates recent developments in growth theory to problems and ideas that first engaged R. F. Harrod, E. Domar, and their neoclassical successors. The body of ‘new growth theory’ began by finding special ways to assume that there are constant returns to capital. It is shown that this is a very nonrobust assumption, thus not a good basis for growth theory. More promising is the attempt to create a genuinely endogenous theory of the process of innovation. This notion has always been present in the literature or just beneath the surface. Current ideas, for all their ingenuity, may be too mechanical.