The interpretative weakness of [Arrow-Debreu economies] lies in the information handling requirements implicitly placed on the agents without regard to cost or even feasibility. [...] Optimizing responses of economic agents are simultaneously feasible only if the proper prices are already known to them. But these prices must somehow themselves be the result of the same responses. Thus there is something circular in the description of events. The market participants must be endowed with extrasensory perception (if acting simultaneously) or with supernatural premonition (if acting successively).
Relating to my post on completeness, expectations, and efficiency, I cite the canonical statement of Tjalling C. Koopmans, American Economic Review, Vol. 64, No. 2, Papers and Proceeding (May, 1974), p. 327 (JSTOR-Link): 