We discuss the emergence of money in a Kiyotaki and Wright (J Polit Econ 97:927–954, 1989) environment through two computational methodologies. One assumes that agents are fully rational and coordinate on Nash equilibria. The other considers boundedly rational agents whose choices are guided by a classifier system à la Marimon et al. (J Econ Dyn Control 14:329–373, 1990). We apply the two methodologies to study the conditions under which individuals can learn to play speculative strategies—to accept a high-storage-cost good as money. Our analysis suggests that, while in both types of systems money can emerge along a dynamic path, boundedly rational agents make conflicting choices at times, even when the classifier system provides clear information about the likely gains of a trade.