The cost of the great number of prototypes that fail to scale far exceeds the value of the few that do go on to be workable technologies.