Under certain assumptions, monopolistic competition with CES preferences is efficient, as first discovered by Dixit and Stiglitz. One assumption, invariably left implicit, is that there are, at any given point in time, no bounds on the number of products that can be discovered. But square wheels do not work, and round wheels keep getting rediscovered. Giving away patents to entrepreneurs who happen to be the first to discover a product generates an inefficiently large amount of variety. The stock of undiscovered products is a commons that can attract too many discovery attempts. Perpetual patents can be efficient, but only when combined with just the right tax on patent-protected monopoly profits. Such a tax is, however, too crude an instrument in an economy with even the least amount of heterogeneity.