Back in 2012 on the EconTalk podcast, Coase stated that he didn’t appreciate the “theorem” that came to bear his name1. Emerging from his 1960 paper The Problem of Social Cost, it elevated a hypothetical world without transaction costs over the much messier world in which they actually exist. In the paper itself, that frictionless world served as a benchmark; Coase’s broader concern was how law, rights, and institutions shape outcomes once the costs of bargaining are brought back in.

In theory, if bargaining were costless, the initial allocation of rights would not prevent the parties from reaching an efficient outcome on their own. But, as Coase put it, “… it’s an imaginary system. There always are transaction costs.” In the real world, those costs can prevent bargaining altogether, which is why Coase placed so much emphasis on the two-way nature of the problem2 and on how the assignment and enforcement of rights can shape the resulting allocation of resources.3

Things feel a little different now than they did in 2012. AI agents, both enterprise4 and personal5, are flooding into the market, and when used well they can already remove meaningful amounts of everyday friction. An agent can search across thousands of options, compare terms, coordinate with services and other agents, and carry out a task that might otherwise require a long series of small decisions and interactions. And this is just the beginning. If agents continue reducing the cost of search, coordination, negotiation, and execution, the gap between Coase’s hypothetical world and our own begins to narrow.

What if the cost of transacting fell to little more than a handful of tokens? What if agents drove transaction costs so low that parties themselves could simply bargain around the legal rule? What if private negotiation became so cheap that policy increasingly set only the starting point, not the final allocation of resources? What if the “imaginary” world described by Coase turned out to be real?