The Second Economy

In the AI Story

Arthur wrote the original essay during the early cloud-computing era, when the substrate's operations were largely invisible to conventional economic measurement. GDP tracked physical production and human services but missed the digital layer's expanding activity. Yet the second economy was already handling enormous volumes of coordination: supply-chain logistics, financial transactions, telecommunications routing, energy-grid management. Each domain was being automated not through single-system replacement but through interconnected networks of specialized systems learning to coordinate without human intermediation.

By 2017, Arthur updated the framework explicitly for AI: 'The main feature of this autonomous economy is not merely that it deepens the physical one. It's that it is steadily providing an external intelligence in business—one not housed internally in human workers but externally in the virtual economy's algorithms and machines.' Business processes could draw on vast libraries of intelligent functions that 'greatly boost their activities—and bit by bit render human activities obsolete.' The shift from substrate to competitor was underway.

What changed with generative AI was directionality. The second economy stopped being purely responsive—executing tasks the physical economy assigned—and became generative. It began producing outputs the physical economy had not requested: code, analysis, design, strategy, creative work emerging from the system's capabilities rather than human direction. The substrate was no longer merely supporting the visible economy. It was competing with it, offering cognitive work that human professionals performed, at a fraction of the cost and a multiple of the speed.

Arthur connected this to Keynes's 1930 prediction that technological progress would solve 'the economic problem'—producing enough to meet human needs. AI was bringing Keynes's prediction to fruition with a twist: the problem was no longer production but distribution. 'The economic problem is now one of distribution rather than production,' Arthur argued. 'The problem isn't generating jobs but providing access to what's produced.' The shift from production-problem to distribution-problem is categorical—solved not by making more but by restructuring the mechanisms through which output reaches people who need it.

Origin

Arthur developed the second-economy concept through observation of how business-to-business digital networks were quietly reorganizing global commerce in the 2000s. The framework built on his earlier work on increasing returns by recognizing that the digital substrate itself exhibited positive feedback: more digital processes attracted more integration, more integration generated more data, more data improved the systems, better systems attracted more processes. The second economy was not a static infrastructure but a growing, learning, self-organizing system.

The 2011 essay appeared before the deep learning revolution, before conversational AI, before the current generation of large language models. Arthur's prediction that the second economy would approach the physical economy's size by 2025 was based on observing the growth trajectory of digital coordination systems and extrapolating their positive feedbacks. The prediction's accuracy—and if anything, conservatism—demonstrates the framework's rigor. Arthur was not guessing but calculating based on the dynamics he had formalized over decades.

Key Ideas

The second economy is autonomous. Digital systems coordinate with each other without human intermediation, forming an economy-within-an-economy operating at machine speed.

External intelligence is the defining feature. Cognitive capability resides not in human workers but in the digital substrate, available on-demand to institutions that connect.

The directionality reversed. The substrate evolved from responsive execution to generative production, competing with rather than merely supporting human cognitive work.

Distribution replaces production as the binding constraint. When AI solves the production problem, the economic challenge becomes ensuring access to what is produced.

The second economy is now the economy. The digital substrate is absorbing function after function from the physical layer, and the question of who captures its value defines economic distribution.