On Long Read Sunday, NLW reads Venkatesh Rao's 'Cognitive Mercantilism.' The essay argues that today's AI panic repeats the zero-sum thinking capitalism overturned 400 years ago. It assumes intelligence is scarce, demand for it is fixed, and most minds are about to become useless. The early modern consumer revolution suggests the opposite may happen.
AI panic is cognitive mercantilism, and history suggests demand for intelligence has no ceiling.
Mercantilists treated wealth as finite, ordinary people as labor to be commanded, and a farm-free majority as useless mouths. Capitalism proved them wrong: agricultural productivity freed most people from farming, and new desires drove a virtuous cycle of voluntary work and expanding demand. Venkatesh Rao argues that today's AI fears make the same mistake about intelligence. If demand for intelligence is as elastic as demand for material goods turned out to be, then AI doesn't create billions of useless minds. It frees them to become explorers, and the politics shifts from preserving cognitive jobs to securing a fair share of cognitive abundance. NLW adds that a good destination still requires managing the transition, and that managing it is different from trying to stop it.
Beyond security and job loss, people worry about what it means for them when an intelligence we long thought unique to humans becomes abundant. NLW's framing: we've faced this kind of question before, and 400 years of economic history suggests something quite wonderful might happen instead.
— Venkatesh Rao, "Cognitive Mercantilism". The core claim of Rao's widely circulated essay, which grew out of reading Joyce Appleby's The Relentless Revolution. If market capitalism was the first planetary intelligence, he argues, the parallels between pre-capitalist economics and today's AI fears are closer than you might think.
Appleby argues capitalism wasn't an inevitable human instinct. It was a historically contingent invention that escaped a zero-sum worldview built on land, bullion, and commanded labor. The breakthrough was seeing ordinary people as sources of demand and initiative, not merely factors of production.
Wealth is scarce and finite. Ordinary people can't be trusted with freedom. Economics is about controlling supply, while demand from the poor barely registers. And most people are good for nothing but farm labor, so removing them from the fields just creates useless mouths to feed.
Swap wealth for intelligence and the parallels are 'almost embarrassingly direct.' National and corporate strategy becomes a treasure hunt for genius brains, intellectual property, and a favorable balance of knowledge trade, and the treasure now includes AIs alongside patents and copyrighted culture.
In Rao's account, cognitive mercantilists see democratized intelligence as dangerous because the masses will misuse it or fall for nonsense. On the supply side, the priority is preserving scarcity and high prices for brains and AI. Ordinary people are treated as passive consumers, never as real sources of demand for intelligence.
Agricultural productivity shrank farming from a large majority of workers to a tiny minority, and the result was abundant food, not a world of useless mouths. Rao, who 'can barely keep a houseplant alive,' argues that most people will likewise no longer need to be significant producers of intelligence to become significant consumers of it.
NLW revisits his May episode 'The New Jobs AI Will Create.' Job-loss thinking assumes we already demand as much as we ever could. He calls that 'completely ahistorical.' Human demand is nearly limitless, constrained mostly by physical limits like hours in a day and calories we can eat.
Price (I wanted it but it cost too much), access (I couldn't get it), complexity (the system was too confusing), continuity (I'd benefit from help all the time), personalization (I value something made for me), and relational (I want it to be more human and trusted). Each one is pent-up demand that AI could release.
Tea, sugar, tobacco, and colorful Indian cottons created new desires. Annual working days rose from about 250 to 300 before the Industrial Revolution, much of it voluntarily. Productivity created surplus income, surplus income fed demand, and demand drove more productivity: a virtuous cycle.
Appleby's account points to a second force: the desire to share in an expanding universe of consumption. Once people saw that their labor could buy more than subsistence, the political question became how to secure a fair share of the surplus. Rao calls this a better model for the AI transition than 'automation replaces jobs.'
Education, consulting, research, and software were all organized around intelligence being expensive to produce. AI undermines that assumption. Demand for intelligence isn't a fixed appetite. It grows as the cost of satisfying it falls, making previously unaffordable projects and curiosities worth pursuing.
Rao argues that people who seem incapable of thought often lack the tools, access, or leverage to do anything with their thinking, so they don't bother. Abundant AI changes that math: a modest investment of attention can now unlock enormous possibilities, so the incentive to pay attention rises.
NLW's own explanation for why AI makes us work more: at work there has always been an infinite backlog, and 'done' was a negotiation shaped by natural limits on our capacity. When you can run tireless parallel copies of yourself, the whole backlog becomes unlockable, and the only constraint is imagining what belongs on it.
NLW says his backlog idea covers only the anxiety side. Rao describes the excitement side: he is willing to work significantly harder than a year ago because much more is within reach. And just as abundant food produced elaborate culinary cultures rather than killing interest in cooking, abundant intelligence may produce more ambitious cultures of thought.
Industrial labor never made preserving farm jobs its lasting demand. Rao expects AI politics to center instead on a fair share of cognitive consumption: access to intelligence, the ability to command it, and the institutional rights to participate in an intelligence-rich economy. He calls it a labor politics of abundance rather than scarcity.
Classical Malthusians feared population would outrun food. AI Malthusians fear intelligence will outrun humanity's capacity to absorb it, leaving redundant minds and collapse. Rao says both camps, and the cognitive mercantilists, share one assumption: that demand for intelligence is fundamentally limited.
Rao's most radical extension, which NLW says deserves its own essay. The point isn't that AIs become conscious consumers. It's that the systems AIs build, and the systems built around them, create new pathways and needs for consuming intelligence.
Mercantilism was a philosophy of statesmen and rentier elites, while actual merchants fought it and gave birth to capitalism. Rao argues that AI-industry leaders behave like cognitive mercantilists in proportion to their political and cultural ambitions. To that kind of elite, open weights and global diffusion register mainly as threats to status and leverage.
Rao sees a possible transition from a world where 80% of brains do subsistence cognition to one where 95% meet previously unfulfillable cognitive demands. AI's biggest consequence may not be efficiency. It may be that most minds, human and artificial, are freed from producing intelligence just to survive.
NLW notes that pieces like this pull readers past answering the work memo and into 'fuzzy, foggy places' of thought. Rao's open question is whether people avoid that exploration for lack of capacity, or for lack of the surplus that gives them time and reason. His answer is the latter, and in that world people are explorers, not replaced.
NLW stresses this isn't an argument for or against any particular AI safety concern. You can be excited about cognitive abundance and still worry about the period when AI takes over work our brains used to do. 'Transitions between orders… are never without cost,' but dealing with a transition's challenges is very different from trying to stop it.
