This post was originally written in Chinese and translated to English by Claude Opus 4.6. The original version is here.
Recent progress on AGI makes it feel like the real thing is coming. If most people no longer need to work because of AGI, how should society allocate resources?
The standard answer right now is market economics. Hayek's argument is well-established: no central planner can process the dispersed, tacit, local information that only the people on the ground possess. So we have to let prices emerge from countless bilateral transactions to indirectly aggregate all that information. Intervening in the market is structurally impossible at the information level. He laid out this logic clearly in his 1945 paper The Use of Knowledge in Society.
But what about big data plus AGI? If a system could read everyone's real-time consumption, location, social, and search data, and reason well enough to reconstruct anyone's reservation price — even know what you want better than you do — could dispersed information still serve as a moat?
Once that premise breaks down, non-market resource allocation, or a planned economy of sorts, starts to seem technically feasible.
Before getting into AGI, though, we need to be clear on what "planned economy" actually means. The Chinese high-school politics textbook says it's "the state making unified plans and unified allocation of resources," but in practice it's far murkier. Here's the simplest example: does Uber's surge pricing count as market or plan?
The mechanism behind Uber is precisely what Hayek declared impossible. A central algorithm collects real-time location, network status, and historical data from millions of drivers and riders, aggregates that information every second, and directly outputs a price.
And yet mainstream economics classifies Uber as a market without a second thought. Something that was theoretically impossible is already being used in our everyday lives, and it's still called a market. That makes the boundary between market and plan rather subtle.
Does "having prices" make it a market?
A naive intuition I've held for a long time: if there are prices, it's a market economy.
Under this view, the Soviet Gosplan allocating fridges by quota or exchanging grain coupons for flour is a planned economy, while Uber, Meituan, and Amazon — where prices exist, transactions happen, and people participate voluntarily — obviously qualify as markets.
That intuition doesn't hold up well. In 1936 the Polish economist Oskar Lange, in On the Economic Theory of Socialism, envisioned a form of market socialism. His model has prices too. The difference is that the prices don't emerge from decentralized bargaining; they're announced directly by a central planner.
Lange's logic: the central planner doesn't need to compute everything. The planner just announces prices, observes excess demand or supply, and adjusts accordingly. This mechanism, known as tâtonnement, could in theory perfectly simulate a market and achieve optimal allocation.
At the time Lange won the debate on theoretical grounds — his math was airtight. But he lost in practice. As Hayek countered, there was simply no central planner in the real world capable of collecting massive amounts of information in real time and adjusting prices on the fly. What the Soviet Union actually implemented was crude quantity planning; Lange's version never got off the ground.
The upshot: looking at whether prices exist isn't enough. Lange's planned model had prices too. The real question is who sets the price, based on what information, and toward what objective.
Three criteria
While hashing this out with Grok the other day, we distilled three criteria:
- Where does the price come from? Does it emerge from decentralized bilateral bargaining, or does a central authority set it?
- Who optimizes? Does each participant optimize their own utility or profit, or is there an actual global optimizer?
- How are quantities allocated? Are they the result of voluntary bilateral trade, or assigned by a central authority?
| Classical market | Classical plan | |
|---|---|---|
| Price origin | Decentralized emergence | Administrative pricing or no prices |
| Optimizer | Each participant | Central planner |
| Quantity allocation | Voluntary trade | Directed allocation |
These three are structural criteria. We don't need to consider ownership of the means of production or the role of the state — just how the system actually operates at the mechanical level.
With these criteria in hand, we can sort a few commercial forms that look market-like but aren't purely so.
Category A: Inside the firm — Coase's planned islands
The simplest scenario first. How Amazon allocates warehouse space internally, how Google assigns GPUs/TPUs across teams, how ByteDance distributes shared-platform compute to its various apps. Nearly all of this runs on internal administrative scheduling, not internal bidding.
Coase addressed this in his 1937 paper The Nature of the Firm. His question: if markets are so efficient, why do firms exist at all, rather than everyone operating as independent contractors coordinating through market contracts?
His answer: markets have transaction costs. Negotiating contracts, bargaining, and monitoring execution all cost money. For certain coordination problems, forming a firm and using internal administrative scheduling is cheaper than repeatedly contracting on the open market.
Firms are, by nature, planned islands in a market ocean. Mainstream economics accepts this without controversy.
What's interesting is today's scale. The big firm in Coase's era was Ford or Standard Oil. Today Amazon has 1.6 million employees, operates in over a dozen countries, and pulls in $600 billion in annual revenue. Its internal scheduling volume already exceeds what the Soviet Gosplan managed in the 1970s.
Walmart is similar. Leigh Phillips and Michal Rozworski, in their 2019 left-wing book The People's Republic of Walmart, argued that if Walmart's internals can run efficiently as a $500-billion-a-year planned economy, nobody can declare that planning at scale is infeasible.
Setting aside whether that conclusion overstates things, their observation lands on a fact. Today's mega-corporations genuinely run Coasean planned economies inside their walls — they just don't call them that.
This category is the easiest to classify: informal planned islands.
Category B: Two-sided platforms — the real contested ground
Uber, Didi, Meituan, DoorDash, and platforms like them. This is where it gets complicated.
Take Uber:
- Drivers and riders both participate voluntarily.
- But the price is computed and announced unilaterally by Uber, with no bilateral negotiation.
- Drivers can't pick among multiple orders with different bids; they can only accept or reject the single order dispatched by the system.
- Riders likewise can only accept the driver the system assigns.
- Uber itself is optimizing a global objective function that combines match rate, GMV, and retention.
Against the three criteria: price is set by the center, a global optimizer exists, and quantities are assigned by the center. All three lean toward planned economy.
And yet, in mainstream economics, it's still a market. A two-sided market.
There's a full body of theory behind this. Jean-Charles Rochet and Jean Tirole's 2003 paper Platform Competition in Two-Sided Markets defined this model — simultaneously serving two sides and pricing between them — as a new species. Tirole won the 2014 Nobel Prize in Economics in part for this work. Since then, antitrust law and regulatory policy have defaulted to treating platforms as a special kind of market with network externalities. In 2018 the U.S. Supreme Court in Ohio v. American Express wrote the two-sided market framework directly into antitrust jurisprudence.
The mainstream argument is simple in skeleton: no matter how powerful the central matchmaker is or how smart the algorithm, as long as participants can freely enter and exit and individually optimize, it's still a market — just one with a more assertive intermediary. Under this view, Uber is simply market infrastructure with more compute and bigger scale.
This is the consensus across global economics and policymaking. Chinese official discourse aligns too: a 2019 article in Qiushi (the CPC's flagship theory journal) titled A Rational View of Internet Platform Monopoly characterized the platform economy as an upgraded version of the market economy. Chen Long, in If Hayek Woke Up, What Would He Think of the Digital Economy?, argued that improved information-processing capacity strengthens firms' planning ability, but this ultimately reinforces rather than terminates market competition.
But there's a group of people who completely disagree with that characterization.
Evgeny Morozov in New Left Review's Digital Socialism?, the aforementioned The People's Republic of Walmart, and Tsinghua professor Yan Yilong's essay asking whether Jack Ma's "planned economy" means the same thing Chen Yun meant all point to the same thing.
When the matchmaker holds all the information, pricing power, and matching authority, "voluntary" degrades to a formality. You can decline a single dispatch or a surge price, but you can't decline the platform's power to set the price for you in the first place. This kind of formal voluntariness is a far cry from what Hayek meant by a market.
Their view is that Uber has essentially engineered the central planner Lange imagined in 1936. The irony is rich: Lange won the theoretical argument against Hayek but couldn't make it work in the twentieth century. It was Silicon Valley's capitalists, racing to capture market share, who actually built his system of real-time price announcements, global optimization, and central matching.
For now Lange's reading remains a minority position. Mainstream economics and antitrust still classify Uber as a special market with a strong intermediary, network externalities, and two-sided pricing.
But this minority has put its finger on a blind spot the mainstream framework currently cannot answer: are voluntary participation and individual optimization really sufficient to define a market? When the intermediary becomes sufficiently omniscient, does the definition break down?
Category C: Open auction — no controversy
Google's and Meta's ad auctions and high-frequency stock trading are unambiguously markets — textbook cases of market design, in fact.
Although the auction rules and reserve prices are set by the platform, bids are made independently by each participant, and the winner pays a price close to their own valuation. Prices emerge from competitive bidding, each participant optimizes individually, and allocation follows the auction outcome. The platform merely designs the rules; market mechanisms function normally.
Credit scoring is similar. It serves as a pricing input that determines loan rates. In essence it's information intermediation, not a rationing mechanism. Big data just makes traditional rating and credit assessment more granular.
This category poses no challenge to the mainstream framework.
Conclusion
Putting the three categories together: firm internals are Coasean planned islands; open auctions are well-designed pure markets; and the real contested ground lies with two-sided platforms. The mainstream sees them as two-sided markets; the minority sees them as Lange-style planning dressed in market clothing.
Under the mainstream definition, a planned economy is a regime where a center directly determines prices, quantities, and who gets what. Soviet Gosplan qualifies. North Korea qualifies. Grain rationing during the Cultural Revolution qualifies. By that definition, Uber isn't one, Walmart isn't one, Amazon isn't one.
But reality keeps shifting. As central matchmakers grow more powerful and participants' voluntariness starts to look more like accommodation, is it possible for a market to quietly cross some threshold and become a plan? Mainstream economics tends to answer: as long as there's voluntary participation, it's still a market.
Before AGI actually arrives, that line can still be held. But if a system truly emerges that can precisely predict every individual's reservation price, the market's foundations of voluntariness and individual optimization may be completely hollowed out.
By current standard definitions, the planned economy has not been revived. But the technological premises at the heart of the Hayek–Lange debate have changed beyond recognition, and that transformation is quietly unfolding in our daily lives.
References:
- Friedrich Hayek. The Use of Knowledge in Society (1945)
- Oskar Lange. On the Economic Theory of Socialism (1936)
- Ronald Coase. The Nature of the Firm (1937)
- Jean-Charles Rochet & Jean Tirole. Platform Competition in Two-Sided Markets (2003)
- Leigh Phillips & Michal Rozworski. The People's Republic of Walmart (2019)
- Evgeny Morozov. Digital Socialism?, New Left Review 116/117 (2019)
- Yan Yilong (鄢一龙). Is Jack Ma's "Planned Economy" the Same Thing Chen Yun Meant?
- Chen Long (陈龙). If Hayek Woke Up, What Would He Think of the Digital Economy?
- Qiushi (《求是》). A Rational View of Internet Platform Monopoly (2019)
- Ohio v. American Express Co., 584 U.S. __ (2018)