The second law of thermodynamics holds that entropy in a closed system increases over time. Organisations are not closed thermodynamic systems in the strict sense, but they obey the same basic principle: complexity continuously accumulates. Every decision, process, edge case, and regulation adds complexity to the system, and each of these additions is individually rational, because most of them solve a real problem at the time of introduction. The trouble is that they are almost never removed, because removal requires understanding the full set of dependencies that have grown up around them, and that understanding degrades as the system grows. Complexity, once added, is very often permanent.
The question is what this means for the organisation of economies and the role of the state.
Biology offers an instructive parallel. In biological ecosystems, individual organisms become less fit over time: they age, their environment changes, better-adapted competitors emerge. It’s difficult for the organism to reverse this process. What the ecosystem does instead is kill the organism, reallocating environmental resources to fitter alternatives. The key insight is that harshness at the level of the individual produces health at the level of the ecosystem. Death is the reset mechanism. Fitness is the selection criterion.
Capitalism operates on an identical principle, with firms in the role of organisms and markets in the role of the ecosystem. Companies compete for customers, customers choose between providers, and firms that fail to deliver value efficiently eventually die through bankruptcy, acquisition, or irrelevance. In commercial terms, fitness means delivering maximum value to customers with minimum resource consumption, and the market selects for it ruthlessly, which is how competitive economies produce efficient allocation of resources across the system as a whole.
The link to thermodynamics is that accumulated complexity is the primary enemy of commercial fitness. If all organisations accumulate complexity over time, and accumulated complexity drives up costs and degrades responsiveness, then every incumbent firm gradually becomes less fit, even as it appears to be running normally. The trouble starts when a new entrant arrives with no legacy at all. Netflix started with streaming video into people’s homes from a blank slate while Comcast was buried under decades of set-top boxes, bundled packages, truck rolls, and byzantine contractual obligations. Tesla entered the auto industry unencumbered by dealer franchise laws, union agreements stretching back generations, and supplier relationships with their own accumulated dependencies. In each case, the challenger won not by being smarter about managing complexity, but by not having any.
This mechanism is the means by which societies can make technological progress without carrying legacy systems forward indefinitely. The incumbent has to integrate every new technology with existing systems, which is expensive, slow, and dangerous. The challenger can build on new technology from scratch. AI is just the latest example in a long line of disruptive technologies, and although it brings benefits to incumbent firms, new entrants can use it to replicate in weeks what it took incumbents years to build.
The critical point is what happens next. The challenger offers better products at a lower cost. Customers migrate to the challenger. The incumbent shrinks or dies. The legacy system, with all its accumulated complexity, is simply deleted. Nobody has to migrate the old COBOL codebase to a modern language or untangle decades of accumulated dependencies. The market solves the legacy problem not by fixing the old system but by making it irrelevant.
In contrast to private markets, the state has no competition, no customer choice, and no death mechanism. The feedback loop between accumulated complexity and organisational death is entirely severed, and the consequences are measurable. The US Federal Register ran to about 20,000 pages in 1960 and exceeded 90,000 by 2023. The US tax code has grown from approximately 400,000 words in 1955 to over 10 million today. The number of federal agencies and sub-agencies is somewhere north of 400, though nobody can produce a definitive count, which is itself a fine illustration of the problem.
Without a death mechanism, there is no renewal without expensive migration either. Government agencies still run critical infrastructure on COBOL, because the accumulated complexity is so deep that replacement costs run to tens of billions of dollars and years of effort.
The result is that state-provided services get slower, more expensive, and worse over time. Healthcare is the most vivid example. The NHS, which started as a lean and elegant idea in 1948, now employs over 1.5 million people, and nobody would describe its current administrative apparatus as lean or elegant. Britain spends an increasing share of GDP on healthcare and gets outcomes that lag behind countries spending less. Singapore, by contrast, relies on individual health savings accounts with government subsidies rather than direct state provision, and consistently achieves among the best health outcomes in the world at a fraction of the cost. The difference is structural: Singapore’s system harnesses market competition to keep providers lean, while the NHS is a monopoly provider accumulating complexity with no competitive pressure and no death mechanism.
If the system cannot die, perhaps it can be reformed. This is the hope that periodically grips governments of all political persuasions, usually around election time, and it has a fundamental problem: the more complexity a system has accumulated, the less anyone understands about how it fits together, and therefore the harder it is to safely remove any part of it. The people who understand the system have no incentive to simplify it, because their roles, budgets, and institutional power are functions of its current size. Turkeys do not vote for Christmas. The people who have the motivation to cut lack the knowledge. They are performing surgery on a patient whose anatomy they do not understand.
The recent attempt by DOGE illustrates this trap. Silicon Valley’s entire skillset is building simple systems from scratch on a blank slate and letting the market decide which is best. Reforming a vast legacy system from the inside is the opposite of this: it is the difference between architecture and archaeology, and DOGE brought architects to an archaeological dig.
If the state cannot shed complexity, and accumulated complexity degrades outcomes over time, then the thermodynamic prescription is straightforward: minimise the complexity the state takes on in the first place.
This argument has a natural boundary. Defence, law enforcement, the legal system, and core infrastructure are genuinely public goods that only the state can provide, and some industries are natural monopolies where competition is structurally impossible. The thermodynamic argument has nothing to say about these. It applies specifically to the large category of goods and services that the state provides despite the existence of viable private alternatives: healthcare, education, housing, transport, energy, and many others.
Within that category, private provision should be preferred wherever viable, because only private provision is subject to competition, survival of the fittest, and the death mechanism that keeps complexity in check. Where the state does need to intervene, it should prefer direct wealth transfers over service provision. A wealth transfer is a much simpler operation: the state collects tax and sends a cheque. Tax collection is already well-solved, and the transfer itself requires almost no additional institutional infrastructure beyond means-testing and fraud prevention. Service provision, by contrast, requires the state to build and operate exactly the kind of complex system it is structurally worst at maintaining: a complexity factory with a permanent workforce and no off switch.
The prescription is not that people should be denied access to services. It is that the state should give people money and let them buy services from providers operating in competitive markets, where the death mechanism keeps providers lean and subject to renewal. The end result for the citizen is better: access to more services at a lower overall cost to the state, delivered by a system that is structurally capable of staying efficient over time.
The honest version of this argument is uncomfortable for everyone. The left has to accept that system size is a structural risk independent of intentions, and that building institutions to deliver what cash transfers could accomplish more simply is thermodynamically reckless. The right has to accept that the private sector companies aren’t necessarily intrinsically better run. The free market just kills companies when they inevitably decay.
A system with no death mechanism must be kept small, not because big government is morally wrong, but because it is thermodynamically unsustainable. You can fight entropy locally and temporarily. You cannot win if you try to fight it globally, forever. The only question is how complex our Governments can grow before they cease to function effectively at all, and the answer to that question has nothing to do with left or right. It is a function of physics.