Balance of Power #

Abstract #

This is my remix of Vitalik’s Dec 30 essay of the same title.

Modern civilization depends on powerful institutions—corporations, governments, and mass movements—to drive progress, yet these same forces threaten freedom and flourishing as they concentrate power. This essay examines how Big Business, Big Government, and Big Mob each become dangerous through specific mechanisms: corporations through profit optimization that diverges from social welfare and through homogenization driven by scale; governments when they shift from neutral frameworks to active agents pursuing their own agendas; civil society when diverse institutions collapse into unified mobs.

Historically, diseconomies of scale and technological diffusion counterbalanced concentration. Today, automation and proprietary systems that distribute functionality without control have weakened these checks, enabling super-exponential growth where small advantages become permanent dominance.

The solution is “mandated diffusion”—forcing powerful actors to share technological capabilities through adversarial interoperability, copyleft licensing, standardization requirements, and similar mechanisms. Coupled with “defensive acceleration” that democratizes security technologies, this approach preserves rapid progress while preventing hegemonic control, embodying a pluralist ethic that prohibits hegemony while encouraging impact.

Introduction #

We find ourselves in a peculiar predicament. We want innovation, prosperity, and progress—yet we’re terrified of the very institutions that deliver them. This isn’t irrational anxiety. It’s recognition of a genuine problem, one that becomes more urgent with each passing year.

Let me be more specific. When I say we fear our most powerful institutions, I mean three distinct forces:

Big Business gives us smartphones and efficient logistics, but also predatory monopolies and companies that reshape governments for profit.

Big Government provides law, order, and public services, but also surveillance states, censorship, and the capacity for systematic oppression.

Big Mob—civil society in its less organized forms—produces Wikipedia and charitable movements, but also lynch mobs, cancel culture, and the kind of revolutionary excess that consumed the French Revolution’s later stages.

The conventional response is to shrug and accept this as the price of civilization. But I want to argue for something more ambitious: that we can identify the specific mechanisms by which power concentrates and becomes dangerous, and that we can design interventions to preserve the benefits while mitigating the harms.

The key concept is balance of power—both within each force and between them. But we’re getting ahead of ourselves. First, we need to understand exactly what we’re afraid of.

Concerns #

Government #

There’s a straightforward reason to fear government that distinguishes it from other powerful institutions: governments have guns, and they’re legally permitted to use them against you. Mark Zuckerberg might annoy you, but he cannot imprison you. Your government can.

This is why centuries of liberal political theory have obsessed over a single question: how do we get the benefits of government—dispute resolution, law enforcement, public infrastructure—without giving someone the arbitrary power to destroy our lives?

Here’s the central insight, expressed as simply as possible: The government should act like a game, not like a player.

Let me explain what this means. A game is a stable framework, a playing field with consistent rules. A player is an agent actively pursuing its own goals within that framework. When your government functions as a game, it provides predictable legal structures that let you plan your life and resolve disputes. When it functions as a player, it picks winners and losers based on its own agenda.

This distinction shows up in various forms across political philosophy:

Libertarianism says the game should have essentially three rules: don’t defraud, don’t steal, don’t kill—and that’s it.

Hayekian liberalism (named after economist Friedrich Hayek) argues that when government must intervene, it should set goals rather than methods. Want to reduce carbon emissions? Tax carbon and let the market figure out how. Don’t mandate specific technologies.

Rule of law emphasizes that government should act through general laws rather than case-by-case decisions, making the system more game-like and less dependent on who’s currently in power.

The subsidiarity principle states that decisions should be made at the most local level capable of handling them—avoiding concentration of decision-making power.

Recent research on authoritarian governments found that even among non-democratic regimes, the ones classified as “institutionalized” (operating through predictable procedures) deliver consistently better economic growth than “personalistic” ones (operating through individual leader’s whims). The game-versus-player distinction matters even when the game isn’t democratic.

Now, you might wonder: doesn’t government sometimes need to be a player? Doesn’t an army facing an external threat need decisive leadership rather than just a stable framework? The answer is yes—but this was understood even in ancient Rome. The Romans would elect a temporary dictator during emergencies, granting extraordinary powers that would automatically expire when the crisis ended. The key was making the transition explicit and limited.

Corporations #

Let me propose a useful distinction for understanding corporate problems. Corporations can be criticized on two grounds:

1. They’re evil (pursuing profit at society’s expense)
2. They’re lame (producing boring, homogeneous outcomes)

Corporations are optimization machines. As they grow more powerful, the gap between “maximize profit” and “maximize user welfare” or “maximize societal benefit” grows wider.

Consider social media platforms. Early Facebook and Twitter connected friends and facilitated conversation. As these platforms grew and faced pressure to maximize engagement (and thus advertising revenue), their algorithms increasingly promoted content triggering strong emotional reactions—outrage, anxiety, tribal affiliation—because such content kept users scrolling. The platforms’ optimization capability grew dramatically through better AI, more data, and refined psychological models. But this optimization wasn’t aligned with user wellbeing. The platforms became extraordinarily good at achieving their goal (maximize engagement), which turned out to be quite different from what users actually wanted (meaningful connection).

The pharmaceutical industry shows a similar pattern. When drug companies were smaller, their profit motive aligned reasonably well with patient welfare—make effective drugs, sell them to doctors. As companies grew larger with more sophisticated marketing capabilities, a different strategy emerged: spend more on marketing than research, focus on drugs that can be heavily advertised rather than breakthrough treatments, extend patents through minor modifications. The opioid crisis provides the darkest example: Purdue Pharma had the resources and reach to run a coordinated campaign convincing doctors that OxyContin was less addictive than it was, while knowing from their own data that addiction was rampant. A smaller company lacking such capabilities couldn’t have caused comparable harm even if it wanted to.

But there’s a second, equally important criticism that gets less attention: corporations being lame. This is the complaint about architectural monoculture—every city looking identical, Hollywood producing ten interchangeable movies, Starbucks making urban spaces feel homogeneous across continents.

The word “soulless” is fascinating here because it bridges both criticisms. We call it soulless when corporations addict people for clicks or pollute rivers (evil), but we also call it soulless when they make every city look the same or produce formulaic entertainment (lame).

I want to argue that both forms of soullessness share two root causes:

Commonality of motive: Corporations are all strongly motivated by profit. When many powerful actors pull in the same direction without countervailing forces, you get convergence.

Commonality of agency: Large companies have more incentive to shape their environment than small ones. The benefits of environment-shaping scale with size. A $1 billion company will spend far more on lobbying or cultural manipulation than a hundred $10 million companies combined—and it will also contribute more to the feeling of urban homogeneity.

Here’s the mathematical intuition: monopolists price above marginal cost because they can benefit from “bending” the market price through restricting supply. The amount of bending you can profitably do is proportional to your market share. But this logic extends far beyond pricing. It applies to lobbying, to cultural manipulation campaigns (remember De Beers and the diamond engagement ring tradition?), to any situation where reshaping your environment brings benefits proportional to your size.

Investors amplify both dynamics. Consider the difference in incentives: a startup founder might genuinely prefer building a $1 billion company that helps the world over a $50 billion company that harms it. But investors are insulated from these non-financial consequences. In competitive markets, this creates a selection pressure: investors willing to pursue the destructive $50 billion path earn higher returns, while those who stop at the ethical $1 billion earn lower returns and eventually lose access to capital.

There’s a second mechanism at work. Investors with stakes in multiple companies can subtly coordinate their portfolio—encouraging these nominally independent firms to behave more like divisions of a single mega-corporation. This effect has natural limits, though: investors can only push coordination as far as their ability to monitor what actually happens inside each company allows.

Market competition addresses commonality of agency (by creating multiple independent actors), but only addresses commonality of motive to the extent that competitors have genuinely different, non-profit motivations. Sometimes they do—companies often sacrifice profits to release innovations openly or uphold deeply held values or aesthetic visions. But this isn’t guaranteed.

If commonality of motive plus commonality of agency creates soullessness, then what is “soul”? I’d argue that soul is simply pluralism—the set of things about corporations that aren’t homogeneous between them.

Mob #

When people praise civil society—the part of society that’s neither profit-driven nor governmental—they invariably describe it as consisting of many independent institutions pursuing different goals. Think of charities, Wikipedia, professional associations, hobbyist communities, religious organizations, local sports leagues. When people criticize populism, they envision something fundamentally different from healthy civil society: a charismatic leader who claims to speak for “the people” as a unified whole.

The danger isn’t that populism mobilizes common people—that can be valuable. The danger is the fiction of unity it requires. Populism treats “the people” as having a single will, discoverable through the leader’s intuition rather than through the grinding work of negotiation between different groups with legitimately different interests. This fiction has several pernicious consequences:

First, it eliminates the space for legitimate disagreement. If there’s a unified “will of the people” and the leader embodies it, then opposition isn’t just wrong—it’s a betrayal of the people themselves. You’re either with us or against us, part of the authentic people or part of the corrupt elite/foreign influence/internal enemy.

Second, it concentrates power through direct emotional connection rather than institutional mediation. The leader appeals directly to millions, bypassing the organizations, professional bodies, local associations, and intermediate institutions that normally translate, filter, and moderate political demands. These intermediate institutions aren’t just obstacles to overcome—they’re how diverse interests get represented and reconciled. Without them, you get mass mobilization without the stabilizing friction that prevents it from becoming dangerous.

Third, populism inverts the healthy relationship between leaders and institutions. In well-functioning systems, leaders operate within institutional constraints—they may push boundaries, but the institutions retain independent authority. In populist movements, institutions exist only to the extent they serve the leader’s connection with “the people.” Courts, media, universities, civil service—all are legitimate only insofar as they align with the popular will as the leader defines it.

The word “mob” captures something important here. A mob is what happens when many people act collectively without the structure that diverse, independent institutions provide. It’s not that the individuals in a mob are bad people—it’s that they’re acting through a mode of organization (or disorganization) that eliminates the friction, debate, and reality-checking that institutions provide.

Balance Between Forces #

So far we’ve discussed balance within each force. But balance between forces matters equally.

Capitalist democracy represents a theory of balance between Big Government and Big Business. Entrepreneurs get legal tools to challenge aggressive government action and enough concentrated capital to act independently. Simultaneously, governments retain the power to regulate corporations. Neither dominates.

There’s a school of thought that lionizes billionaires that pursue unconventional, detailed personal visions rather than directly maximizing profit. SpaceX’s Starship, for instance, was created neither by profit motive nor government mandate, but by an individual with resources and vision. This represents threading the needle: getting capitalism’s benefits without its worst features.

My own views on philanthropy follow similar logic. I strongly support billionaire philanthropy—but specifically the kind that counterbalances other forces. Markets often won’t fund public goods. Governments often won’t fund things lacking elite consensus or whose benefits don’t concentrate in any single country. Some things fall into both gaps simultaneously. Wealthy individuals can fill these gaps.

But billionaire philanthropy becomes harmful when it stops counterbalancing government and instead takes it over. This has happened in Silicon Valley recently. Powerful tech CEOs and venture capitalists have become less libertarian, less focused on creating alternatives to government (“exit”), and more engaged in directly pushing government toward their preferred ends—while simultaneously making the world’s most powerful governments even more powerful.

I much prefer the early 2010s Silicon Valley attitude of building alternatives as opposed to the current attitude of merging with power. The former expressed balance of power; the latter represents two forces that should balance each other instead forming a unified bloc.

Balance can exist between all three corners of the triangle. The Enlightenment concept of the “Fourth Estate” positioned the press and civil society as checks on government power. Meanwhile, governments fund schools and universities, shaping (but not determining) the content of education. Media reports on business; business figures fund media. These mechanics are healthy as long as no single direction of flow dominates.

The Economies of Scale Problem #

Here’s a simple explanation for America’s rise in the 20th century and China’s rise in the 21st: economies of scale. Both are large countries where you can easily scale businesses and institutions to hundreds of millions of people. This is often cited as a criticism of Europe—many small countries with different languages and institutions make continent-wide scaling difficult.

Economies of scale drive progress. But they’re also dangerous. Here’s why:

If I have twice your resources, I’ll make more than twice your progress (due to economies of scale). Next year, I’ll have (let’s say) 2 times your resources. The year after, 4 times. Eventually, one actor controls everything.

With proportional growth, if you start with 10 units and I start with 12 units, and we both grow 10% per year, we’ll maintain roughly the same ratio forever. Small initial differences stay small.

With super-exponential growth (economies of scale), the larger actor grows faster. A 20% initial advantage becomes 50%, then doubles, then becomes overwhelming. Small initial differences become decisive.

Historically, two forces counterbalanced economies of scale:

1. Diseconomies of scale: Large institutions face internal conflicts, communication costs, geographic distance costs, bureaucratic inefficiency.

2. Diffusion: People move between companies and countries, taking ideas with them. Poor countries trade with rich countries and catch up. Industrial espionage happens. Innovations get reverse-engineered. You can use one social network to advertise another.

Think of it as a race where the cheetah and turtle are competing. Diseconomies of scale slow the cheetah down. Diffusion pulls the turtle forward, closer to the cheetah.

Recently, several forces are disrupting this balance:

• Rapid technological progress: Makes the super-exponential curves much steeper than before
• Automation: Enables global-scale work with few people, reducing coordination costs
• Proprietary software/hardware: Historically, giving someone your product meant they could inspect and reverse-engineer it. Now we can distribute functionality without distributing control.

Basically: economies of scale are increasing, while diffusion of control (not just ideas) is decreasing.

This gives us our central question: How do we maintain a flourishing civilization with rapid progress in the 21st century without extreme power concentration?

The answer: mandate more diffusion.

The next question is obvious: what does mandated diffusion actually look like in practice?

Forcing Diffusion #

What does “mandate more diffusion” actually mean? Let me start with policy examples:

• EU standardization mandates (like requiring USB-C charging): Make it harder to build proprietary ecosystems that don’t interoperate
• Bans on non-compete agreements: I support these because they force “tacit knowledge” inside companies to be partially open-source—when employees leave, they can apply learned skills elsewhere. Non-disclosure agreements limit this but are fortunately quite porous in practice.
• Copyleft licenses (like the GPL): Require any software built on copylefted code to itself be open-source and copylefted

We could develop other mechanisms inspired by the EU’s carbon border adjustment. Imagine charging a tax on products proportional to how proprietary they are—if you share the technology with us or open-source it, the tax drops to zero. Or we could revive Harberger taxes on intellectual property, where rights-holders must continuously name a price they’d sell for, with anyone able to purchase at that price.

But there’s also a more chaotic strategy we should embrace: adversarial interoperability.

As technology writer Cory Doctorow explains, adversarial interoperability means creating products or services that plug into existing platforms without permission. Think of third-party printer ink, alternative app stores, or independent repair shops using compatible parts.

Concrete examples:

• Alternative clients for social media: See the platform’s content and post your own, but with completely different filtering and algorithms you control
• Browser extensions: Like ad blockers, but filtering AI-generated content, engagement bait, or other unwanted elements
• Decentralized exchanges: Censorship-resistant bridges between fiat and crypto, mitigating the chokepoint features of centralized financial systems

Much of web2’s value capture happens at the user interface level. If you can make alternative interfaces that remain interoperable with the platform and its users, you stay part of the network while opting out of its value extraction.

Sci-Hub provides an interesting case study. By providing unauthorized access to academic papers, it has arguably improved fairness and open access in science—forcing diffusion where the formal system created artificial scarcity.

A third strategy comes from ideas on “Plurality”—ways to enable cooperation across difference without requiring unified goals. These tools would let open-source communities, collections of countries, and other non-unitary groups share economies of scale and remain competitive with centralized behemoths, while maintaining their independence.

Piketty documented how capital compounds faster than labor income (r > g), concentrating wealth even when income remains distributed. His solution was a global wealth tax. Ours differs: instead of redistributing wealth after it concentrates, we target the mechanisms generating concentration.

I’d argue that this approach is better because it more directly targets the dangerous thing (extreme growth coupled with exclusion), and if done well it can even be efficiency-increasing. It also has the advantage that it does not limit itself to targeting one type of power. While a global wealth tax may prevent concentration of power among billionaires, it would do nothing against powerful dictatorial governments or other transnational entities, and it would perhaps leave us even more defenseless against them.

A global decentralized strategy of forcing technological diffusion—telling people “either you grow with us, and share access to your secret sauce and your network on a reasonable schedule, or you grow entirely alone, and we shut you out”—would address power concentration in a different way.

Making Multipolarity Safe #

You might worry that pluralism creates vulnerability. What if advancing technology enables single actors to cause catastrophic harm? Isn’t the solution to concentrate power more for security?

This is where defensive acceleration (d/acc) becomes crucial. The strategy involves building defensive technology that keeps pace with offensive capabilities, in ways that are open and available to everyone. When defense is democratized, we reduce the security-driven pressure to concentrate power.

Consider end-to-end encryption. When widely available, it prevents both governments and corporations from monopolizing secure communication. Or vaccine technology: when mRNA platforms are open-source, no single nation can monopolize pandemic response. These technologies don’t prevent innovation—they prevent the concentration of control over innovation’s benefits.

Think of it as a complementary approach: forced diffusion prevents dangerous concentration; d/acc makes that diffusion safe rather than catastrophically risky.

Toward a Pluralist Ethic #

Let me close with a moral framework:

Slave morality says: You are not allowed to be powerful.

Master morality says: You are commanded to be powerful.

A pluralist synthesis might say: You are not allowed to be hegemonic, but you are encouraged to be impactful and to empower others.

This is another formulation of the “power to” versus “power over” distinction that has existed for centuries.

You can have power to without power over through high diffusion toward the outside world. Or you can build things that minimize their potential as levers of power.

In Ethereum (a blockchain platform where participants “stake” cryptocurrency to help secure the network), the decentralized staking pool Lido provides a useful example. Despite controlling roughly 24% of staked ETH, people fear it much less than they would almost any other entity with 24% control. Why? Because Lido isn’t a single actor—it’s internally decentralized with several dozen operators and a “dual governance” design giving staked ETH holders veto power over decisions. Lido deserves credit for this effort. Simultaneously, the Ethereum community has insisted that even with these safeguards, Lido shouldn’t control all of Ethereum’s stake—and it remains far from that.

More projects should think explicitly about their “decentralization model” alongside their business model—how they avoid concentrating power and the risks associated with such power.

Sometimes decentralization is straightforward. Few worry about English language dominance despite its global reach, because anyone can use English without asking permission and no one controls who gets to speak it. Open protocols like TCP, IP, and HTTP similarly distribute power—they’re essential infrastructure, but no single entity can leverage them as tools of control. These systems succeed precisely because they separate functionality from authority.

Other cases are harder because the use case demands intentionality and capacity to act. Consider content moderation on social platforms. A completely decentralized approach where no one can remove anything fails—you get harassment, illegal content, spam that makes the platform unusable. But heavy-handed centralized moderation creates its own dangers: arbitrary censorship, political manipulation, suppression of dissent. Or consider emergency response systems: they need coordination and authority to function, but concentrated authority invites abuse.

The challenge isn’t choosing between centralization and decentralization as absolutes. It’s designing systems that maintain enough coordination to achieve their purpose while preventing that coordination from becoming a weapon. This might mean:

• Separating different types of power: A platform might let a thousand moderators make independent decisions (distributing day-to-day power) while reserving only rare, extreme cases for central authority
• Building in exit rights: Users can leave with their data and social graph intact, limiting the platform’s ability to trap them
• Creating internal checks: Multiple independent parties must agree before consequential actions occur
• Accepting incompleteness: Choosing to solve 80% of problems in a distributed way rather than 100% of problems through centralized control

None of these solutions are perfect. Each involves tradeoffs. A distributed moderation system will be less consistent than a centralized one. Exit rights reduce network effects that make platforms valuable in the first place. Internal checks slow decision-making. Accepting incompleteness means tolerating problems you technically have the power to solve.

But these tradeoffs are features, not bugs. The friction they create—the inability to act swiftly and decisively in all cases—is precisely what prevents power from concentrating to dangerous levels. Figuring out how to get flexibility’s upsides without power concentration’s downsides will remain an important challenge. There’s no formula. Each domain requires careful thought about what kind of coordination it genuinely needs and what safeguards prevent that coordination from becoming domination.

But it’s a challenge we must address. The alternative—accepting inevitable concentration as technology advances—means accepting that the future belongs to a few dominant actors. The tools already exist: standardization mandates, adversarial interoperability, copyleft licensing, defensive acceleration technologies. What remains is the will to deploy them systematically, and the wisdom to design new institutions that embody pluralist principles from the start.

A future of distributed capability and genuine choice is worth fighting for. Not because it’s easier or more efficient, but because it’s the only way to preserve both progress and freedom as we navigate an era of unprecedented technological power.

Credit #

Original article https://vitalik.eth.limo/general/2025/12/30/balance_of_power.html