Path Dependence

In 1873, Christopher Latham Sholes patented the first commercially successful typewriter. The machine worked, but imperfectly, its metal type bars frequently jammed when commonly paired letters were struck in quick succession, so to mitigate this, Sholes rearranged the keyboard such that frequently used letter combinations were more physically separated. There, the goal was not efficiency, but reliability under a specific mechanical constraint. The resulting layout became known as QWERTY, which you can see in every keyboard today as the default configuration. But why is that?

By the 1890s, typewriter manufacturers had solved the jamming problem, so the mechanical constraint that justified QWERTY's original design no longer existed, but by then, millions of people had already learned to type on QWERTY keyboards. All the training materials, typing courses, and professional typists assumed the layout, and manufacturers kept producing it because that's what typists expected, and typists kept learning it because that's what manufacturers produced.

So, as much as the original problem disappeared, the suboptimal solution for the extinct problem remained, because switching costs had become an inconvenience. Better alternative layouts, like Dvorak, Colemak, and others, which demonstrably improved typing speed and reduced finger strain, never gained more than niche adoption. And so QWERTY won not by being best, but by being first.

This is path dependence: the tendency of systems to converge toward outcomes determined by early contingencies, because later changes become prohibitively expensive, inconvenient, politically impossible, or socially illegible.

The Mechanism

Path dependence emerges from three related phenomena:

Increasing returns to adoption: The more people use a standard, the more valuable it becomes. Not because the standard itself improves, but because compatibility, expertise, and infrastructure all accumulate around it. This creates a gravitational pull toward existing choices, even suboptimal ones.

Switching costs: Changing course requires simultaneously updating many interdependent components. These costs grow over time as more systems integrate with the original choice. Eventually, the cost of switching exceeds the benefit of improvement.

Lock-in: Once sufficient investment has been made in one path, alternatives become inaccessible even if they're superior. The system is locked into its trajectory, not by physical constraint, but by the weight of accumulated decisions.

These forces compound and early adopters choose based on immediate constraints, while later adopters choose based on compatibility with early adopters, and even though the original constraints may become irrelevant, the choice persists because too many other choices now depend on it.

Scale and Scope

Path dependence operates at every level of organization, from technical standards to political institutions.

Technical standards: Railroad gauge, electrical voltage, video encoding formats, programming languages. In each case, early choices made under specific constraints become permanent infrastructure, even after those constraints disappear. The United States uses 120V electrical systems because that's what Edison's first power stations delivered. Europe uses 230V because that's what later, more efficient generators provided. Both work, neither is clearly superior for modern use, and the cost of unifying them globally is now astronomical.

Operating systems and software: Microsoft Windows dominated not because it was the best operating system in the 1990s, by most technical measures, it wasn't, but because it achieved critical mass first. Software developers wrote for Windows because that's where the users were, and users bought Windows PCs because that's where the software was. The feedback loop became self-reinforcing, and while better alternatives existed and continue to exist, the installed base of Windows expertise, software, and compatibility makes migration nearly impossible for most organizations.

Urban infrastructure: Cities build around their dominant transportation networks. Once roads, sewers, and power grids are laid, future development must accommodate them. Many European cities fixed their street widths centuries before wheeled traffic existed at scale. Streets were designed for walking, horses, drainage, and defense, not speed or throughput. When automobiles arrived, the underlying geometry could not change: buildings were built in stone long before cars existed, property lines had already hardened into law, and utilities were buried and layered over centuries. By the time automobiles arrived, demolition was prohibitively expensive, so instead of rebuilding cities for cars, cars were redesigned for cities (smaller vehicles), which were not planning preferences but mechanical necessities. So, the original choice persists not because it is optimal for modern transport, but because rewriting the physical substrate would require destroying everything built on top of it.

Financial infrastructure: The SWIFT network for international payments was designed in the 1970s for a world of correspondent banking, where transfers moved through multiple intermediaries over several days. The system remains the default today, even though the original technical constraint no longer exists.

Legal and regulatory systems: Laws accumulate over time, each new rule assuming the existence of previous ones. Modern tax codes are not designed systems but archaeological strata, each layer preserving the concerns and compromises of its era. Simplification becomes nearly impossible because every deduction, credit, and exemption has a constituency that depends on it. The system doesn't optimize for clarity or efficiency, but for compatibility with its own history.

Institutional Path Dependence

Political systems are particularly vulnerable to path dependence because they encode power relationships. Once established, institutions create their own constituencies: bureaucracies that administer them, industries that benefit from them, populations that adapt their behavior to them.

The modern administrative state emerged in response to industrial-era coordination problems: railroad regulation, food safety, labor standards. These were real problems requiring institutional solutions. But the bureaucratic structures built to solve 1920s problems persist into an era where they often create more friction than they resolve.

This is not a mere moral failing, it is a mechanical failure, as each regulation assumes prior regulations, and each agency coordinates with other agencies, and so each funding stream depends on previous budgets. Changing one piece requires changing everything it touches. Over time, the cost of reform exceeds the cost of maintaining a suboptimal status quo, and the system continues on momentum alone.

This is why institutional decay often looks like increasing rigidity: the system is not becoming more authoritarian out of ideology, but out of path dependence: the only way to maintain control over a world that no longer fits the original framework is to force it back into alignment. Regulation piles on regulation, not because each new rule is needed, but because the system cannot adapt and can only constrain.

Why Escaping Path Dependence Is Hard

The usual prescription for path dependence is "make better initial choices." But this misunderstands the problem. Early choices are made under uncertainty, with incomplete information, and optimized for constraints that later prove irrelevant. So the issue is not bad decision-making at the start but the impossibility of correction later.

How to escape path dependence?

Coordinated migration: Everyone switches simultaneously, so no individual bears disproportionate switching costs. This is nearly impossible at scale. The European Union spent decades and billions of euros coordinating a currency transition, and even that only partially succeeded because opt-outs were allowed. Coordinating a technological or institutional transition without an authority figure is harder.

Parallel systems: Build the new system alongside the old, let people opt into it gradually, and let the old system atrophy naturally. This works when switching costs can be made low enough that individuals can migrate without losing accumulated value, but most path-dependent systems have high exit costs by design.

Crisis: Disasters create temporary windows where the costs of maintaining the old system spike above the costs of switching. Wars, economic collapses, and technological disruptions force resets, but crisis-driven transitions are chaotic, often replacing one set of path dependencies with another, equally arbitrary set.

Degrees of Freedom and Path Dependence

Path dependence is a dimensional collapse. Early in a system's development, many trajectories are possible. Choices are made, some deliberate, others contingent, and each choice narrows the possibility space. Eventually, the system has so many interdependent components that only one viable path remains.

This is why systems with higher degrees of freedom are more resilient. If components can vary independently (if you can change one part without changing everything it touches) then path dependence weakens. Then, the system can adapt piecewise rather than requiring total reconstruction.

This is also why non-delegation matters. Delegation creates dependency chains: you depend on the intermediary, the intermediary depends on its infrastructure, the infrastructure depends on regulatory frameworks. Each link in the chain adds path dependence. When the system fails, you cannot simply opt out, you're locked in by everything that depends on it.

Self-custody, encryption, and portable identity are not just defensive strategies, they're also ways of reducing path dependence. If you control your own assets, you can switch systems without permission. If your communication is encrypted end-to-end, you're not locked into a particular platform. If your identity and reputation are portable, you can exit one institution without losing your accumulated social capital.

Network Effects vs. Path Dependence

Network effects and path dependence are often confused because both involve increasing returns. But they differ in an important way:

Network effects create value. A phone network becomes more valuable as more people join, not because the technology improves, but because you can call more people. This is genuine value creation from coordination.

Path dependence creates lock-in. QWERTY doesn't become more valuable as more people learn it. It becomes harder to escape, but the layout itself remains suboptimal. The returns are not from value creation but from switching cost amplification.

The difference matters because it determines whether lock-in is desirable. Network effects justify coordination around a standard, everyone driving on the same side of the road creates real value. Path dependence does not: when a standard is suboptimal and cannot improve, rising switching costs trap users in a system that no longer serves them.

Modern platforms exploit this ambiguity. They claim network effects ("our platform is valuable because everyone is here") to justify natural monopolies, but often what they've actually created is path dependence ("you can't leave because your data, contacts, and history are locked in"). Real network effects create value from coordination. Artificial lock-in extracts value from dependency.

This is why portability matters, if your social graph is portable between platforms, then platforms compete on service quality rather than holding your relationships hostage. The network effect remains: you want to be where your friends are, but the path dependence disappears. You're no longer locked into a failing platform because your accumulated social capital can move with you.

What Makes Systems Escapable

Not all path dependence is equal, some systems are intrinsically harder to escape than others.

Physical infrastructure has high escape costs because it's embedded in space. Cities built around highways cannot easily transition to rail without demolishing and rebuilding. But digital systems can coexist. You can use multiple messaging apps, multiple payment systems, multiple identity providers. The switching cost is not physical demolition but attention and habit.

Legal frameworks have high escape costs because they're tied to territory and enforced through monopoly on violence. You cannot easily opt out of national regulations without leaving the country, and leaving the country means abandoning property, relationships, and professional credentials built under that system.

Protocols have low escape costs when designed for interoperability. Email is path-dependent in that everyone uses it, but you're not locked into a specific email provider. You can switch providers without losing your contacts or your ability to communicate with others. The protocol is standardized; the implementation is competitive. This is the structure that minimizes path dependence while preserving network effects.

The lesson is that systems should be designed with escape hatches:

• Optionality: Multiple coexisting systems rather than one mandatory monopoly.
• Interoperability: Standards that allow different implementations to communicate.
• Portability: The ability to migrate your data, identity, and relationships between providers.

These are not luxuries, they are design choices and mechanisms that prevent temporary solutions from becoming permanent prisons.

The Enlightenment and Path Dependence

The original Enlightenment broke a massive path dependence: the fusion of religious and political authority. For over a millennium, European governance was built on divine-right monarchy: institutions, laws, social hierarchies, and economic relationships all assumed this structure, and questioning it was not just politically dangerous but socially unintelligible, as most people didn’t have the language or concepts to imagine a different system.

The printing press spread ideas, but also broke the lock-in, as once information could flow outside church and state control, the path dependence weakened. Alternative governance models became thinkable, then possible, then implemented. But the transition took centuries and was chaotic, yet it succeeded because the switching costs dropped enough that parallel systems could emerge.

We're in a similar moment now. The institutions built in the Enlightenment era are path-dependent on nation-state monopolies, centralized intermediaries, and geographic governance, and these made sense in a world of slow communication, limited coordination, and physical constraint, but they make less sense in a world of instant global communication, cryptographic identity, and network-based coordination.

But path dependence is strong, as the entire architecture of modern life: citizenship, banking, professional credentials, social ties, and tradition, they all assume nation-states and intermediaries. So, you cannot simply opt out without losing everything you've built.

What's needed are systems designed explicitly to reduce lock-in:

• Platforms built on open protocols rather than proprietary lock-in.
• Identity and reputation systems that are portable across institutions and jurisdictions.
• Governance structures you can opt into and out of without losing your accumulated value.
• Financial systems where you control your own assets rather than depending on institutions that can freeze or confiscate them.

These are not utopian visions, but engineering requirements for systems that can adapt without requiring coordinated global transitions or waiting for crisis.

The Future Is Not Determined

Path dependence is not destiny, past dependencies are inherited, but they can be changed by the force of will and coordination. Future path-dependencies can be avoided by smart and ethical design choices we make today.

Of course, the question whether the cost of change is worth the benefit of escape remains. For QWERTY, probably not, the efficiency gains from Dvorak are real but small, and the coordination costs are enormous. For institutional frameworks designed for a pre-internet world, the answer is different. The misalignment between existing institutions and the world they're trying to govern is growing, not shrinking. The cost of staying on the current path is increasing faster than the cost of switching.

This is why building alternatives matters. Not to replace the existing system overnight, coordinated migration at global scale is impossible, but to create parallel paths that people can opt into when the switching costs drop below the staying costs.

And switching costs are dropping: cryptography makes self-custody cheaper, open protocols make portability easier, digital tools make coordination across distance nearly free, the infrastructure for low-path-dependence systems is being built in real time.

The legacy systems will not disappear overnight, path dependence ensures they'll persist long after they stop serving their original purpose, but they no longer need to be the only option, and once people have genuine alternatives, and once exit becomes cheap enough that institutional failure no longer traps society, the path dependence loosens.

The future should not be locked by the past, it should remain free to change, and keeping it that way is a design choice. So the question is whether we accept permanent lock-in, or whether we build systems that can adapt to change.