4. Blockchains

Whereas most technologies tend to automate workers on the periphery doing menial tasks, blockchains automate away the center. Instead of putting the taxi driver out of a job, blockchain puts Uber out of a job and lets the taxi drivers work with the customer directly.1

—Vitalik Buterin

Why Computers Are Special: The Platform-App Feedback Loop

In the 1989 sequel to Back to the Future, the main character travels from 1989 to 2015. Flying cars zip across skyways, but people still use phone booths. Smartphones don’t exist.

This is common in pre-internet science fiction: almost no stories foresee the fantastic success of computers and the internet. Why do storytellers so consistently get this wrong? Why do portable internet-connected supercomputers arrive, in reality, before flying cars? Why do computers and the internet improve so much faster than everything else?

The explanation is partly technological. The laws of physics allow us to shrink transistors, the smallest unit of computing machinery, and therefore pack more computing power into smaller volumes. The rate that describes this process is known as Moore’s law,2 named after Gordon Moore, a founder of the chip company Intel. Moore’s law states that the number of transistors that can fit on chips roughly doubles every two years. History proves the rule: a modern iPhone has more than 15 billion transistors, compared with a 1993 desktop PC, which has only about 3.5 million. Very few technologies experience more than a thousand times improvements like this. The physical constraints in other engineering fields are harder to overcome.

The remainder of the explanation is that an economic phenomenon is also at work: a reciprocal relationship between applications, or apps, and the platforms that underpin them. The iPhone today contains many more transistors and other components than the original iPhone, but it also has many more apps. Those apps are far more useful and advanced than the earliest available apps. New apps help sell more phones, which leads to increased reinvestment in phones and, in turn, back into apps. This is the platform-app feedback loop. Platforms, like the iPhone, enable new applications. New apps make platforms more valuable. The back-and-forth creates a positive feedback loop of compounding improvements.

Technological advances and platform-app feedback loops make computers faster, smaller, cheaper, and more feature rich. These forces recur throughout the history of computing. Entrepreneurs created word processors, graphic design programs, and spreadsheets for PCs. Developers put search engines, e-commerce, and social networking on the internet. Builders brought messaging, photo sharing, and on-demand delivery services to mobile phones. In each case, investment alternated between platforms and apps, creating rapid, multiyear growth.

The platform-app feedback loop applies to both community-owned and corporate-owned platforms. Protocols like the web and email benefited from the feedback loop, as did the open-source operating system Linux. On the corporate side, Microsoft benefited from similar loops in the 1990s as developers built apps for Windows computers. App developers are doing the same for Apple’s and Google’s mobile operating systems today.

Sometimes multiple trends converge and amplify one another, like constructive interference between overlapping waves. Social networks were a killer app for mobile phones; they helped make the devices popular. Meanwhile, cloud computing offered flexible infrastructure that startups could use to quickly scale up their apps, such as social networks, so they could support billions of users. Mobile phones made everything accessible and affordable. Together these trends combined to bring us the magical handheld supercomputers that are ubiquitous today yet which most science fiction failed to imagine.

Major computing cycles3 typically come along every ten to fifteen years. Mainframes dominated in the 1950s and 1960s. Minicomputers reigned in the 1970s. Then came PCs in the 1980s. The internet took off in the 1990s. And, most recently, mobile phones became ubiquitous starting in 2007, when the iPhone launched. There is no rule that says this pattern needs to continue, but there is a logic to it: Moore’s law suggests that it takes roughly ten to fifteen years to improve computing power by a hundred times, and it also takes about that long for many research projects to mature. If the ten-to-fifteen-year pattern continues, then we’re in the midst of another cycle.

Multiple trends will drive the next cycle. Artificial intelligence is one of them. The sophistication of AI models appears to be growing at an exponential rate, a function of the number of parameters in their underlying neural networks. The pace of improvement suggests that future models will be much more powerful than already impressive ones out in the market. Another breakout will be new hardware devices like self-driving cars and virtual reality headsets. These devices are advancing rapidly thanks to improvements in sensors, processors, and other components. Big companies like Apple, Meta, and Google are making significant investments in these areas.4 These are the consensus bets—the conventional picks—for what’s next in computing. Almost everyone agrees on their significance.

Blockchains are different. They’re a non-consensus bet. While plenty of people recognize their potential—including me—much of the establishment disregards them. In fact, a prevailing view in the tech industry assumes that the only vectors of technological improvement that matter are the ones incumbents are already focused on: bigger databases, faster processors, larger neural networks, smaller devices. The view is myopic. It puts too much weight on technologies originating from established institutions while ignoring ones that come from elsewhere, from the long tail of outside developers.

Two Paths to Adoption: “Inside Out” versus “Outside In”

New technologies follow one of two paths:5 “inside out” or “outside in.” Inside-out technologies start inside Big Tech. They are the more obvious of the two, coming out fully baked from inside established institutions and getting better at the rate at which corporate employees, staff researchers, and others on the payroll improve them. They tend to need significant capital and formal training, which raises barriers to entry.

Most people recognize the value of inside-out technologies even before they exist. It’s easy to imagine that internet-connected pocket-sized supercomputers might be popular, as Apple proved with the iPhone. It’s also easy to imagine that people might want machines that can learn to act intelligently and do all sorts of tasks, as university and corporate research labs showed with AI. Incumbents pursue these technologies because they see obvious potential.

Outside-in technologies arrive, in contrast, on the fringes. Hobbyists, enthusiasts, open-source developers, and startup founders hatch them outside the mainstream. The work usually involves less capital and formal training, which helps level the playing field with insiders. A lower bar also causes insiders to take these technologies and their proponents less seriously.

Outside-in technologies are much harder to see coming, and they’re routinely underestimated. Their builders work out of garages, basements, dorm rooms, and other unconventional spaces, outside official hours. They tinker after work, during breaks, and on weekends. They’re motivated by a distinct philosophy and culture that can look strange to the outside world. Other people don’t get them. The outsiders launch products half-baked, without clear uses. Most onlookers dismiss their technologies as toylike, weird, unserious, expensive, or even dangerous.

Software is an art form, as you’ll remember: just as you wouldn’t expect all great novels or paintings to come from people at established institutions, so you shouldn’t expect all great software to come from them either.

Who are the outsiders? Picture a counterculture-loving, twentysomething Steve Jobs attending the Homebrew Computer Club,6 a den of microcomputer-obsessed geeks that hosted monthly meetups in California in the 1970s. Picture Linus Torvalds as a student7 at the University of Helsinki in 1991, coding up a personal project that would become his namesake Linux operating system. Or picture Larry Page and Sergey Brin dropping out of Stanford and moving into a Menlo Park garage in 1998 to turn their web-link-cataloging project,8 BackRub, into Google.

The value of outside-in technologies is often unclear before their invention—and may remain so for many years afterward. The web started out half-baked when Tim Berners-Lee concocted it at a Swiss physics lab in 1989, but it grew exponentially as it attracted developers and entrepreneurs who saw its potential. As my technologist friend Sep Kamvar jokes, if you asked people at that time what they needed to make their life better, they likely wouldn’t have said a decentralized network of information nodes that are linked using hypertext. And yet, in retrospect, that’s exactly what they needed.

Hobbies fuel future industries. Open-source software started out as a niche anti-copyright movement before going mainstream. Social media began as a pastime among idealistic blogging enthusiasts before the world embraced the idea. That T-shirt- and flip-flop-wearing hobbyists spawn large industries may seem like an amusing eccentricity of the tech industry, but hobbies are important for a reason. Businesspeople vote with their dollars: they are mostly trying to create near-term financial returns. Engineers vote with their time: they are mostly trying to invent interesting new things.

Hobbies are what the smartest people spend time on when they aren’t constrained by near-term financial goals. I like to say that what the smartest people do on the weekends is what everyone else will do during the week in ten years.

These two modes of tech development—inside out and outside in—are often mutually reinforcing, as you can see in the combination of trends that powered the growth of computing over the last decade. As mentioned earlier, mobile, an inside-out technology pioneered by Apple, Google, and others, brought computers to billions of people. Social, an outside-in technology cobbled together by hackers like the Harvard dropout Mark Zuckerberg, drove usage and monetization. Cloud, another inside-out technology, spearheaded by Amazon, allowed back-end web services to scale.9 The two modes can unleash powerful forces when they line up, like nuclei fusing.

Blockchains are a classic outside-in technology. Most incumbent technology companies are ignoring blockchains, and some of their employees even dismiss and ridicule them. Many people neglect blockchains because they don’t even think of them as computers. Startup founders and independent groups of open-source developers are driving the technology’s development. In this way, industry outsiders are leading this new computing movement, just as they did for the early protocol networks like the web and open-source software like Linux.

Blockchains Are a New Kind of Computer

In a 2008 paper, Satoshi Nakamoto, a pseudonymous inventor or team of inventors (the identity remains unknown), introduced the world’s first blockchain.10 Although he didn’t call his invention a blockchain at the time—he used the terms “block” and “chain” separately—the community that formed around his ideas would eventually stick the two words together. His paper described a new kind of digital money, Bitcoin, as “an electronic payment system based on cryptographic proof instead of trust, allowing any two willing parties to transact directly with each other without the need for a trusted third party.” To remove the trusted third party, Nakamoto needed a way for the system to run computations independently. To this end, he described a new kind of computer, a blockchain.

Computers are an abstraction,11 defined by what they do rather than what they’re made of. Originally, “computers” referred to people who perform calculations. In the nineteenth and twentieth centuries, the word started referring to machines that can calculate. Alan Turing, a British mathematician, set a more rigorous foundation in a famous 1936 paper12 on mathematical logic in which he investigated the nature and limits of algorithms. In it, Turing defined what computer scientists would today call a state machine, and what everyone else would simply call a computer.

A state machine consists of two parts: (1) a place to store information and (2) a means to modify that information. Information stored is called state, equivalent to computer memory. Sets of instructions, called programs, specify how to take one state, an input, and produce a new state, an output. I like to describe computing through the lens of language, since more people can read and write than can program. Imagine nouns represent state or memory: things that can be manipulated. Verbs represent code or programs: actions that do the manipulating. As you’ll hear me repeat, anything you can dream up, you can code, which is why I compare coding to creative activities like fiction writing. Computers are extremely versatile in this way.

A state machine is the purest way to think about a computer. Nakamoto’s blockchain is not a physical computer, like a PC, laptop, phone, or server. It is a virtual computer—meaning it is a computer in function, not in conventional physical embodiment. Blockchains are a software abstraction that overlay on top of physical devices. They’re state machines. Just as the meaning of “computers” once shifted from people to machines, so too has the term since encompassed not just hardware but software as well.

Software-based computers, or “virtual machines,” have been around since IBM developed the first one13 in the late 1960s and released it in the early 1970s. The IT giant VMware later made the tech popular in the late 1990s. Today, anyone can run virtual machines by downloading so-called hypervisor software on a PC. Companies commonly use virtual machines to streamline the management of corporate data centers, and they’re key to the operations of cloud service providers. Blockchains extend this model of software-based computing to a new context. Computers can be built in many different ways; they are defined by their functional properties, not by what they look like.

How Blockchains Work

Blockchains are by design resilient14 to manipulation. They are built on top of a network of physical computers that anyone can join but that is extremely difficult for any one entity to control. These physical computers maintain the state of the virtual computer and control its transitions to new states. In Bitcoin these physical computers are called miners, but the more common term today is “validators” since what they’re really doing is validating state transitions.

If state transitions sound too abstract, an analogy may help. Think of Bitcoin as a fancy spreadsheet, or ledger, with two columns. (It’s more complex than that, but bear with me.) Each row of the first column has a unique address. Each row of the second column contains the number of bitcoins held at that address. State transitions update the rows in the second column to reflect all the transfers of bitcoin executed in the latest batch. That’s the gist, really.

If anyone can join the network, how does the virtual computer arrive at a single source of truth about its state? Phrased differently, if the spreadsheet is open to all, how can anyone trust the numbers that appear in its rows? The answer: through mathematical guarantees involving cryptography (the science of secure communication) and game theory (the study of strategic decision-making).

Here’s how a proposed state becomes the next state of the computer. During each state transition, the validators run a process to reach a consensus on the next state. First, the validators do as their name says: they validate, making sure every transaction comes with an appropriate digital signature. The network then randomly selects one validator to bundle together qualifying transactions to create the next state. Other validators check to make sure the new state is valid, that all the bundled transactions are also still valid, and that the computer’s core commitments have been upheld (for example, in the case of Bitcoin, that there will never be more than twenty-one million bitcoins). Validators effectively cast their vote for a new state by building on it as the transition to the next state starts.

The process is designed to ensure everyone is working off the same, valid version of history—to reach consensus. If a validator (or subset of validators) tries to cheat, the other validators have every opportunity to catch it lying and outvote it. The rules of the process are set up in such a way that you would generally need a majority of validators to collude for it not to work.

In our simplified example above, the new master copy spreadsheet is the one proposed by the winning validator. Of course, in reality, there is no spreadsheet. There are only state transitions—the essence of computation. Each state transition is called a block, and the blocks are chained together so that anyone can verify the complete history of the computer by examining the blocks. Hence the name blockchain.

State transitions can contain more than just numbers representing simple account balances. They can hold whole sets of nested computer programs. Bitcoin comes with a programming language, called Bitcoin Script, that software developers can use to create programs that modify the transitions between states. This programming language is, however, limited by design. It mostly enables people to send funds between accounts or to create accounts controlled by multiple users. Newer blockchains like Ethereum, the first general-purpose blockchain, which made its debut in 2015,15 allow developers to program in much more expressive programming languages.

The addition of advanced programming languages to blockchains is a major breakthrough. It’s analogous to Apple’s introducing an app store to the iPhone (except where mobile app stores are curated, blockchains are open and permissionless). Any developer in the world can write and run apps, ranging from marketplaces to metaverses, on blockchains like Ethereum. This is a very powerful property that makes blockchains far more expressive and versatile than an accountant’s notebook. This is why it’s wrong to think of blockchains as mere ledgers for tabulating numbers. Blockchains are not databases; they’re full-fledged computers.

Running apps on computers takes resources, though. Both application-specific blockchains like Bitcoin and general-purpose ones like Ethereum need people to pay for the computing power that validates state transitions, and so they must give people a reason to invest in these networks. To that end, Nakamoto introduced a clever twist: the system’s digital currency—in Bitcoin’s case, bitcoin—would itself be the source of funding for the computers that power it. Other blockchains have since copied the design.

Every blockchain has its own set of internal incentives to get people to participate. In most systems, every new block, or state transition, gives away a small bounty to a lucky validator. (“Validator” can refer to computers that vote on state transitions or to the person or group operating those computers.) The validators that behave honestly—the ones that faithfully verify digital signatures and propose only valid changes to the blockchain—get rewarded. This financial incentive encourages the validators to continue supporting the network and behaving honestly. (Money also flows into blockchains through fees charged to users; more on how this works, and how tokens are valued, in the chapter “Tokenomics.”)

Blockchains are permissionless, so anyone with an internet connection can participate. Nakamoto designed the original blockchain, Bitcoin, this way because he believed that existing financial systems were elitist, favoring privileged intermediaries, like banks. Instead, he wanted to put everyone on an equal footing. Requiring an application or screening process would introduce new privileged intermediaries, re-creating the problems he associated with the existing system. But this design had a complication: if any computer could vote, then spam and bad actors could easily overwhelm the network.

Nakamoto’s solution was to charge a “fee” to participate. To vote on the next machine state, a miner would need to perform computational work, which costs energy, and submit proof that it did that work. This system—aptly called proof of work—enabled open, permissionless voting while also filtering out spam and other nefarious schemes. Other blockchains, like Ethereum, have adopted another system, called proof of stake (PoS). Instead of requiring validators to spend money on electricity, proof of stake requires them to “stake” collateral, meaning to put money at risk in escrow. If the validators behave honestly, they earn monetary rewards. If they get caught lying—by voting for contradictory state transitions or proposing multiple conflicting state transitions simultaneously, for example—their collateral gets “slashed,” or confiscated.

One of the main criticisms of Bitcoin is its excessive energy consumption, which could harm the environment. While clean energy sources, such as excess renewable energy from dams and wind turbines, can mitigate the environmental effects of proof of work, a better approach can be to replace proof of work altogether with less energy-intensive systems,16 like proof of stake, which eliminate environmental objections to blockchains.

Proof of stake is as secure as proof of work, if not more so, while also being cheaper, faster, and far more energy efficient. Ethereum finished transitioning from proof of work to proof of stake in the fall of 2022, and the results have been dramatic. The next chart shows Ethereum proof-of-stake energy consumption compared with other popular systems.

Many blockchains mentioned in this book, with the notable exception of Bitcoin, use proof of stake. In the future, I expect proof of stake will power the most popular blockchains. Concerns over energy consumption shouldn’t hold anyone back from using this powerful new technology.

Neither should the popular misconception that blockchains enable secrecy and anonymity. “Crypto,” a word connoting statecraft and intrigue, literally means “encoded” or “hidden.” Confusion over how the word is used to describe the industry leads people to believe, mistakenly, that blockchains hide information, and therefore that they’re perfectly suited for illegal conduct. This inaccuracy is common, for example, in TV and movies that depict criminals using cryptocurrency to secretly transfer money. It’s also dead wrong.

In fact, everything that happens on popular blockchains like Bitcoin and Ethereum is public and traceable. As with email, you can sign up using a fake identity, but there are companies that specialize in de-anonymizing, and it’s straightforward for law enforcement18 to do so. Blockchains are so public by default that their innate transparency could actually hinder adoption. This may seem counterintuitive, given the erroneous public perception of crypto as a black box, but it’s true. People may be reluctant to use blockchains for certain activities if they fear doing so will expose sensitive information, such as salaries, medical bills, or invoices. Some projects are working to solve this problem by giving users the option to make transactions private. The most advanced projects employ cutting-edge cryptography—especially innovations like “zero knowledge proofs”19—which enables auditing of encrypted data that can mitigate the risk of illegal activities20 and satisfy the needs of regulatory compliance.

Blockchains are “crypto” not because they enable anonymity (they don’t) but because they’re based on a mathematical breakthrough from the 1970s21 called public key cryptography. The main thing to know about public key cryptography is that it lets multiple parties who have never before communicated perform cryptographic operations with one another. The two most common operations are (1) encryption, which encodes information so it can only be decoded by the intended recipient, and (2) authentication, which lets a person or computer sign information, proving it’s authentic and actually came from that source. When people describe blockchains as crypto, they mean it in the latter sense of “authenticated,” not “encrypted.”

Public and private cryptographic key pairs are the foundation of blockchain security. People use private keys, numbers they keep private, to create network transactions. Public keys, in contrast, identify public addresses where transactions come and go. A mathematical relationship ties the key pair together such that it is easy to derive the public key from the private key, but it takes vast amounts of computing power to derive the private key from the public key. This is what enables a blockchain user to send money to someone else by signing a transaction that basically says, “I give you this money.” The signature is analogous to signing a check or legal document in the offline world, but it uses math to prevent forgery instead of handwriting.

Digital signatures are widely used behind the scenes in computing to verify the authenticity and integrity of data. Browsers make sure websites are legitimate by checking digital signatures. Email servers and clients use digital signatures to ensure messages aren’t spoofed or manipulated in transit. Most computer systems will verify that software downloads are coming from the right source and haven’t been tampered with by confirming digital signatures.

Blockchains use digital signatures too. They use them to operate trustless, decentralized networks. “Trustless” may sound confusingly ambiguous, but when people say this in blockchain contexts, they just mean that blockchains need no higher authority—no intermediary, no central corporation—to oversee transactions. Through their consensus processes, blockchains can securely verify the senders of transactions all by themselves, and no one computer has the power to alter the rules.

Well-designed blockchains use incentives to get validators to behave honestly. Sometimes they also punish misbehavior, as in Ethereum’s case. Again, consensus systems are the basis of blockchains’ security assurances. If the costs to attack a blockchain are high enough and if most of the validators act honestly in accordance with their financial self-interest (as is true for the most popular blockchains), then the system is secure. In the unlikely event of a successful attack, participants could split, or “hard fork,” the network and roll back the blockchain to a previous checkpoint—another deterrent for attackers.

Even if some users are dishonest and would rather game a blockchain for profit, the system keeps everyone honest. This is the genius of the system: a set of incentive structures that makes it self-policing. Through well-calibrated economic rewards, blockchains get users to keep one another in check. And so, even though they may not trust one another, they can trust in the decentralized, virtual computer they are collectively helping to secure.

In practice, this trustlessness enables people to design networks that operate very differently from traditional online systems. Most internet services, like online banks or social networks, require you to log in to access your data and money. Companies keep your data and log-in credentials in their databases, which can be hacked or misused. Corporate networks use cryptography in some places but mostly rely on perimeter security, an approach involving a stack of technologies, like firewalls and intrusion detection systems, designed to keep outsiders and unauthorized parties away from internal data. The model is like putting a wall around a fort that’s stocked with gold and then trying to protect only the wall. It doesn’t work. Data breaches are so common they barely make the news anymore. The perimeter security model heavily favors attackers; it takes only a single gap for an attacker to break in.

In contrast, blockchains let you store data and money, but you can’t log in, because there’s nothing to log in to. Instead, if you want to do something like transfer money, you submit signed transactions to the blockchain. You keep your private data private;22 you don’t have to share it with any service you don’t want to. Unlike corporate networks, blockchains have no single point of failure. There are no internal servers to “break into,” as there are in typical internet services. Blockchains are open, public networks. “Breaking into” one, if you can even call it that, would require taking over a majority of the nodes on the network—an extraordinarily expensive and entirely impractical proposition.

A key concept in security is the “attack surface,” which refers to all the places an attacker might find vulnerabilities. The security philosophy of blockchains is to use cryptography to minimize the attack surface. In the blockchain model, there is no gold to be stolen from inside the fort. Data that needs to be private is encrypted. Only users (and anyone they authorize) have keys to decrypt the data. The keys need to be secured, of course, and users can choose to have third-party software custodians do this for them. The difference is that these custodians are solely focused on security. In the corporate model, all sorts of random businesses with little security expertise are tasked with storing and managing data. A hospital secures health records, an auto dealer secures financial records, and so on. Blockchains unbundle security from business functions and let specialists like custodians do what they do best.

When you hear about alleged blockchain hacks, these almost always refer to attacks on institutions that use crypto, or else they refer to old-fashioned phishing attacks on individuals. They don’t usually refer to hacks of blockchains themselves. In the exceedingly rare instances in which blockchains actually do get hacked, they almost always involve small, obscure, insecure blockchains. A successful attack can disrupt transaction processing or enable attackers to “double spend” the same money in multiple places. These attacks are known as 51 percent attacks23 because their conspirators must gain control of more than half a system’s validators to be successful. Feeble systems, like Ethereum Classic and Bitcoin SV, have succumbed to 51 percent attacks. Successfully attacking a major blockchain, like Bitcoin or Ethereum, would, in comparison, be so prohibitively costly as to be infeasible.

That hasn’t stopped people from trying. There have been many attempts to attack popular blockchains like Bitcoin and Ethereum, but none have come close to succeeding. The tech is battle tested. These blockchains are, in effect, the world’s biggest bug bounty programs. Hacking them could yield a massive financial prize, enabling attackers to transfer large sums of money, worth hundreds of billions of dollars, to themselves. But this has never happened. The security assurances of well-designed blockchains work not only in theory but also, so far, in practice.

Why Blockchains Matter

What would motivate someone to write software that runs on blockchains instead of traditional computers, like web servers or mobile phones? We’ll cover the answer to this in greater detail throughout part 3, but let’s quickly review blockchains’ novel properties.

First, blockchains are democratic. They are accessible to everyone. Blockchains inherit the ethos of the early internet, providing an equal opportunity to participate. Anyone with an internet connection can upload and execute whatever code they want. No user is privileged above any other, and the network treats all code and data equally. It’s a fairer framework than the gated status quo of today’s tech industry.

Second, blockchains are transparent. The complete history of their code and data is publicly available for anyone to inspect. If the code and data were available only to some people, that would put other participants at a disadvantage, which would undermine the egalitarian promise of the technology. Anyone can check a blockchain’s history and be assured that a valid process generated the system’s current state. Even if you don’t personally audit the code and data, you know others can and probably have. Transparency begets trust.

Third and most important, blockchains can make strong commitments about their future behavior—that any code they run will continue to operate as designed. Traditional computers can’t make commitments like these. Traditional computers are controlled by individuals or groups of people, either directly, in the case of personal computers, or indirectly, in the case of corporate computers. Their commitments are weak. Blockchains invert this relationship, putting the code in charge. The consensus mechanism described earlier and the immutability of their software makes blockchains resistant to human intervention. You don’t need to trust the promises of people or companies when using them.

Engineers at companies like Google, Meta, and Apple think about computers as machines they can set to do their bidding. Whoever controls the computer controls the software. The only assurances users receive about how the computers will operate are long “Terms of Service” legal agreements written by the software providers which mean little and almost no one bothers to read, let alone negotiate. (As the saying goes, “The cloud is just someone else’s computer.”)

Blockchains are different. They’re remarkable for what they cannot do as much as for what they can do. A blockchain can resist manipulation, a feature that may contribute to the misconception that they’re more like databases than computers. Blockchain software runs on other people’s computers, but—and this is the key—the software is in charge. A person or company can try to manipulate the software, but it will resist tampering. The virtual computer will continue operating as intended, despite attempts to subvert it.

This resistance to tampering goes not just for blockchains but also for the software that runs on top of them. Applications built on programmable blockchains like Ethereum inherit the platform’s security guarantees. This means apps—social networks, marketplaces, games, and more—can also make strong commitments about their future behavior. The entire tech stack, blockchains and anything built on top of them, can make these strong commitments too.

Critics who fail to appreciate the power of blockchains tend to have different priorities. Many people, including lots of workers within Big Tech companies, care about improving computers along familiar dimensions, such as memory and computing power. They see blockchains’ abilities as constraints—as weaknesses rather than strengths. It’s hard for people accustomed to free rein to appreciate that computers could improve on a dimension that is designed, in part, to undermine their authority.

Breakthroughs that fall outside the norm often get dismissed for the same reason that skeuomorphic thinking is more prevalent than native thinking in the early developmental stages of a new technology: preconceived notions hold innovation captive.

Still, you may wonder, why do computers and applications that can make strong commitments about future behavior matter? As Nakamoto showed, one reason is to create a digital currency. A requirement for successful financial systems is trust in their long-term commitments. Bitcoin commits that there will never be more than twenty-one million bitcoins, a commitment that makes bitcoins credibly scarce. Bitcoin also guarantees that people can’t play tricks like “double spending,” or using the same money in two places at once. These commitments are necessary but insufficient conditions for Bitcoin’s currency to have value. (The currency also needs sustainable sources of demand, a topic I’ll discuss in “Sinks and Token Demand.”)

Commitments don’t carry the same weight on traditional computers because the people or organizations who control them can simply change their minds. If, hypothetically, Google used the standard servers in its data centers to mint GoogleCoins and declared there will only ever be twenty-one million coins, nothing would bind the company to that commitment. Google management could change the rules, and the software, whenever it pleased, unilaterally.

Corporate commitments aren’t reliable. Even if Google put a pledge in its service agreements, it could at any time break those terms by revising the agreements, working around them, or shutting down the service (as it has done to nearly three hundred products to date).24 Companies simply cannot be trusted to keep promises to users. Fiduciary duty trumps other concerns. Corporate commitments don’t work, and haven’t worked, in practice. This is why the first credible attempt to create digital money was built on a blockchain and not by a company. (In theory, a nonprofit organization might be able to make long-term commitments to its users, but this has had its own challenges, which I discuss in “The Nonprofit Model.”)

Digital currencies are just the first of many novel applications that blockchains enable. Blockchains, like all computers, are canvases that technologists can use to invent and create. The unique properties of blockchains unlock a range of applications that simply can’t be created on traditional computers. The full range will be discovered in time, but many will involve building new networks that improve on existing networks by offering new capabilities, lower fees, greater interoperability, fairer governance, and shared financial upside.

Some examples include financial networks that commit to borrowing, lending, and other activities on transparent and predictable terms; social networks that commit to better economics, data privacy, and transparency for users; gaming and virtual worlds that commit to open access and favorable economics for creators and developers; media networks that commit to new ways for creators to make money and collaborate; and collective bargaining networks that commit to paying writers and artists fairly when AI systems use their work. I’ll discuss these and other networks, and how they lead to better outcomes, throughout the rest of the book (especially in part 5, “What’s Next”), but first we’ll cover the mechanism by which blockchains enable ownership.