1
July 2017, New York.
Hayden Adams was packing up his desk at the Siemens office. He had been there for one year as a mechanical engineering hire. The termination notice was brief, leaving no room for protest. As he placed his belongings into a box, he had no idea what to do next.
About a month later, his friend Karl Floersch, a researcher at the Ethereum Foundation, called him.
"Forget mechanical engineering. Start studying Ethereum right now. This is the future."
Adams had no idea what cryptocurrency was. He had heard the word "blockchain" before, but could not fathom what it had to do with his life. Yet a man who has just been fired does not have many options. He dipped into his severance pay and began learning Solidity programming.
One day, he came across a short post that Vitalik Buterin had written on Reddit. It described an idea for a decentralized exchange that could operate on a single mathematical formula, without a complex order book.
x * y = k
The quantity of Token A in the pool is x, the quantity of Token B is y, and their product k is a constant. When someone buys A, A leaves the pool and B enters. The price adjusts automatically to keep k constant. No buyers and sellers need to shout bids. The pool itself becomes the counterparty.
The New York Stock Exchange has market makers who stream buy and sell quotes, specialists who intermediate those quotes, and thousands of employees and hundreds of servers supporting the entire operation. Buterin's proposal was to replace all of that with a single line of math. Finance professionals scoffed. How could an exchange function without an order book? Market makers posting quotes, managing spreads, supplying liquidity—the idea of replacing decades of accumulated market structure with a single mathematical formula seemed like a joke.
But Adams, knowing nothing about finance, carried none of that baggage. Just as the Black-Scholes formula had demonstrated, sometimes an outsider's ignorance defeats an insider's inertia. Holed up in a small room in New York, he began translating Buterin's idea into code.
He received a $65,000 grant from the Ethereum Foundation—less than the annual bonus a Wall Street quant fund would pay a single trader. He had no office and no employees. In November 2018, at Devcon 4, a developer conference held in Prague, he unveiled to the world a protocol he had named Uniswap. Adams was the sole presenter, and most of the audience did not yet grasp what it meant.
For the non-specialist, the core idea behind Uniswap can be explained as follows. On a traditional exchange, a trade can only occur when a person who wants to sell apples meets a person who wants to buy them. The seller posts "Selling at 1,000 won" and the buyer posts "Buying at 1,000 won," and the trade is executed. This is the order book model. Uniswap works differently. Imagine two large jars. One is filled with apples, the other with money. Anyone who wants to buy apples puts money into the jar and takes apples out. As apples in the jar grow scarcer, the price of the remaining apples rises. Prices adjust automatically so that the product of the two jars stays constant. There is no need to wait for a seller. The jar is the counterparty.
Two more years remained before the summer of 2020 would arrive.
2
March 2020. The world stopped.
On March 11, the WHO declared a pandemic. On March 15, the U.S. Federal Reserve made an emergency cut to the federal funds rate, bringing it to 0–0.25 percent. Cities locked down, airports closed, and people were confined to their homes.
When interest rates converge on zero, money has nowhere to go. Bank deposit interest vanishes, government bond yields hit the floor, and capital flows into unfamiliar territory in search of higher returns. The same thing had happened in 2008, when capital migrated toward subprime mortgages. In 2020, one of the places capital migrated to was Ethereum.
As of January 1, the Total Value Locked (TVL) across decentralized finance (DeFi) protocols built on Ethereum stood at roughly $688 million. Not a trivial sum, but by Wall Street standards, a rounding error.
By June, TVL had crossed $1 billion. The growth was still gradual. Then one event changed everything.
On June 15, Compound launched the distribution of its governance token, COMP.
Compound is a decentralized lending protocol. Users deposit cryptocurrency and earn interest; other users post collateral and borrow. It replicated a bank's deposit-and-loan functions through a single smart contract—no building, no staff. Its founder, Robert Leshner, had graduated from the University of Pennsylvania with an economics degree and held the Chartered Financial Analyst (CFA) designation—a pedigreed finance professional. He wore suits and met with institutional investors; people called him "the DeFi builder who speaks Wall Street's grammar."
The mechanics of the COMP token distribution were simple. Both lenders and borrowers on Compound received COMP tokens as rewards. As Leshner explained it:
"We are distributing ownership of the protocol to everyone who uses it."
Something peculiar happened. As the market price of COMP skyrocketed, the value of the reward tokens began to exceed the cost of borrowing interest. People deposited funds, borrowed against those deposits as collateral, then deposited the borrowed funds again, repeating the cycle. It was called "spinning the windmill"—recursive leverage. Annual percentage yields (APY) of several hundred percent, sometimes several thousand percent, flashed across screens.
Peel back the layers and these yields broke down into three components. First, real yield derived from the interest borrowers paid. Second, the market value of the reward tokens. Third, the expectation that those token prices would keep rising. The second and third components overwhelmed the first. When token prices rose, yields rose; when yields rose, capital flowed in; when capital flowed in, token prices rose again. A cycle feeding on itself—whether this was sustainable was a question no one asked. The yields were too high to bother asking.
Someone coined a name for this activity: yield farming. The metaphor was agricultural: sow capital, harvest tokens. On Telegram and Discord, the word "degen"—short for degenerate—began to circulate as a self-deprecating badge of honor. It meant degenerate gambler, but those who used it wore it with pride. Their daily routine was staring at monitors at four in the morning, simultaneously tracking gas fees, liquidation risk, and reward token prices.
July TVL: $1.8 billion. August: $3.9 billion. September: $9.3 billion. A ninefold increase in three months.
The COVID lockdowns poured fuel on the explosion. People trapped at home opened their laptops. The stock market had Robinhood; the crypto market had MetaMask. Install a browser extension with a fox-head logo, and you could access DeFi from anywhere in the world without a bank account. A dozen tabs open in Chrome was the landscape of the era: one tab on Etherscan checking pending transactions, the Uniswap trading screen adorned with its pink unicorn, Compound's green dashboard, the Telegram "Whale Alert" channel. Numbers flickered on every screen, and with each flicker, assets grew or shrank.
Someone called this summer "DeFi Summer."
3
Cape Town, South Africa. In a room lit only by the blue glow of a monitor, Andre Cronje was performing the same task yet again. Open a browser tab and check Aave's interest rates. Open the next tab and check Compound's rates. Open yet another and check dYdX's rates. Move stablecoins to whichever offers the highest yield. By tomorrow the rankings will have shifted, requiring another check. The day after that, the same.
Cronje had studied law before pivoting to programming, following an unusual trajectory through telecommunications and fintech before entering the Ethereum ecosystem. The mindless repetition was unbearable.
"Why do I have to do this manually every day? The code should just move the money to wherever the interest is highest."
He carved out weekends to build a tool. He called it iEarn. Users deposited funds and a smart contract scanned multiple lending protocols, automatically routing capital to whichever offered the highest yield. It was the first yield aggregator. A task that bankers had performed for 600 years—deciding where to deploy a client's money—was now handled by a few lines of code. The difference was that bankers had working hours, and code did not.
When he open-sourced the code in early 2020, the response was immediate. There was no marketing and no fundraising. The mere fact that the code worked—and that it delivered interest rates dozens of times higher than a bank—was enough to pull in hundreds of millions of dollars.
In July, Cronje carried out what would become the most radical experiment in DeFi history. When he distributed the governance token YFI for Yearn Finance, he set his own allocation at exactly zero. No early-investor share. No venture capital stake. Only users who supplied liquidity to the protocol could earn tokens.
Fair Launch. It was hailed as the most decentralized token distribution since Bitcoin. Starting from $0, the price of YFI at one point surpassed the price of a single Bitcoin. Cronje could have become a billionaire—he would have been one had he simply set aside 10 percent for himself, as other protocol founders did. Instead, he walked away from a fortune.
In return, tens of thousands of messages poured in every day. Every time a bug surfaced, every time the price plunged, people held him accountable. Death threats were mixed in. Cronje once said in an interview, "I test in prod." Build fast, deploy fast, fix fast when something breaks. It was Silicon Valley's "Move fast and break things" applied to DeFi—except that in a protocol where users' money was at stake, the philosophy produced different consequences. What broke was not code but deposits.
He announced multiple times that he was "leaving DeFi," shut down his Twitter account, and came back. In the most greed-saturated market ever assembled, the architect who had designed the arena was indifferent to money—and burned out all the faster for it.
4
The three architects of DeFi Summer came from different worlds.
Leshner was a suit-wearing insider-innovator. For him, DeFi was not anarchic revolution but a technological evolution toward more efficient capital allocation. He was fond of the metaphor "Money Legos." A single Lego brick is just a piece of plastic, but combined, they can become anything. He was describing the composability between protocols—the ability to take funds borrowed on Compound, deposit them into Uniswap, and then pledge the resulting Uniswap liquidity position as collateral again. This composability is the key to explaining DeFi's speed of growth. Unlike traditional finance, where banks, securities firms, and insurance companies operate in isolation behind their respective licenses, regulations, and systems, DeFi protocols could call upon one another without anyone's permission. Within a single block, lending, trading, and liquidity provision executed in sequence.
Cronje was an anti-establishment hacker. He handed all authority to the community and stepped back.
Adams was a pure builder who knew nothing about finance. A fired mechanical engineer followed a friend's advice and translated a single mathematical formula into code—and it ended up surpassing the trading volume of the world's largest cryptocurrency exchange. When Uniswap's daily trading volume overtook Coinbase in September 2020, the industry was stunned—and for good reason: a company backed by thousands of employees, hundreds of millions in investment, and a license from the U.S. Securities and Exchange Commission had been defeated by open-source code written by a single person in a bedroom. It proved that an exchange could function without an order book, without a listing review process, without a customer support center.
Each embodied a distinct branch of DeFi: Cronje's radical decentralization, Leshner's fusion with traditional finance, Adams's mathematical infrastructure. Their motivations also differed. Cronje started from personal laziness—"I want my money managed automatically." Leshner carried an institutional vision—"a more efficient capital market." Adams was driven by a builder's curiosity—"I want to see if the math works."
Yet the three shared one thing. None of the protocols they built had a credit committee. There was no loan officer flipping through a binder, no risk management director questioning the presale-rate assumptions, no committee chair saying, "Let's approve it conditionally." Instead, there was code. Code verified collateral, code set interest rates, code executed liquidations. What five people had spent two hours doing for 600 years, code dispatched in 12 to 15 seconds. The problem was that code lacked the habit of asking, "Wait—is this really okay?"
5
A protocol executes a loan like this.
To borrow money on a protocol called Aave, a user posts collateral. Deposit $1,000 worth of Ethereum (ETH) as collateral, and the protocol allows you to borrow a certain amount based on a Loan-to-Value ratio (LTV). An LTV of 80 percent means you can borrow up to $800. The interest rate is set by an algorithm. Depending on the pool's utilization rate—the proportion of total deposits that have been borrowed—the rate automatically rises or falls. Supply and demand directly determine the price.
Up to this point, it looks similar to a bank. But a decisive difference exists.
To obtain a real estate project finance loan from a bank in Korea, the relationship manager (RM) must meet the developer, commission an appraisal, assemble a binder, and submit the file to the credit committee. Five people debate for two hours, the loan officer and the RM clash over whether a 73 percent presale rate is realistic, and the committee chair, after much deliberation, declares "conditionally approved." Weeks to months pass before the flow of 68 billion won is decided.
On Aave, it takes roughly 12 to 15 seconds—the block time of the Ethereum network in 2020. The moment the collateral-posting transaction is included in a block, code verifies the collateral's value, calculates the LTV, and approves the loan. No paperwork, no meetings, no stamps. No identity verification, either. The protocol does not know who the borrower is, nor does it need to. An address and collateral are sufficient. The same rules apply whether the borrower is a university student in Nigeria, a hedge fund in New York, or a salaried worker in Seoul. Geographic location, social status, past credit history—everything that human gatekeepers had relied upon as the basis for judgment was stripped away.
"Can this person repay the money?"
The question at the center of capital allocation is now reformulated. A branch manager at the Medici Bank weighed the reputation and repayment history of a Bruges wool merchant. A loan officer at a Korean savings bank scrutinized the developer's financial statements and presale-rate assumptions. A Moody's analyst fed correlation data on CDO tranches into a model. The form of the question changed over the centuries, but the structure remained the same—a human judged, a human decided, and a human bore responsibility.
Aave rewrites the question itself. It does not ask who "this person" is. Instead it asks: "Does this address have sufficient collateral?" The subject of the question shifted from human to code; the object shifted from person to number. Credit—that fundamentally human concept—was replaced by a mathematical threshold called a collateral ratio.
And code does one more thing. When the collateral's value drops and the health factor falls below 1.0, automatic liquidation triggers. A liquidation bot sells half the collateral, repays the debt, and collects a 5 percent bonus. 0.3 seconds. There is no appeals process. There is no committee that considers extenuating circumstances. When a project finance loan at a savings bank goes bad, the bank visits the developer, discusses an auction, and files a seizure petition with the court. Months, sometimes years, pass. On Aave, from trigger to execution, it takes less than a second. Code is cold, and that coldness keeps the system safe—so long as the code functions as intended.
Aave had one more invention: the flash loan. You could borrow millions of dollars without collateral, provided you repaid within the same transaction. Ethereum transactions are atomic—they cannot be split midway. Borrowing, deploying, and repaying must all occur within a single transaction, and if repayment fails, the entire transaction is reverted. As if it had never happened.
This structure is impossible in traditional finance. Walk into a bank and say, "Lend me a billion—I'll pay you back in 0.1 seconds," and you will be stopped at the door. There are application forms to file, credit checks to run, committees to convene. But in the world of code, time is granulated differently. Within a single block, one can borrow, capture an arbitrage opportunity, and repay principal plus fees. Inside the 12-second block time, a structure that no prior era of finance had imagined was operating.
Line up the differences in speed. At the Medici Bank, a bill of exchange took 20 to 25 days to travel from Florence to Bruges. The first government bond subscription at the Bank of England in 1694 took 12 days. Plugging the Black-Scholes formula into a TI calculator to price an option took several minutes. During the 2010 Flash Crash, the Dow Jones fell 1,000 points in 34 minutes. Aave's loan approval takes roughly 12 to 15 seconds; liquidation, 0.3 seconds. The compression of capital allocation speed is mapped out here—from bills of exchange to smart contracts, from 25 days to fractions of a minute.
6
By December 31 of that year, the TVL across DeFi protocols had reached $15.4 billion—22 times the January figure. By protocol: Compound at $2.7 billion, Uniswap at $2.4 billion, Aave at $2 billion, MakerDAO at $1.6 billion, Yearn Finance at $450 million. The number of users had nearly quadrupled from roughly 150,000 in January to 600,000.
On the surface, this looks like a success story. But beneath those numbers, cracks were forming. Behind the thousands-of-percent APYs lurked circular logic. The value of reward tokens depended on new capital inflows; inflows depended on high APYs; APYs depended, in turn, on the value of reward tokens. Vanishingly few protocols generated positive returns after stripping out token rewards. The rest ran on a structure in which new capital paid the returns of old capital—a formation structurally resembling what Charles Ponzi had engineered with postal reply coupons in Boston in 1920.
On October 26, a single transaction appeared on Etherscan. It was an attack on Harvest Finance. Open the transaction and you see an execution trace with dozens of nested internal calls. Flash-loan millions of dollars into existence, dump them into a Curve pool to distort the price of the stablecoin, deposit at the distorted price into a Harvest vault, withdraw immediately, repeat. Finally, repay the flash loan and vanish with the profit. Twenty-four million dollars. It took seconds.
Discord chat rooms erupted in panic.
"Why is my balance dropping?" "Admin?" "It's over."
This was not an isolated incident. In February, the bZx protocol was exploited. In August, YAM crashed 99 percent within 35 minutes. In September, SushiSwap founder Chef Nomi sold $14 million from the developer fund—a "rug pull," the act of yanking the carpet from under users' feet, a term that became permanently etched into the ecosystem's vocabulary. The common thread was singular: whether attacker or insider, none had broken the rules of the code. They had exploited the gaps within those rules.
Remove the gatekeeper and friction disappears. But in the space where friction once stood, gaps remain. The question raised during The DAO hack of 2016—is exploiting a vulnerability in code theft, or the fulfillment of a contract?—was repeating itself four years later, still without an answer. Back then, Ethereum had taken the extreme measure of a hard fork, reversing history—violating its own principle that "code is law." For the attacks of DeFi Summer, not even that remedy was available.
7
Two years later, the largest edifice that code had deemed "safe" collapsed.
Terra was an algorithmic stablecoin. It consisted of a token called UST, pegged 1:1 to the U.S. dollar, and a backing token called LUNA. Conventional stablecoins (USDT or USDC, for instance) deposit dollars in a bank and issue a corresponding amount of tokens. Behind every dollar of tokens sits a dollar of real-world assets. Simple, but robust. Terra's approach was fundamentally different. It aimed to maintain its peg purely through algorithms and arbitrage incentives, without a single dollar of reserves. Stability without collateral. It was like an arch without external support—standing so long as forces remain in equilibrium, but collapsing entirely the moment one side gives way.
The mechanism worked as follows. If UST's price fell below $1, arbitrageurs would buy UST cheaply on the open market and redeem it on the protocol for $1 worth of LUNA. UST would be burned, reducing supply, and the price would return to $1. The reverse worked the same way—in theory.
The founder was Do Kwon, a Stanford computer science graduate. He styled himself "the Master of Stablecoins" and retorted to critics, "I don't debate poor people on Twitter." He once declared, targeting MakerDAO's DAI, "By my hand, DAI will die." It was a conviction that his algorithm was flawless—the same species of conviction that LTCM's Merton and Scholes had harbored immediately after receiving their Nobel Prizes, certain that their model was perfect.
The growth engine was Anchor Protocol. Deposit UST and receive 19.5 to 20 percent annual interest. The marketing line was "DeFi's savings account." More than 70 percent of all circulating UST was locked in Anchor. The problem was that borrowing demand was nowhere near sufficient to cover this interest. The shortfall was subsidized by Terraform Labs and the Luna Foundation Guard (LFG). In other words, interest was being paid with subsidies. If the subsidies stopped, Anchor would wobble; if Anchor wobbled, UST would wobble; if UST wobbled, the entire structure would collapse.
Saturday, May 7, 2022. A massive sell-off began. UST poured out of the Curve pool. The price slipped to $0.985. Do Kwon's response: "I don't really care about what happens on a weekend."
Monday, May 9. Funds began fleeing Anchor. UST crashed to the $0.60–$0.70 range. Do Kwon tweeted: "Deploying more capital—steady lads."
But the algorithm's death spiral would not stop. UST selling → LUNA minting → LUNA value declining → UST confidence collapsing → more UST selling. Eighty thousand Bitcoin—roughly $3 billion—the entirety of LFG's reserves, was poured into defending the peg. It was not enough.
May 12: LUNA's price fell below $0.01. Its circulating supply had exploded from several hundred million to 6.5 trillion—a digital reenactment of the Weimar Republic's hyperinflation of the mark in the 1920s. Friday, May 13: validators halted the network. Delisting notices cascaded. The combined market capitalization of UST and LUNA had been approximately $40 billion. In seven days, not a trace remained. A savings bank takes years to go insolvent; LTCM took months to collapse. In the world of code, seven days was enough.
LUNA's price trajectory can be summarized in a single data point. April 2022 all-time high: $119. May 13: $0.000001. What lay between those two numbers was the failure of an algorithm—and the hubris of the human who designed it.
A chain of bankruptcies followed. Three Arrows Capital (3AC) was wiped out by its LUNA investment and toppled. Celsius and Voyager, which had lent to 3AC, froze customer funds and filed for bankruptcy. At the end of this chain, the collapse of FTX in November 2022 lay in wait—it was later revealed that Alameda Research had used FTX customer funds to cover losses sustained in the Terra debacle.
In South Korea alone, the number of victims was estimated at approximately 200,000. Some had deposited their retirement savings into Anchor's 20 percent interest. Others had deposited university tuition. The marketing slogan "DeFi's savings account" evoked the safety of a bank deposit, but behind it stood an uncollateralized algorithm. Just as savings bank depositors in Korea were unaware that their money was flowing into real estate project finance, Anchor depositors did not know what mechanism their money was resting upon. Six hundred years ago as now, the gatekeeper's failure ultimately inflicts the greatest harm on those who understand the least.
The Joint Financial and Securities Crime Investigation Unit of the Seoul Southern District Prosecutors' Office1 designated it as its first-priority case. Do Kwon fled from Singapore to Dubai, then Serbia, then Montenegro. He was arrested in March 2023 on charges of traveling with a forged passport. He was extradited to the United States to stand trial, and in November 2024, he was found guilty on nine counts including fraud.
The hubris of genius is the most relentlessly recurring pattern in the history of finance. Lorenzo de' Medici neglecting the bank while immersing himself in art in the fifteenth century. Isaac Newton reinvesting in the South Sea Company and losing his entire fortune. Merton and Scholes bankrupting LTCM shortly after receiving the Nobel Prize. Do Kwon was the 2022 edition of the same pattern. The claim that "the algorithm is perfect" was the blockchain translation of "the model is perfect," and the ending was the same.
8
The collapse extended far beyond a single protocol.
After the Terra/LUNA implosion, total DeFi TVL shrank from $229 billion in May 2022 to $81.7 billion in June. Hacking losses surged—over the course of 2022, roughly $3.1 billion was drained from DeFi through hacks and exploits. It was a year in which the slogan "code is law" collided with the reality that "a flaw in the code is a catastrophe."
The deeper casualty was a belief: that code could replace the gatekeeper.
DeFi Summer had proved its point. Lending is possible without a bank, trading is possible without an exchange, and capital can be allocated without a credit committee. A single formula—x * y = k—overtook Coinbase in trading volume. A single smart contract performed the function of a savings bank credit committee in 12 to 15 seconds. A fired mechanical engineer, a South African law graduate, a CFA holder in a suit—none had received a license from the financial establishment, yet the code they built worked.
But what the Terra/LUNA collapse revealed was equally clear. Code, too, has its version of the gatekeeper's failure. Only the form differs. Where human gatekeepers failed through bias and greed, code gatekeepers fail through design flaws and self-reinforcing feedback loops. Between Portinari of the Medici Bank extending excessive credit to the Burgundian court and thereby endangering the parent bank, and Anchor Protocol promising an unsustainable 20 percent interest rate and thereby destroying all of Terra, centuries stand. Yet the structure of the failure is identical. Someone's judgment that underestimated risk collapsed together with the system's safeguards.
If there is one difference, it is this. Portinari had a name, and the Medici headquarters could hold him accountable. The major shareholder of Busan Savings Bank in 2011 had a name, and prosecutors indicted him. Moody's analysts had names, and Congress subpoenaed them. The anonymous developer who executes a rug pull has no name. The address that carried away tens of millions through a flash loan attack has no identity. The whale that captures a governance vote has no obligation. Where the gatekeeper vanished, accountability vanished with it.
And yet the experiment that began in 2020 did not end. TVL, which stood at $15.4 billion at the close of 2020, soared to $181.3 billion by November 2021. It plunged after the Terra crisis, falling to $39.8 billion by the end of 2022—but it did not disappear. Uniswap was still operating. On Aave, loans were still being executed every 12 to 15 seconds. Compound's interest rate curve continued to shift quietly according to supply and demand. The rules written in code did not stop, and capital continued to flow atop those rules.
Leshner stepped back from Compound and founded a company called Superstate. Its business was tokenizing U.S. Treasury bonds and placing them on the blockchain—using DeFi's grammar to handle the assets of traditional finance. Cronje returned, launched new projects, and left again. Adams continued to lead Uniswap Labs, steadily improving the protocol. The subsequent paths of these three men reveal where DeFi's three branches are heading: fusion with traditional finance, the solitude of pure decentralization, and the quiet evolution of infrastructure.
DeFi's lasting creation was a set of rules written in code. The rules were transparent, open to everyone, and running without pause. Aave's LTV parameters were publicly accessible. Uniswap's x * y = k could be verified by anyone. Compound's interest rate curve was specified in the code itself. It was a different world from one in which a savings bank credit committee's decisions were buried between the lines of meeting minutes.
Yet the choices of which rules to write, which parameters to set, and which protocols to trust still belonged to humans. It was humans debating in Aave's governance forum whether to set the LTV at 75 or 80 percent. It was humans deciding whether to supply or withdraw liquidity on Uniswap. Code enforced the law, but it was humans who wrote it—at least until 2020.
Already, seeds of a transformation were becoming visible within DeFi itself. MEV (Maximal Extractable Value) bots were automated agents that reordered transactions to capture arbitrage. Bots that automatically analyzed governance proposals and voted were emerging. Yearn's vault strategies were growing increasingly complex, reaching speeds at which human intervention at every step was no longer feasible.
On the Ethereum blockchain, a new block was produced every 12 seconds. Within those blocks, loans were executed, trades were settled, and liquidations were completed. No one stamped an approval. No one sat in a conference room. And yet, with increasing frequency, observers began to notice that the entities sending transactions inside those blocks were not human. Behind the wallet addresses was not a person but code. Beyond the era in which code enforced rules, the door was opening to an era in which code chose the rules.