Author: Dom Cooke

Compiled by: Deep Tide TechFlow

Introduction: Colossus magazine conducted an in-depth visit to Hyperliquid's Singapore office, producing the most detailed profile of founder Jeff Yan to date. With 11 people, zero VC funding, and annual profits exceeding $900 million—how did this three-year-old protocol capture 37% of the decentralized perpetual contract market share in just two years?

From Yan's journey of running quantitative strategies on a TV in Puerto Rico, to rejecting a VC check with a valuation of $1 billion, airdropping $16 billion worth of tokens to users, and now deploying perpetual contracts for crude oil, gold, and the S&P 500 on Hyperliquid with an independent team, this lengthy article elucidates one question: Why would a physics Olympiad gold medalist give up a life of financial freedom to rebuild the entire financial system?

text:

In the early hours of a Friday in January, in Saint-Léger-sous-Cholet, western France, a 43-year-old man was abducted from his home. He was driven 30 miles to Basse-Goulaine, where he was beaten, tied up, and dumped on the roadside. Twelve hours later, in Verneuil-sur-Seine, a suburb of Paris, three armed men kicked down the door of a house, beat a couple in front of their children, tied the family of four up with cable ties, ransacked the house, and then left by train.

This is the 70th such attack worldwide in less than a year.

Two days later, I boarded a flight to Singapore.

The bodyguard and the plush cat in the office

I came here to visit an 11-person team, but the first person I met after arriving at the office didn't belong to the team. He was a burly American with a buzz cut and stubble, sitting behind a small table in the corner of the break room with an Apple laptop in front of him. His build clearly indicated he wasn't there to write code.

He is a bodyguard.

One of Hyperliquid's co-founders, whose pseudonym is iliensinc (short for Aliens Incorporated), led me from the hotel to the office. As we walked through tree-lined streets, she told me they hadn't always been based in this area of ​​Singapore. The company initially operated from a co-working space in the financial district, but her co-founder—the only one on the team who didn't use an alias—began to attract attention. At first, it was just passersby glancing at him, trying to recognize his face. Then strangers started approaching him. Later, someone followed him into an apartment building elevator. So the company moved to a quieter location, in a building no one would think to look for them.

Even their cleaning lady had no idea what they did. In her mind, she cleaned a company that made plush cat toys. With 34 plush toys in the office, the misunderstanding was understandable. The company mascot was a cat named Hypurr, with 12 of them perched on cabinets, but there were also sharks, lizards, koalas, penguins, and dragons, several perched on Dell monitors like furry gargoyles. Most of the toys were brought by an engineer—his wife wouldn't let him bring them home anymore, so he brought them to work. The team didn't correct the cleaning lady's misunderstanding.

Because Hyperliquid is one of the most profitable companies per employee on Earth. Last year, its 11 employees generated over $900 million in profit. At only three years old, with a market capitalization of $10 billion, it has never accepted a single penny of venture capital. The key figure behind it, Jeffrey Yan, is 31 years old. In an industry where "the greater the success, the easier it is to be held hostage," he has somewhat involuntarily become one of the most recognizable figures.

TV dealer in Puerto Rico

Before Hyperliquid, Yan lived in Puerto Rico and almost single-handedly ran one of the largest anonymous trading businesses in the crypto space, called Chameleon Trading—Chameleon being his high school gaming ID. He started with $10,000 he had saved and grew it at a rate of several thousand percent annually for two and a half years. After telling me about his returns, he immediately tried to dissuade me from thinking he was anything special. I took note of his humility. I also took note of how much money Chameleon had made him. At 27, he was financially independent. To the surfers, bartenders, and waiters of San Juan, he was just another young man in beach shorts.

Now he sits cross-legged in a grey armchair in a guarded office in Singapore, barefoot, wearing black shorts and a dark blue T-shirt, explaining to me why the entire financial system needs to be overhauled. What I want to know is: why did he trade his first life for his second?

"It's not about the money," he said. Yan didn't come from a wealthy family, and his lifestyle showed no interest in the lifestyle of the rich. He wore the same Lululemon shorts and T-shirt every day. He had 15 pairs of shorts and 10 T-shirts, three of each color. There was no trace of wealth in the office either. The furniture was left by the previous tenant. The only things the team had added were two board games in the break room, NFTs on the wall, and the plush cats. I found four books on the shelf and recognized one of them as Frank Slootman's *Amp It Up* , a management book whose core argument is that "most people don't work hard enough." I mentioned it to iliensinc. She shrugged. The book wasn't theirs, but the ideas were. The three unopened bottles of Grey Goose and Macallan in the kitchen weren't theirs either; they were left over from a community event two years ago—an event that didn't meet the minimum spending requirement. The team was drinking tea.

It's not because of a love for crypto. Bitcoin has fallen about 30% from its early October highs; the cryptocurrency that was supposed to replace gold has fallen 30%, while gold has risen 7% in the same period. Most tokens have fallen even more drastically. I asked Yan what he thought of the pervasive pessimism in the industry, and he didn't defend it. "There are definitely a lot of questionable practices in this industry," he said. "People are starting to realize that a lot of things aren't as advertised, which is probably a good thing." He doesn't consider Hyperliquid a crypto company. "Nobody says we're an internet company right now," he told me. "We use crypto, but that doesn't define us."

Of the 11 members on the team, only two, including Yan, had prior experience in crypto. This was intentional. In the early days of the crypto community, in Yan's words, it was primarily comprised of people looking to make quick money. He was building something for the long term, which resonated more with tech-minded individuals than traders. However, this also reflects a supply issue. Hyperliquid recruits from the podiums of the International Mathematical and Scientific Olympiads. Yan won a gold medal in physics at 18. One engineer won a silver medal in informatics, and another was selected for the US national team training camp. Yan wanted to recruit more people, and after my visit, he added two more. But there are already few people at this level willing to work in crypto; many have been drained by scams and empty promises over the years, and recently, AI has further diverted a portion of them.

So what exactly is Yan—someone who has made enough money to do anything—doing here?

There are heavens beyond heavens.

The answer, at least to the outside world, is becoming increasingly clear.

Hyperliquid is a blockchain on which its own exchange is built. In traditional exchanges, the company holds your funds and controls the infrastructure. On Hyperliquid, you hold your funds yourself, and the platform is public. Yan's vision for it, without a hint of irony, is to support all finance. Whether this is ambitious or absurd depends on whether you're looking at the plush cats or the platform's data. Because in the months following my visit, markets that had used trading methods for over 100 years began to shift in subtle but measurable ways.

Hyperliquid started with perpetual contracts in 2023. Perpetual contracts are a type of derivative and the largest single market in the crypto space. Essentially, a perpetual contract is a bet on the price of an asset you never own, and unlike traditional futures, it never expires. This market is 6 to 8 times larger than the spot market, at approximately $7 trillion per month. Until recently, almost all trading volume ran on centralized exchanges. The largest was Binance. No decentralized platform could challenge it. Hyperliquid was the first to grow to approximately 14% of Binance's market share.

Then, in October 2025, Hyperliquid did something centralized exchanges couldn't: anyone could launch a new perpetual contract market on the platform for any asset with a price source. An independent team called Trade[XYZ] was the most active deployer. They started with the silver market. By January, their 24-hour trading volume had reached about 2% of the CME (Chicago Mercantile Exchange, the world's largest derivatives exchange, founded in 1898). Then Trade[XYZ] launched crude oil. Oil had been trading on a market closed on weekends. One Saturday in late February, the US and Israel began bombing Iran. The CME was closed. Hyperliquid wasn't. Daily crude oil trading volume soared from $21 million to $3.7 billion. A month later, Trade[XYZ] launched S&P 500 perpetual contracts, officially authorized by S&P Dow Jones Indices, for 24/7 trading, including weekends.

The most influential product on Hyperliquid is now being built by people who are not on Yan's team and will not be joining in the future.

The founder of Trade[XYZ] requested anonymity. He bought his first Bitcoin for $66 in 2013 and has since been an investor rather than a builder, never intending to start a business. He told me that he wouldn't be in the crypto industry if it weren't for Yan. "Hyperliquid has the opportunity to save crypto," he said.

But none of this explains why Hyperliquid became what Yan described—an industry that "looked like it was going to succeed until it suddenly failed"—nor why he gave up his life in Puerto Rico to validate it. These questions followed me on my first afternoon in the office, when iliensinc and I were chatting in the break room, a plush cat on the table, the air still smelling of ginger and sesame from lunch. She told me that three years ago, when Yan announced the end of Chameleon, the team had asked him the same questions. Her answer didn't start with encryption, but with what kind of person Yan was. She said, "You should ask him about his mother."

There are always people better than you, and there's always something beyond your understanding.

Yan likes to hold meetings outdoors. We sat on the covered terrace with four gray lounge chairs and a coffee table. Cars passed by below. Every few minutes, a gardener started his lawnmower. The pedestrian crossing announcements played intermittently.

Yan tucked his legs under his body. I asked him about his mother, and he thought for a moment. She had a favorite saying, he said, a Chinese idiom: "There are people beyond people, and heavens beyond heavens." She wasn't the kind of mother who pressured her child, but she wanted him to know that no matter how great you think you are, you only see a small part of the outside world.

She raised him and his sister alone in the heart of one of the most lucrative geographical zones in American business history—Redwood Shores, between San Francisco and Palo Alto. Oracle's mirrored headquarters towered over the neighborhood. His neighbors were engineers and product managers whose children were already preparing Yan for the kind of life he would later build. His parents, both Chinese immigrants, divorced when he was in third grade. His father left. His mother was an accountant, working overtime every tax season, which he could sense. "I could tell other families were wealthier than us," he said, "but I never resented it. Going out doesn't cost much."

His school lacked an academic competitive atmosphere. Despite the common saying, his mother never pushed him. No one pushed him at all before he became a teenager. He played outside, went to school, came home, and continued playing. By the standards of his zip code, he was a rare breed: a free-range child.

Photo: Yan and his dog Max at Redwood Shores

In eighth grade, a friend who had just transferred from a private school dragged him into a math competition. His friend wanted a companion. Yan had never seen anything like it. School math was completely different. There were no formulas to memorize, no calculations to grind. You were given a problem, sometimes just a sentence, and then you had to find your own way in. The answer wasn't a number, but a proof—a complete argument explaining why something must be true. Finally, they ranked you, like ranking sprinters. For Yan, this was a fusion of the best of sports and the best of understanding the world.

That summer, he would get up at five every morning, download past competition papers from the internet, and work on them alone in his room. He had no tutor, couldn't afford summer programs, and no one required him to do this. "It turned out I was incredibly competitive," he said. "There was one competition I had no idea about beforehand, where other kids had been running it their whole lives, and I was lagging behind."

A year later, in ninth grade, he made the U.S. Mathematical Olympiad training team—one of the top 50 high school students in the country. He was one of the youngest in the room. He didn't make the national team. He said he didn't care. For three weeks, he sat with teenagers who could stare at three sentences for five hours and find truths invisible to most people. Mathematics doesn't have Federer, Yan told me, but at the highest level, there's something similar to Federer. There's a style to the work, an elegance in the construction of proofs, something he saw up close for the first time at the training camp. "It's like playing football with Tom Brady," he said, "but the nerdy version. Most people don't have that feeling."

The following year, he was eliminated in an intermediate round of math selection. He was 16 and had to wait a full year to try again. I asked if that was his first experience of failure. "Losing is a common experience," he said. "Most people are losers. Usually, there's only one winner."

The problem wasn't losing, it was emptiness. "There was a void," he said. "I should be learning something." So he found some physics textbooks for upper-level students. His school didn't offer physics until the third grade, but he had just learned calculus and understood for the first time what it was for. He discovered the Feynman Lectures. "I watched them like a TV series." Within a year, also through self-study, he became one of the top five young physicists in the United States.

He was selected for the US Physics Olympiad team and went to Estonia (his first trip to Europe), where he won a silver medal. The following summer in Copenhagen, he won a gold medal, ranking 24th in the world. At 18, he returned to the Bay Area and understood that his mother's words about the sky were true. There were exactly 23 people above him.

Harvard and Hudson River Trading

Harvard University covered almost all of his tuition. In the spring semester of his freshman year, Yan took Computer Science 124, Data Structures and Algorithms. This course, primarily taken by sophomores and juniors, was notoriously difficult. Students called it a "necessary evil" in course reviews. One comment warned, "No social life. You'll never have a girlfriend." Out of 150 students, Yan, a freshman, ranked first, and not by a narrow margin.

At Harvard, students are assigned to upperclassman dormitories after their freshman year. Yan was assigned to Pforzheimer, where he became close friends with Scott Wu. Wu was two years younger than him, and Yan had initially met him at a summer program for Olympiad participants. Wu had represented the United States three times in the International Olympiad in Informatics, winning a gold medal on his last attempt, and later co-founded Cognition AI. When Wu was also assigned to Pforzheimer in his sophomore year, he messaged Yan: "Yo, I'm in Pfoho." Yan replied: "Let's go!"

Wu often found Yan in the common room, sitting at the grand piano, teaching himself jazz, repeatedly playing a particular phrase until it was perfect. They played chess, Go, and poker together, spending countless hours discussing what it meant to be "the best at something." Yan would talk about Faker—the greatest League of Legends player of all time—as well as great Go players and the best high-frequency traders. "He was always thinking, what makes someone special," Wu told me, "What is the essence of this field? What does it truly mean to be good at it?"

Wu remembers Yan having an unusual, contrarian way of thinking. Most Harvard students, given the same information and in the same environment, would arrive at roughly the same conclusions. Yan never did. Wu also said he was very humorous. "A real deadpan comedian. He would say things that were completely unexpected, but delivered in the most mundane way."

Yan worked throughout the summer. He interned at Google X, developing tools for the self-driving car project before it became Waymo. He also interned at Tower Research Capital (a trading firm). During his senior year, he worked part-time at Nuro (another self-driving car company), mainly because he felt his four-year college program was at least an extra year.

During his junior year winter break, he and Wu were two of the ten interns in Hudson River Trading's inaugural internship program. HRT is one of the world's most successful quantitative trading firms. Other interns in the same cohort included Alexandr Wang and Jesse Zhang, who later founded Scale AI and Decagon, respectively. The internship was a three-week competition. In each round, Wu and Yan took the top two spots.

After graduating with a bachelor's degree in mathematics and a master's degree in computer science, Yan joined HRT full-time at the end of 2017, assigned to the US stock algorithm team. He and his manager sat down for a weekly meeting. This manager had mentored many newcomers. These meetings usually followed a pattern: the newcomer would hit a wall in the code, they would solve it together, and the newcomer would go back to hitting the next wall. Yan didn't hit walls, his manager recalled. He came in with ideas. The meetings were efficient, but something about it bothered his manager. It took him a while to figure it out. Yan had done everything right, but these things seemed irrelevant to him. Eight months later, when Yan said he wanted to leave, his manager understood. The resignation email he sent to Yan was unusually warm by HRT standards.

Yan likes HRT. He believes trading is the purest game you can play in real life. Whether you're right or wrong, the market will tell you. You're competing with some of the world's smartest people, and in this brutal game, you create an extremely valuable product for the world: a highly liquid and efficient market. But he spent eight months improving an already excellent system in a company that would thrive without him, which meant he couldn't answer the question that kept nagging at him: What value have you added to the world?

Encrypted Christmas

In December 2017, the answer came to him. Bitcoin was approaching $20,000. Coinbase was the most downloaded app in the US. Billions of dollars poured into ICOs like Jesus Coin. It was a crypto Christmas. Yan first heard about Bitcoin during his internship at HRT, when two former partners came to introduce it to interns. At the time, no one was impressed. But while still at HRT, he discovered the Ethereum yellow paper, which described a computer whose computational results were universally accepted and which no individual could shut down. He was exposed to finance every day and could see what finance was running on. That paper described a way of replacing trust with code. "I felt I could do something that could completely change finance."

Around April 2018, he left HRT to create a prediction market where users could bet on the weather, elections, or sporting events—anything with a result. It ran on a blockchain, with no single entity controlling the funds. The core of the architecture was an idea Yan believed he and his co-founders had first conceived: off-chain matching and on-chain settlement, because Ethereum was too slow for real exchanges. Funds were held in smart contracts, governed by code, but users experienced a fast and smooth experience. The promise of cryptographic decentralization meant no friction. He and his college roommate Brian Wong (who also left HRT) developed it at Binance Labs' first incubator in San Francisco, calling it Deaux.

Kalshi was founded in 2019, following the same direction. Polymarket followed suit in 2020. Today, Kalshi and Polymarket are valued at over $40 billion.

Deaux reached 100 users.

As Yan finished speaking, the sky over Singapore suddenly cleared. Large, heavy raindrops began to fall, the kind that could fill drains in minutes. From the balcony, we could hear the rain pounding on the street below, and the hissing of tires over puddles made the cars even louder.

“That thing had absolutely no chance of succeeding,” he continued. By the time Deaux launched, Bitcoin had already fallen more than 80% from its peak. Jesus Coin died and never came back to life. Nobody wanted to bet on tomorrow’s weather. More importantly, Yan and Wong barely considered regulation. Kalshi later spent three years battling regulators before launching his product.

When Deaux closed down, Scott Wu was the only person on Earth who felt sorry for it. He was one of five regular customers.

Yan returned most of his $450,000 investment. He was still in his non-compete period at HRT, so he went to Lake Tahoe, California, and skied with a friend who was also in his non-compete period until the snow melted. Then he went to China, Japan, and Peru, traveling on a budget. He tried to convince me that there are actually quite a few skills involved in being a tourist. He didn't have those skills.

At the end of 2019, his non-compete period ended, and Yan moved to Puerto Rico—where he could legally reduce his capital gains tax rate to near zero. He had $10,000 and a gut feeling that something big was about to happen.

His girlfriend went to Puerto Rico with him. They shared a one-bedroom apartment by the sea for less than $2,000 a month, but "sharing" meant a certain degree of cohabitation, and Yan didn't make time for that. He didn't have a monitor, so he commandeered the television and set up camp in the living room. For the first year or so, she could get about 30 minutes of his attention each day. The rest of the time belonged to the trading algorithms scrolling on the television.

Yan works at least 14 hours a day, easily reaching 100 hours a week. He started with Python scripts, writing code to connect to cryptocurrency exchanges and letting the program trade for him around the clock. He monitors them, adjusts the logic, tracks the data, and starts over if he's not satisfied.

He can do this because the openness of crypto is unprecedented in traditional finance. In HRT's US stock trading, placing an order on one exchange requires connecting to 13 exchanges across three data centers in New Jersey, complying with a host of SEC regulations called Reg NMS, retrieving CME futures data via a microwave link in Chicago, and incurring tens of millions of dollars in infrastructure costs. In the crypto market, everyone—whether you're an HRT employee or someone working with a television—connects to the same HTTP infrastructure originally used for web pages. You only need a single server on AWS.

For nearly two years, his girlfriend had no idea what was happening on the other side of the television. Their lives remained unchanged. Same rent, same food. She knew he was passionate and hardworking, and assumed he was doing well, but there was no material evidence. Then, one Friday evening in the summer of 2021, she tried to take him out for dinner that had been booked a week earlier. He wouldn't go.

“You don’t understand,” he told her, “if I don’t fix this bug now, I’ll lose $100,000.”

I didn't even want Chameleon Trading or a bottle of Vitamin C.

After that night, Yan decided to turn it into a real company. He needed someone who could do everything except coding. There was someone in Harvard's Pforzheimer dorm who seemed to manage everything in life perfectly—a skill so foreign to him. But the last time he heard of iliensinc, he was in Asia, working as a strategist at a venture capital firm, flying between Tokyo, Seoul, and Hong Kong.

When he contacted her, he found she was in San Francisco. COVID had brought travel to a standstill, and the job that used to take her all over Asia had turned into midnight phone calls from her apartment. Yan explained what he needed. He didn't provide a job description, a title, or even much of what she would be doing. But having spent three years investing and observing founders, she felt that whatever Yan was describing, he wasn't the kind of person you should short.

The company officially adopted a name—Chameleon Trading. iliensinc began joining Zoom meetings with the exchange's business development team, adding a layer of professionalism to what was essentially just a man's operation on the San Juan beach. Beneath the giant market makers (companies like Jump Trading, Tower, HRT, and Jane Street), there exists a tier of anonymous trading firms whose size is impossible to fully verify. Chameleon is one of the largest among them.

By 2022, Yan couldn't sit still. After four years in the crypto industry, dabbling in various markets—centralized and decentralized—he began to genuinely care about the industry itself, not just his own profits and losses. Bitcoin gave the world a way to hold and transfer funds without intermediaries. Ethereum gave the world a computer that no one could shut down. In between, they had laid almost everything needed to rebuild the financial system. But the industry had done almost nothing. The two largest exchanges, Binance and Coinbase, were both centralized. Crypto had been reintroducing what it was supposed to eliminate.

That summer, iliensinc arranged a team outing at a hotel in the English countryside. She had grown Chameleon to six people. Yan gave her a budget of one Bitcoin. The team flew to London, visited the British Museum, and spent a few days at a country estate. Their leader, away from the screen for the first time—at least for everyone's memory—wasn't entirely comfortable.

Back in Puerto Rico, trading continued. But Yan told the team they were going to start building something new. He wasn't sure what. He had some ideas, but none of them convinced him. All he knew was that Satoshi Nakamoto's original vision for Bitcoin was being quietly buried by the very industry Nakamoto had created, and this bothered him more than someone who had made millions off of everything that industry had failed to build.

According to his team, Yan was breathing too much fresh air outside.

In November 2022, FTX, the world's third-largest cryptocurrency exchange, valued at $32 billion, collapsed in nine days. It had been lending user deposits to Alameda Research, a trading company run by the founder's girlfriend. When users tried to withdraw their funds, the money was gone. Less than six months earlier, Terra, a $50 billion crypto ecosystem, went to zero in three days. It attempted to build a dollar-pegged currency solely based on its own system logic. The algorithm intended to maintain the peg instead accelerated the collapse. The two largest projects in the industry's history died within the time it takes for the sun to orbit halfway around the Earth.

Yan had seen enough. He told his six-person team that the deal was over. They could disagree, but Chameleon was finished. If he was wrong, they could always go back and do the deal. Several people did disagree, and several later left. But this didn't change Yan's decision. No investors needed to be consulted, no board needed to be persuaded—it was his money, his decision, and he had a new mission.

"I was overly confident that FTX would be the end of centralized exchanges," Yan told me, "but that belief was useful; it gave me the determination to attack this huge market."

Build a chain from scratch

The market he was referring to was perpetual contracts. This concept originated from an insight made by economist Robert Shiller in the 1990s. Traditional futures contracts have an expiration date. Upon expiration, traders either receive the underlying asset (oil, wheat, pork) or close their position and open a new one, incurring a fee each time. Shiller asked an obvious question: if almost no one trading pork futures actually wants pork, why force an expiration date?

Traditional markets already had viable solutions, and there was no reason to change. In 2016, a crypto exchange called BitMEX saw the reason. Since then, perpetual contracts have become the mainstream trading method in the crypto market. Contracts never expire. Traders can use high leverage positions, typically 10 or 20 times their principal. The fees and liquidation costs they generate have made centralized crypto exchanges among the most profitable companies in the industry.

By the end of 2022, no one had created a usable decentralized version. The reason was the underlying technology. In most modern markets, transactions run through order books. Buyers quote prices they're willing to pay, sellers quote prices they're willing to accept, and transactions are completed when they match. The more market participants, the smaller the bid-ask spread. This is roughly how it works, from the New York Stock Exchange to Binance. But order books do more than just process transactions. They also have to keep up with the constant stream of price updates from traders—often adjusted many times before a transaction is executed. Existing blockchains don't do this well. They're too slow, too expensive, and too cumbersome. Every update costs money and requires confirmation. Running an order book on top of that is like trying to run the New York Stock Exchange with dial-up internet.

By the end of 2022, Yan and his team had reviewed every blockchain used by other projects, and none came close to their needs. So they built their own. Three months later, Hyperliquid had a self-developed blockchain capable of supporting exchanges. Yan then spent the next six months on Twitter defending what Hyperliquid offered, explaining why it was better than existing industry solutions.

The problem with exchanges is that they're useless when no one's using them. A buyer arriving in an empty market can't find a seller. The traditional approach is to hire market makers, ensuring a counterparty for everyone. You pay them with cash, equity, or token splits. Several market makers approached Hyperliquid. One told iliensinc directly that his company was kingmakers. "If you don't pay us, you'll never get started."

They didn't give it to anyone. Hyperliquid launched at the end of February 2023, and its user base in March and April consisted mainly of NFT collectors who had never traded perpetual contracts before, placing $10 orders and learning leverage through demo trading competitions. There were no serious users.

Then, in May, Yan put the strategies that made Chameleon one of the most successful anonymous trading operations in the crypto space into an on-chain vault called HLP (Hyperliquidity Provider). You can deposit $10 or $10 million. There are no management fees, no performance-based commissions. The vault runs automated strategies, and every dollar of profit flows to the depositor. All accounts are on the blockchain. If you deposit $10, you can watch it grow in real time. If FTX is built like this, Alameda's holes will be visible to the whole world.

HLP kills two birds with one stone. Exchanges gain liquidity. Users providing liquidity gain something traditional finance has never offered. An early Hyperliquid user described it as, for the first time in history, an ordinary person can invest in a high-frequency trading strategy with zero fees.

"I was willing to pay Jeff a 2% management fee plus a 50% commission to get involved in this," they told me. "And what happened? An unknown nobody with no connections, sitting anywhere in the world, gained access to one of the best market-making strategies in the crypto space. People still don't understand how special that is."

But few understood at the time. By fall, crypto prices were soaring daily, and depositors watched their HLP balances dwindle while Bitcoin climbed. The algorithm was doing its job, and the trades themselves were profitable, but because everything was on-chain, it couldn't hedge its exposure to the broader market. Traditional market makers would offset this risk in another market. HLP wasn't designed to do that. So while it won trade after trade, it was effectively shorting a continuously rising market. People were furious. Other projects attacked Hyperliquid on Twitter and Discord, and Yan retaliated. It was early enough that he still took it to heart.

But HLP wasn't the final answer. Yan created it to funnel liquidity before independent market makers arrived; he saw the opportunity as obvious to them. Demand exceeded supply, and wide spreads meant anyone willing to quote could easily make money. He wrote documentation, posted long threads on Twitter explaining market making, and guided market makers through the onboarding process. Most hesitated. Every other exchange offered them money. Yan refused, and HLP couldn't scale to fill the gap. "Alameda is the core of FTX's operation," he said. "We didn't want HLP to become the core of Hyperliquid's operation."

Data is rising, and so are complaints. Theoretically, market makers should arrive at any time. But if they don't, and users leave first, then it's all over.

However, there is one group of people you can always count on – venture capitalists.

Their analysts were using the exchange themselves, quietly using their private time to approach the partners one by one, saying: "This thing is really good." The partners picked up the phone. Yan and iliensinc didn't do any promotion, no pitch deck. The protocol generated transaction fees, but Yan insisted from the beginning that the team wouldn't take a single penny. VCs called to ask if there was a deck, and Yan and iliensinc chatted, eventually realizing that there really wasn't.

By January 2024, the fund started visiting him in person. iliensinc was familiar with the process; she had experience as an investor. She began going over the terms and conditions Yan needed to know and be aware of. After about two weeks, he was cooperative. "It felt almost natural," he told me, "The VCs were contacting me; it was probably time to raise funds."

His only condition was that he would only consider letters of intent from investors valuing the company at $1 billion. Hyperliquid had been online for less than a year. The team was burning through hundreds of thousands of dollars every month, all from Yan's personal savings. When an investor met his conditions, Yan spent a weekend thinking it over.

He asked entrepreneurs and VCs what they were really after when they raised funds. But they couldn't convince him that their money was more valuable than the money itself. At one point, he said that refusing felt right. Once he felt right, there was nothing more to say.

On Monday morning, he told iliensinc, "We're not taking it."

"What the fuck?"

She couldn't believe it. She was the one managing the money, watching it burn every day. Now a fund had offered about $100 million, which he rejected after two weeks of preparation. The rest of the team couldn't accept it either.

He called the fund and declined. They didn't believe him either. They thought he must be accepting someone else's terms. No. Hyperliquid isn't a company. It's a protocol, and neutrality has been a priority from day one. "If Bitcoin had taken a VC round," he said, "I really don't think it would be Bitcoin. Its entire value proposition would be destroyed." And he didn't need that money. To this day, Yan still pays for much of the team's expenses himself.

On January 28, 2024, he posted four lines on Twitter:

There are no investors. There are no paid market makers. The team does not charge commissions. There are no insiders.

Token airdrop: The largest wealth transfer in crypto history

Hyperliquid holds only one meeting a day, a morning stand-up meeting. I observed one on my second day in Singapore. The team was crammed around an engineer's screen. A dragon-shaped plush toy was perched on the monitor. They were testing a new feature called "combined margin," and the conversation mainly revolved around potential problems. There were also long stretches that weren't really conversations. Yan would cross his arms, look down, and stare at his bare feet. The engineer next to him would do the same. These silences weren't awkward, nor were they brief; no one in the room found it strange.

Part of the reason is personality. The team is young, between 24 and 31 years old, and almost all of them are extremely intelligent introverts. But when I asked Yan if he reads books in his spare time, he implied that it wasn't just shyness.

"I read far fewer books than conventional wisdom would suggest," he said with a laugh, wearing dark-rimmed glasses. "To read a book and have it permanently change you is extremely time-consuming. The return on investment isn't great."

He shifted his jaw—a habit I later recognized—like adjusting ear pressure on a plane. One particular risk of writing about young tech people is that sooner or later they'll tell you they don't read. I appreciated Yan adding that he reads about a book every two months and looks forward to one day sitting down and reading all the ones he hasn't yet. He then went on to explain why reading more books is so important, and so on.

“If you’re not the first person to do something,” he said, “it’s probably not worth your time. I genuinely believe that. If you operate on that assumption, then reading isn’t very helpful. If there are already useful references about what you’re doing, it’s very likely that it’s already been done. If it’s already been done, why would you do it?”

In late 2023, Hyperliquid faced yet another problem with a ready-made script in the crypto industry. Yan, as usual, wasn't interested in that script. A crypto project's token grants holders a stake in the project's success. Deciding who gets the tokens first, and on what terms, is typically done through a points program. The project announces that users can earn points by using the platform. Users assume these points will later be converted into tokens. Then they flock in, hoping to accumulate as many tokens as possible before the conversion.

The problem is that most of the influx of users aren't actually users. They're professional teams that reverse-engineer formulas, run automated strategies to maximize rewards, and then leave. The real users—those the points program is supposed to serve—are left with the scraps.

Hyperliquid's version launched on November 1, 2023. Users trade on the platform, accumulating points weekly, but the formula wasn't published. Nobody knew how it worked. Every Friday, iliensinc announced the week's points, and a ritual formed around it. Users would watch their IDs appear "typing" on Discord, then gather to compare how much they'd received, share screenshots, and build theories about how the system worked. "Rewarding real users is key," Yan said. "It's hard to define what a real user is, but Hyperliquid's points program probably reduced the percentage of bots from 99% to 20%."

Around the same time, those market makers Yan refused to pay outright began to show up. One of them—one of the largest market makers on Binance—was wary of new platforms after FTX. But he had acquaintances who spoke highly of Yan, and he first met Yan and iliensinc at a conference in Singapore in September 2023. “Jeff is ambitious but not arrogant,” the market maker told me. “He described what he was doing very tactfully, and all the criteria were met.” He walked out and messaged the team: We should integrate. They went live two weeks later.

What this market maker saw after connecting confirmed what the user discovered. The meticulous design of the infrastructure is something only traders notice. Hyperliquid has a built-in "speed bump" mechanism that makes it harder for even the most aggressive quantitative firms to outmaneuver other market makers. This feature was later copied by the entire industry. The effect is that market makers can demonstrate deeper liquidity without needing to achieve extreme latency to survive. Yan actually chose to sacrifice some trading volume—the kind that comes from institutional investors outmaneuvering each other—in exchange for better prices for ordinary users. This trade-off reduced Hyperliquid's own revenue.

During the same conference, Token2049, Yan and iliensinc decided to relocate. With the regulatory outlook for crypto derivatives in the US uncertain, Yan told me that setting up in the US felt like an unnecessary risk. A lawyer I interviewed described that period as US regulators "trying everything they could to drive the technology out of the country." iliensinc considered Hong Kong, Switzerland, and Singapore, ultimately choosing Singapore. Modern, secure, and undisturbed.

In the spring of 2024, the team moved here. This city suited Yan perfectly; it was boring. He had only two modes: work and fitness. He swam, ran, did whatever exhausted him without getting injured. This principle stemmed from a scooter accident in Puerto Rico—leaving a scar on his face that kept him from touching a keyboard for a week. Exercise was to clear his mind so he could continue coding. His only leisure time was Sunday mornings. The rest of the week belonged to Hyperliquid. He even cut his own hair because going to the barbershop took too much time.

He didn't see anything abnormal about it, or rather, he felt that most people were unhealthily lax in their work attitude. "I think people are generally too soft," he said. "The brain is an organ. If you need to work more hours, you can train it."

He had learned not to impose this system on the team. They ate lunch together every day, a family-style meal, around a black wooden table. Thursdays were for Chipotle. There are no Chipotle restaurants in Singapore, so they gave the recipes to the chef, who now made them for them. The lunch conversations usually veered into what the team had been watching or listening to lately. At this point, Yan would often fall silent, seemingly lost in thought—and he probably really was.

By the spring of 2024, Hyperliquid's perpetual contract daily trading volume had exceeded $1 billion, and the infrastructure began to creak under user pressure. One afternoon, the alarm system was triggered and kept beeping. The platform couldn't handle the influx of users. This was Hyperliquid's first outage. But everyone outside only cared about one thing: when would the Hyperliquid token arrive?

In May, Yan posted a roadmap for the next six months on Twitter. It was full of technology goals. There was no mention of tokens.

In the past few months, Hyperliquid expanded from derivatives to spot trading. The first token launched was called Purr, named after the cat. Spot trading was a necessary step—to issue Hyperliquid's token, the team needed a spot market to trade it. But it brought problems that derivatives exchanges had never faced. When trading perpetual contracts, no one needs to hold the underlying asset; you're betting on price. When trading spot, someone must hold the assets in custody. Yan didn't want to do that. The key was that users control their own assets.

To solve this problem without becoming a custodian, he realized he had to stop treating Hyperliquid as an exchange built on a blockchain and start treating it as a blockchain with an exchange built in. The chain the team built to run the exchange—already processing hundreds of thousands of orders per second—could be programmable. It would become an open system on which anyone could write code and build financial applications, just as thousands of developers were already doing on Ethereum. The difference was that Ethereum was too slow to run a decent exchange—which was precisely why Yan built his own chain in the first place.

If he opens this chain, assets can enter Hyperliquid through a decentralized bridge protected by the protocol itself, without requiring any single party to act as custodian. Anyone building on the programmable layer can access the exchange's order book and all its liquidity. A developer could build a lending platform, a stablecoin, or a mobile trading app, directly connecting to the same market where professional market makers quote billions of dollars daily.

Yan dislikes analogies. He'll tell you that Hyperliquid has no equivalent in traditional finance, and that people are more inclined to cram new things into old categories than to understand them as they are, which is a mistake. But for those of us who aren't Yan, it's like Amazon building its cloud service to power its e-commerce platform, only to find that the cloud service is bigger than e-commerce. Yan first used the phrase "Hyperliquid will carry all finance" in that Twitter post.

He had been hesitant about making this change. He told me that subconsciously he didn't want to sign the commitment. Embedding the virtual machine into Hyperliquid was a massive undertaking, and the team didn't know if they could do it. They didn't know how much work would need to be done from scratch. But at some point, he said, the answer became clear. If they didn't do it, they would spend years piecing together various components, something like Binance, something like Ethereum, but neither, and then they would regret it.

The community was in an uproar. They were expecting an airdrop. Instead, they got a tweet about infrastructure. A thousand upvoted comments quoted a Breaking Bad meme: "We had a good thing." "I hate this. You betrayed us." Users didn't want blockchain. They wanted money. Team member Xulian—who joined after a 15-minute user interview that lasted an hour and a half—beared the brunt of the anger. "Jeff was thinking about what's best for the long term," he told me. "We really don't care how something looks in the short term."

The noisy team, in iliensinc's words, eventually got tired of arguing. The team spent the next six months developing the spot market, building the programmable layer, testing on the testnet, and preparing for staking. Then, on November 29th, a Friday, HYPE arrived.

Hyperliquid airdropped 31% of its total supply to approximately 94,000 early users. There were no conditions, no lock-up period. If you used the platform and earned points, you woke up that morning with tokens in your wallet, richer than you were when you went to sleep. At the opening price, this airdrop was worth over $1 billion. At its all-time high, it reached $16 billion. This was the largest wealth transfer in cryptocurrency history, with every dollar going to a user.

The team's share was 23.8%, smaller than the community's, and it was to be unlocked over several years. On the day of the airdrop, they received nothing. The VCs received nothing either. If they wanted tokens, they had to buy them on the open market, at the same price as everyone else, on Hyperliquid—because it wasn't listed anywhere else. That's another thing you have to pay for.

That morning, Yan didn't need to explain anything on Twitter. "Woke up to a six-figure airdrop," one user wrote. Another replied, "HYPE changed my life today. Enough to live comfortably for years, help my family, and go all in on the bull market." Someone else said, "Seven-figure airdrop, thanks Jeff."

“I feel great,” Yan told me. “It’s rare for early participants to really share in the gains and gain meaningful ownership of the network.”

I asked him how he felt since then, with everything now having a publicly listed price.

"It's terrible," he said.

Jelly Jelly attack and the encirclement of major exchanges

On a Wednesday evening in late March 2025, iliensinc's computer started alarming. She was on the phone. She hung up. On the screen, the HLP—Hyperliquid's community vault—was showing a decreasing balance.

A trader tested Hyperliquid's defenses a few days ago with a small, coordinated position. Now the test is over. He opened three positions on JellyJet—a niche token with a market capitalization of approximately $15 million and a daily trading volume of $72,000. One large short and two long positions. The short position was designed to be liquidated. The trader was shorting a token he was about to pump, and when that position collapsed, someone else would be left holding the bag. It was like pulling the pin on a grenade and handing it to someone else.

HLP took over the positions. On Hyperliquid, when the order book can't absorb traders' liquidations, the community vault takes over the positions and gradually closes them out. This is a routine procedure under normal circumstances. But Jelly Jelly had almost no order book, and when HLP was stuck and unable to exit, the attackers went on a buying spree on the open market. The price surged by over 500% in less than an hour. For every point it rose, the vault's losses increased by one cent.

iliensinc stared at the screen as the losses surpassed $5 million, $8 million, $12 million. Nothing in the system could stop it. No one had ever designed such a scenario: someone using a token with a market capitalization of $15 million as a weapon.

Validators from Asia and Europe went live. Hyperliquid's blockchain is secured by about twenty validators, independent operators who verify every transaction and gain voting power by staking a large amount of HYPE as collateral. Many people were using Hyperliquid before the token existed. They could see what was happening on the same public ledger—visible to anyone in the world—and they didn't believe it was a transaction. Within minutes, everyone voted to delist Jelly Jelly, settling positions at the price before the manipulation began. Every user holding a legitimate position was unharmed. The only one who lost money was the attacker.

This incident exposes a question Hyperliquid critics have been waiting to ask: How decentralized is the system if a dozen or so validators can overturn market prices and settle contracts according to their chosen numbers? Yan didn't shy away from the question. The small validator set is intentional. A system that releases upgrades every few weeks can't coordinate a thousand participants for each upgrade. The set will grow over time, but not at the expense of the speed that brought Hyperliquid to where it is today.

“The fix took a month. It’s bad to learn from an attack instead of having someone tell you beforehand,” Yan said. Hyperliquid—which has never paid market makers and whose team has never received fees—was willing to pay up to a million dollars for a vulnerability report. “But these people clearly weren’t trying to inform us of the problem. They wanted to exploit it.”

At the time of the attack, Binance and OKX—two of the world's largest centralized exchanges—had listed Jelly Jelly perpetual contracts on their platforms. A Twitter user tagged Binance co-CEO He Yi, urging her to list the token. "If you list Jelly Jelly," they wrote, "Hyperliquid might be finished." He Yi replied in Chinese: "Okay, received."

This is the reward of ambition. You leave a beach in Puerto Rico where nobody knows you. You build from scratch with a television and your own savings. You turn down $100 million. You give billions to strangers. What do you get?

war.

In 2023 and 2024, Hyperliquid was too small to be noticed. The airdrop changed everything. A market cap of $4.2 billion, then $9 billion, and then more—meaning every major player in the crypto industry could see a future where Hyperliquid would steal their lunch. Binance announced its own decentralized exchange. Coinbase and Robinhood started offering futures products. New protocols targeting Hyperliquid launched one after another. Then someone followed Yan in the elevator of his apartment building.

This may seem insignificant, but violent attacks against crypto holders nearly doubled in 2025. In France, the co-founder of a hardware wallet company had a finger sawed off, and photos were sent to his business partners as ransom. A family in Canada was waterboarded. Crypto transfers are instantaneous, irreversible, and require no bank approval. Someone with a wrench and a wallet address can wipe out a fortune.

Yan moved to a safer place, hired bodyguards, and was, in a way, trapped in the safest island city on Earth. Two private security guards accompanied her when she traveled. iliensinc began testing the team: what to say if a stranger asks where you work. This is why almost everyone who spoke to me for this article used a pseudonym.

Stress test on October 10th

I asked Yan what the most difficult moment of 2025 would be, and he didn't mention Jelly Jelly, his competitors, or even his bodyguards. He said it was the API server.

Throughout the summer, Bitcoin surpassed $100,000, and Hyperliquid's monthly trading volume exceeded $400 billion. The servers connecting market makers and the blockchain began to strain. Too many market makers were connecting, each sending massive amounts of orders, cancellations, and updates, and the infrastructure relaying all of this couldn't keep up. An order that should have been executed instantly took three seconds.

The chain itself never crashed. User funds were never at risk. But three seconds—in a market where milliseconds determine everything—was a warning sign. "If congestion occurs when things aren't extremely volatile," iliensinc said, "then it's unacceptable when that kind of event actually happens." Yan barely slept for weeks. He went to bed at 1:30 a.m. and was woken up at 3 a.m. by a ping saying things were down again. The team rewrote the servers from scratch.

On October 10th, that happened. Trump threatened to impose tariffs on 100% of Chinese imports, and within 24 hours, over $19 billion in leveraged crypto positions were liquidated—the largest such event in the industry's history. More than 1.6 million traders were caught in a self-reinforcing waterfall—forced selling drove down prices, triggering more liquidations, which in turn pushed prices even lower.

Hyperliquid experienced zero downtime and no withdrawal suspensions. The rebuilt servers held up. The subsequent fixes after Jelly Jelly held up. HLP liquidated billions of dollars in positions, making a profit of $40 million. But because every transaction on the Hyperliquid blockchain is public, anyone can count its liquidations. Other exchanges simply don't report liquidation data with the same precision. Binance only publishes one transaction per second. The data aggregators that major media outlets rely on use data provided by exchanges, and that data is misleading. Media reports claim that Hyperliquid handles more liquidations than any other exchange. It appears to be the most dangerous trading venue simply because it is the most transparent.

Three days later, while others in the crypto industry were still tallying their losses, Yan's team released an upgrade that would define the next phase of Hyperliquid: HIP-3. HIP-3 allows anyone who has staked 500,000 HYPE to deploy a new perpetual contract market on the platform, setting their own parameters, choosing a price source, and retaining half of the transaction fees.

By the end of the year—its second full year of operation—Hyperliquid had generated approximately $900 million in profit. Not a single penny went to the team. 99% was automatically converted to HYPE and burned, permanently removed from circulation, returning almost all of the platform's revenue to anyone holding the tokens.

I asked iliensinc what she thought about 2025, and she said, "It feels like we've grown up."

The End of Finance

On my last afternoon in the office, Yan and I sat at the black dining table next to the kitchen—next to those unopened bottles of whiskey—where the team ate lunch together every day. I still had some questions.

Hyperliquid has been distributing its components over the past year. HIP-3's Builder Codes allowed independent developers to build trading applications on the platform's order book and retain a portion of every transaction fee generated by their users. Paradigm co-founder Matt Huang called it "a fantastic way to franchise a user experience." These teams have earned over $70 million since October 2024.

HIP-3 went even further. Within six months of its launch, seven independent teams deployed hundreds of markets, mostly for assets unrelated to crypto: crude oil, gold, stock indices, and forex. The largest deployer, Trade[XYZ], has grown 38% weekly since October 2025, with a cumulative trading volume exceeding $130 billion and covering 192,000 traders. Markets created by independent deployers now account for half of Hyperliquid's total trading volume. In February 2026, HIP-4 was announced. Once launched, anyone can deploy options or prediction markets on the platform. HIP-3 opened Hyperliquid to any asset with a price. HIP-4 will open it to any event with a result.

The most impactful product on Hyperliquid is now being built by people who aren't on Yan's team and won't be joining in the future. I asked him what he thought about this. What should the team do, and what should be left to others?

"It's a dynamic question, and I don't think there's a single right answer," he said. "The most important perspective is philosophical. Are you building a financial super app, like Robinhood, or are you building a financial system?" He admitted he didn't know which would win. "But I think a financial system accessible to everyone is a better outcome for the world. A system built on open tracks, not owned by any single company."

"In building it, we often thought: What do we need to do to enable others to come to Hyperliquid and succeed, and own their own business? When people compete and own their own things, the system becomes more resilient and scalable."

He said the path of least resistance was to do everything in-house, keeping everything within one company. They chose the opposite path. "It's more difficult, but we care about how we get to our goal because the path determines what we ultimately build."

The founder of Trade[XYZ] told me that he feels there might come a day when nobody knows they're using Hyperliquid. "Maybe in the final state, it's just the infrastructure and liquidity layer of finance," he said, "while Interactive Brokers, Phantom, or something else, ultimately face the users. That's actually pretty great."

Huang of Paradigm invested heavily in the public market shortly after the HYPE token launched. "What's even more amazing," he told me, "is that this was done by an 11-person team." Eleven people, and with almost no AI. There are dedicated AI laptops in the office running the latest models. They're only used for exploring ideas. "We're closely monitoring the capabilities of AI," Yan said, "it's not good enough to write important code yet."

I asked Yan about the biggest shadow hanging over all of this. Hyperliquid has accumulated over $4 trillion in trading volume since 2023. It has captured 37% of the decentralized perpetual contract market. And while doing all this, users of the world's largest capital markets couldn't even touch it. Americans were kept out.

The obstacle was the Dodd-Frank Act, a law passed in the US after the 2008 financial crisis that required every derivatives transaction to go through a regulated intermediary. Ironically, Hyperliquid's public ledger already provided what regulator Dodd-Frank aimed to achieve: real-time visibility into all leverage in the system. But until the US Commodity Futures Trading Commission (CFTC) enacted new rules, Americans had no legal path to trade derivatives through decentralized protocols. Yan didn't assemble his own policy team, consistent with his philosophy. A month after my visit, the Hyperliquid Policy Center was established as an independent nonprofit, led by Jake Chervinsky—a prominent crypto lawyer with a decade of experience in the field. The Hyper Foundation, an independent organization supporting the Hyperliquid ecosystem, donated 1 million HYPE tokens (worth $28 million) as seed funding.

Yan acknowledged that Hyperliquid had grown so large that the "build it and ignore everything else" strategy was no longer viable. "There are people lobbying against it," he told me. "I can't say with much confidence what the final outcome will be. But regulation ultimately reflects the will of the people, and I'm optimistic about the future."

Go and the endgame

There's a question I've been pondering for a whole week: You don't really think Hyperliquid can handle all of finance, do you?

He laughed. For someone who cuts his own hair, he laughed more than you'd expect. "I mean, 'all' is a bit of an exaggeration," he said. "That's our vision. But it's very difficult to achieve, and decades-long goals sound arrogant."

"That's the difference between Go and chess," he continued. "In chess, the stronger you are, the more moves you can calculate. In Go, there are far too many possibilities. The focus is more on developing an intuition for the next move, rather than trying to exhaust the entire search tree."

He sensed I wanted to hear more, so he changed his approach. He always tried to live by this principle: be very certain you're heading in the right direction, and do your best in the present moment, but don't necessarily know exactly where you're going.

That Friday evening, the team went to a Chinese restaurant in a hotel. The engineer who had colonized the entire office with plush toys couldn't make it. Everyone else was there, including me. We were led through a quiet lobby, down a corridor, to a private room decorated with dark wood paneling, featuring a carved screen and a round table. Behind a partition at the far end of the room, several armchairs surrounded a coffee table. We sat there and had tea first.

The room was cold, the air conditioning set as if for an even hotter night. Someone handed the youngest engineer a blanket. He draped it over his shoulders and noticed it said Christian Dior. So he and Yan started talking about luxury brands—neither of them clearly knew anything about the subject. One of them pronounced LVMH as "LHVM." Neither corrected the other. iliensinc, wearing her Ralph Lauren hat, sighed.

As we moved to our table, the rotating tray began to spin and never stopped. Dishes were served one after another until a large blue-and-white porcelain bowl arrived, at which point the table fell silent. The bowl contained a shallow layer of water covering pebbles and small leaves—a miniature koi pond. A bowl of white, scallop-shaped noodles sat in the center, with three small orange fish swimming in circles in the moat between the two bowls. The waiter explained the dish to us. These fish, he said, only work for five minutes after a 30-day break. We watched them swim around and around. Then they were taken away, beginning another month-long vacation.

We left around 9:15, walking into the light rain. I said goodbye and got into a taxi to the airport. A few minutes from the hotel, the taxi began to climb a left bend on the highway, and as it rounded the corner, the financial district came into view: HSBC, JP Morgan, Standard Chartered, Deutsche Bank, Citi, their signs gleaming against the black sky. Then the road straightened eastward, and they receded one by one in the rearview mirror until only the wet pavement remained. Yan drove in the opposite direction back to work, where his bodyguard was waiting.