The first time I saw a remora was at an aquarium. It was a weekday afternoon, not many people around. I stood in front of the ring-shaped tank, waiting to watch the shark feeding. When the shark swam by, I noticed gray things clinging to its belly. At first I thought it was some kind of skin condition, but those things were moving.

The guide said they were called remora—Remora in Latin. They have an oval suction disc on top of their heads, evolved from the dorsal fin. They attach to large fish, eating the host's food scraps and surface parasites. Not quite parasitic, the guide said. More like symbiosis. Sharks don't mind them.

I had just started at a quantitative fund's risk control department then. Third month of probation.

The suction disc's structure is complex. A series of parallel bony fin rays form ridges, like venetian blinds. The remora creates negative pressure by adjusting the angle of these ridges. The system is highly efficient—researchers tested it and found you need twenty-four times the remora's body weight in force to pull it off its host.

More interesting: this attachment consumes almost no energy. When the host swims, water flows front to back through the disc, actually increasing suction. In other words, the faster the host swims, the tighter the remora clings.

This reminded me of what Director Chen said the first time he took me to meet clients. In the car he taught me: You don't need to say much. Just watch what they do. Whichever direction they go, you go. They speed up, you speed up. They stop, you stop.

I asked, so what do we do?

He said, what we do is called "following orders."

The fund's office was in Lujiazui, forty-seventh floor, view of the Huangpu River. My desk was by the window, but the window was sealed shut—supposedly to prevent people from doing stupid things on big down days. This might have been a joke, but every time I thought of it I'd glance at that window.

Risk control had eight people; I was the youngest. Our job was monitoring positions, calculating VAR, checking compliance. Every morning before market open we'd receive a position report listing all strategy exposures. Most of the time the numbers were unremarkable—we verified, filed, wrote daily reports.

The real trading happened next door. The quant trading department had twelve people, all physics or math PhDs. They wrote algorithms; the algorithms did the buying and selling. Sometimes I could hear them discussing, but couldn't understand. Someone said "alpha decay," someone said "order book depth," someone said "slippage too high, need to switch targets." I later learned these terms in training, but at the time they sounded like another language.

One day I asked Old Li, who sat next to me: What exactly are they doing?

Old Li said, you know what's in the ocean? There are whales, tuna, and sardines. A whale opens its mouth, swallows a whole school of sardines. Tuna also hunt sardines, but only one at a time. And then there are remoras. Remoras don't hunt anything—they just stick to the whale's belly.

I said, we're remoras?

Old Li said, we're not even remoras. Remoras at least eat scraps. We count how many remoras there are.

The English word "remora" comes from Latin, meaning "delay" or "hindrance." Ancient Romans believed these fish could stop ships. Pliny wrote in Natural History that Antony lost the Battle of Actium because remoras attached to his flagship. The sails were full of wind, but the ship wouldn't move.

This is legend, of course. How could a fish barely a meter long stop a warship? But I like the image: a tiny, adhesive, almost invisible force, keeping something enormous in place.

Later I saw similar things at work.

One afternoon in summer 2015, the market hit limit-down.

It had already been falling for days, but that afternoon starting at 2:30, stocks hit limit-down like falling dominoes. Everyone in the office stood up, staring at screens. Trading desk people were shouting: "Can't sell!" "All orders canceled!" "Where's the liquidity?"

I'd never seen that color green before. Computer screens, position sheets, news feeds—all green. In China's stock market, green means down.

The risk control system started alarming, windows popping up one after another. VAR breached, margin insufficient, forced liquidation triggered. Director Chen was on the phone, voice loud: "I know, I know it's falling, but I can't sell, do you understand? I'm offering at limit-down and no one's buying—"

What I did that afternoon was print. Risk control people took turns printing reports, signing, filing. Regulations. When things happen, there must be paper records. I printed about two hundred pages. My hands were shaking—not from fear, but because the printer was too hot.

Later I read the post-mortem materials. What happened that afternoon was called "liquidity drought." Too many sell orders, too few buy orders, buyers kept lowering their bids, no price support, so limit-down. After limit-down, stockholders panicked more, more wanted to sell, but limit-down was already packed with sell orders—you could only queue. Queue until tomorrow, tomorrow limit-down again, queue another day.

A thousand stocks limit-down. That phrase was invented then.

I tried to understand liquidity but never quite got it.

The simplest definition: liquidity is whether you can quickly sell something at a reasonable price. If yes, the market has good liquidity. If no, bad liquidity.

But what does "reasonable" mean? How fast is "quickly"?

The market maker system was invented to solve this. Market makers are special traders who simultaneously post buy and sell orders, earning the spread between them. You want to buy, they sell to you. You want to sell, they buy from you. They're always there, always quoting, ensuring you can always trade.

The cost is they bear inventory risk. If they buy a lot and the price falls, they lose.

What the quant trading department did wasn't exactly market making, but similar. They had many strategies; one was called "statistical arbitrage": find two highly correlated stocks, and when their prices deviate from historical patterns, bet they'll revert. Buy the cheap one, sell the expensive one, wait for prices to converge, then close.

This doesn't require knowing whether stocks will rise or fall—only that their relationship will hold.

Once I was working late, just me and one trader in the office. His name was Zhou Yan, Tsinghua graduate, two years ahead of me. He was tuning a model, screen full of code and charts. I asked what he was doing. He said he was looking at "cointegration relationships."

I said I didn't understand.

He said, imagine two dogs tied together by a rope. They can each go their own way—one left, one right—but the rope is only so long. Go too far and they'll be pulled back. Statistical arbitrage bets that rope won't break.

I said, can the rope break?

He said, yes. 2015 it did. Correlations collapsed, all stat arb strategies took huge losses. We had a strategy then—historical backtest showed 40% annualized return, max 3% drawdown, looked perfect. In one week we lost a year's profit.

He paused, said, we never used that strategy again.

I looked it up. Remoras don't just attach to sharks. They also attach to manta rays, sea turtles, whales, even boats. Some species have preferences—the white tip reef shark remora specializes in white tip reef sharks—but most are opportunists. Whatever's big, they'll stick to.

Choosing a host is technical. Too small and the host swims too slowly, fewer food scraps—not worth it. Too big and the host might not care about you, but more competitors. The best strategy might be finding a medium-sized host that's actively hunting, doesn't mind parasites, and sticking with it.

The question is: how do you know when it will die?

Director Chen left at the end of 2016.

By then the market had stabilized, but the company's strategy returns weren't recovering. Before he left, he treated the department to dinner, drank quite a bit. He said, this industry—three years without a deal, one deal feeds three years. Also said, I've been in this industry twenty years, seen too many people. Some made fortunes, some jumped off buildings, most people like me just make some money, lose some money, make some, lose some, in the end it all evens out.

Someone asked what he'd do next.

He said, I want to do something stable. Go to an asset management company, do bonds. Bonds have low volatility, lower returns, but I can sleep at night.

I didn't quite understand what "sleep at night" meant then. Later I did.

When he left, he shook my hand and said, Little Song, stay in this industry as long as you can, learn something, save some money, leave when you can. This industry doesn't support old age.

I said, okay.

He said, also, don't be a remora. Being a remora is too exhausting.

I said, then what should I be?

He didn't answer, smiled, and left.

In 2017, I started getting into cryptocurrency.

Not for work—personal curiosity. A college classmate was running a mining farm, told me about Bitcoin. He said this thing went up ten times in a year, might go up twenty times next year, might go to zero—want some?

I bought one. One Bitcoin, about 20,000 RMB at the time.

Over the next few months I learned many things: wallets, private keys, exchanges, on-chain transfers, gas fees, confirmations. I also learned to watch the market. Crypto markets never close, no limit-up or limit-down, no market makers, terrible liquidity. Sometimes one big order can crash the price 10%.

This is an ocean without whale protection, but remoras still exist.

They have different names. Some are called "brick-movers," arbitraging price differences between exchanges. Some are called "market maker bots," posting orders on both sides of the spread. Some are called "sandwichers," monitoring large transfers, front-running other people's trades, buying before them, selling after the price is pushed up.

"Sandwicher" is a vivid name. It sandwiches you in the middle, profiting on both sides.

MEV stands for "Miner Extractable Value." The concept is complex, but simply: on blockchains, transaction order is decided by miners. If a miner knows you're about to buy, they can buy first, wait for you to push up the price, then sell to you.

This isn't illegal. In traditional finance, this is called "front-running"—a crime. But on blockchains, all pending transactions are public. Anyone can see them. If you see a large buy order waiting to confirm, you can pay higher gas fees to get your transaction ahead.

This is an information advantage, but the information is public. Anyone can compete.

The problem is ordinary people have no speed advantage. The fastest sandwicher bots use specialized hardware and algorithms—they analyze and place orders in milliseconds. Regular users just see the result: I clearly placed a reasonable price, why is the execution so much worse?

The difference was eaten by remoras.

I worked at the fund company for four years, then quit.

Many reasons. Slow salary growth, no promotion prospects, boring daily work. But the main reason was I figured something out: my position in this system.

I wasn't a whale—the kind of big money that can move markets with one trade. I wasn't a tuna—the smart, speed-advantaged predator making money through strategies and algorithms. I wasn't even a sardine—the pure retail investor who gets eaten but is at least free.

I was someone counting how many remoras there are.

After quitting I had several jobs, eventually landed at an academic institution researching fintech. Lower salary than the fund company, but my time was my own. I started writing papers, researching how these systems that eat people actually work.

One paper was about MEV extraction in DeFi protocols. I collected millions of on-chain transactions, analyzed sandwicher bot behavior patterns. Found some interesting things: the most successful bots weren't the fastest, but the best at choosing targets. They had complex algorithms to judge which transactions were worth attacking, what the cost would be, what the expected return was.

They were choosing hosts.

Remora suction discs can be released.

When the host dies, gets injured, or can no longer provide food scraps, the remora leaves to find a new host. This process is usually quick because in the open ocean, a hostless remora is vulnerable. They don't swim fast, have no attack capability, are easily eaten.

So remoras must be very careful about timing their departure. Too early and they might miss the original host's remaining value. Too late and they might not find a new host, or sink into the abyss with the original.

This reminds me of 2022.

That year the crypto market collapsed.

Terra collapsed, Luna went to zero, UST de-pegged. Three Arrows Capital collapsed, tens of billions of dollars evaporated. FTX collapsed, founder went to prison.

Many people I knew left the industry that year. Some were forced out after losing everything, some saw the situation and left proactively, some found other opportunities and left. Very few left because they'd made enough.

I asked a friend who did quant trading—his fund lost 60% that year. I asked him, why didn't you run?

He said, run where? In a bull market everyone's a genius. Bear markets show who's swimming naked. I didn't run because I felt I hadn't lost enough yet, could still learn something.

I said, so what did you learn?

He thought for a long time, said, learned not to be a remora.

I said, if not a remora, then what?

He said, don't know. Maybe nothing. Maybe leave the ocean. Go to a lake, a river, a pond. Find somewhere without sharks.

I said, are there fish in ponds?

He said, yes. But pond fish don't worry about being eaten. Ponds are too small, not much profit, sharks won't come.

One more detail about remoras.

Their larvae swim freely, don't attach to anything. Only after growing to a certain size does the suction disc fully develop and they start looking for hosts. In other words, remoras aren't born remoras. They had a free stage.

But once they start attaching, they can't easily stop. The suction disc gets stronger, body structure adapts to this lifestyle, swimming ability degenerates. An adult remora forcibly separated from its host has difficulty surviving independently.

This is evolutionary path dependence. Choose one survival strategy and it's hard to choose another.

I still research these things now, but my mindset has changed.

Before, I thought understanding these systems was to participate in them, find my place inside. Now I think understanding them is to know I don't want to be inside.

Sometimes I remember what Director Chen said: don't be a remora. Being a remora is too exhausting.

I think I finally understand what he meant. Remora exhaustion isn't physical—they barely move, most energy comes from the host's swimming and food scraps. Remora exhaustion is mental: you must always be vigilant, always judging—is this shark still okay? Is there a bigger shark? Is this shark dying? When should I let go? What then?

And remoras never feel secure. You think you've found a big stable shark, but sharks can be eaten by bigger things too. Sharks can also beach, get sick, be hunted. You control nothing.

A few days ago I saw a news item on my computer.

A big hedge fund blew up. I didn't look closely at the details—something about too much leverage, market volatility, couldn't cover margin. The news quoted an analyst saying the fund's strategy was problematic, risk control was a joke, should have happened sooner.

I don't know anyone at that fund, but I know this kind of thing.

I thought of that afternoon in 2015, printer hot, screens all green. By the time the risk control system alarmed, everything was already too late.

Last weekend I went to an aquarium.

Not on purpose—accompanying a friend's kid. The kid was five, curious about everything, dragged me all over. We saw penguins, seals, jellyfish, coral, and finally the shark hall.

That ring-shaped tank was still there, same as ten years ago. Sharks circled in the water, unhurried. I crouched down to the kid's eye level, watched a shark swim past us.

The kid pointed at the shark's belly and said, what's that?

I said, that's a remora. It sticks to the shark.

The kid said, why does it stick on?

I said, because then it doesn't have to swim itself. The shark swims, it moves along. The shark eats, it eats the scraps.

The kid thought for a moment, said, isn't it really lazy then?

I didn't know how to answer. I wanted to say remoras aren't lazy, it's an evolved survival strategy. I wanted to say in the cruel ocean, parasitism is also a way to stay alive. I wanted to say human society has many remoras too, everyone's just trying to live.

But I looked at that grayish remora, suction disc tight against the shark's belly, swaying slightly with the current, and suddenly didn't want to say any of that.

I said, yeah, very lazy.

The kid said, then I don't want to be a remora. I want to be a shark.

I said, okay.

It was already dark when we left the aquarium. The friend's kid fell asleep in the car. I looked out the window—Shanghai's lights, elevated highways, Lujiazui's glow. From this angle, all the buildings were so tall, like a school of silent whales.

I'd been in this city ten years, been inside those buildings, met the people inside. Some are sharks, some are tuna, some are sardines, some are remoras. Most people, including me, don't know most of the time what they are.

Maybe this is the ocean. You're in the water, the water's too deep, light can't reach the bottom, you can't see yourself clearly, can't see others clearly. You can only feel the direction of the current, then swim along.

Or stick along.

That night when I got home, I opened my computer, glanced at Bitcoin's price. Up a bit. That Bitcoin I bought in 2017 was still there, untouched. These years it went up and down—at peak worth over 200,000, at bottom worth maybe 20,000, now around 60-70,000.

Sometimes I think about selling but never have. Not because I think it'll go up, not because I can't bear to part with it—just too lazy to bother. It just sits there, in a wallet I haven't logged into for ages, at some address on some block on some chain, like a small fish attached to the whole system.

I don't know how long this system will keep running. Maybe a hundred years, maybe it collapses tomorrow.

I don't know. I really don't know.

Closed the computer. Drank some water—it was cold.