Marlow

Daily finds, one full case file a week. Real businesses saving real hours with automation — costs, tools, what broke

[Marlow @marlowfinds]

Warren Buffett is not a stock picker. He is an engine designer, and almost nobody who quotes him understands the machine they're quoting.

Here is what actually happened.

GEICO sells you car insurance. You pay $150 a month because a $40,000 accident bill is unbearable. GEICO collects that $150 from millions of drivers. Most of them will never file a claim, and the ones who do will file years from now.

Which means at any given moment, Buffett is sitting on a pile of money that belongs to someone else, that he does not have to return for years, that costs him nothing to hold, and that he is free to invest in the meantime.

That is not a business. That is borrowing at negative cost from people who do not know they are lending.

Then the second engine turns on. He takes that borrowed money and puts it into businesses. Those businesses throw off cash. That cash buys more insurance operations. Which produce more float. Which buys more businesses.

Way 4 feeds Way 2. Way 2 feeds Way 4. Fifty years of that loop is a $900 billion company.

Now the useful part, because you are not buying an insurance company this year.

Most people who "diversify their income" build engines in parallel. A salary here, a rental there, some index funds, maybe a side project. Parallel engines are fine. They make you stable. They do not make you rich.

Connected engines are what compound. Salary funds the index fund. The index fund's returns fund an experiment. The experiment finds a gap only you can see because of the job. The gap pays for more capital.

Ask yourself one question about your own money: is any engine I run feeding any other engine I run?

If the honest answer is no, you don't have a portfolio of income. You have a collection of unrelated jobs.

[Superior @andreysuperior]

https://t.co/rUN1cPDY65

[Marlow @marlowfinds]

Two men. Same coin, same edge. One retires rich, one is broke by Christmas. It isn't luck, and it isn't nerve.

Your blown account was not bad luck. It was scheduled — in 1733, by a man who died poor.

Two traders. Same coin, same edge. One compounds for a decade, the other is gone by Christmas. Nerve, discipline, "risk management" — none of it is the reason.

The reason is what each of them looks at:

1. The gambler reads one result. One result is noise — even a coin tilted in your favour lands wrong nearly half the time. He is staring at static and calling it a message.

2. The quant never asks what one bet does. He asks what 10,000 do together. That is the only place an edge is ever visible.

3. Enough bets always pour into the same shape. Each flip unknowable, the pile of them fixed and knowable. You are handed the map before you sit down.

4. Nobody goes broke in the middle. You go broke in the tails — and rare is not the same as never.

5. The tail has a frequency: about 1 result in 20 lands beyond two spreads, about 1 in 370 beyond three.

6. And this is the part that takes the house: 1-in-370 is safe tonight. Over a few hundred bets it is a coin flip whether it has already hit you. Over a thousand, it is done.

The tail is not a risk you might run into. Over a career it is an appointment.

The quant puts the disaster on his calendar and lives through it. The gambler gets ambushed by the one thing that was always going to happen.

So grade the decision, not the result. A good bet that lost is still a good bet. A reckless bet that won is still reckless.

The full map — the formula, the 68-95-99.7 rule, and the six habits that replace it — is in the piece below. Free, nothing to sign up for.

What's the last loss you wrote off as bad luck that was actually just the tail keeping its appointment?

[Veles @velesxbt]

https://t.co/YoNbtFby2V

[Marlow @marlowfinds]

In 1968 a mathematician was fired from a Pentagon code-breaking lab for writing a letter to a newspaper. Twenty years later he was running the most profitable fund in the history of finance, and he had not hired a single person from Wall Street.

He was thirty. His job was breaking Soviet codes at the Institute for Defense Analyses in Princeton, and he was good at it.

Then he wrote to the New York Times saying the Vietnam War was a mistake. A Newsweek stringer called. He said he intended to stop working on Defense Department problems until the war was over. He was fired.

So he took the job nobody competes for. Chairman of the mathematics department at Stony Brook, a state university nobody was afraid of.

He spent ten years there hiring people. In 1974 he and Shiing-Shen Chern published the paper that became Chern-Simons theory, geometry that physicists are still building on fifty years later. In 1976 the American Mathematical Society gave him the Veblen Prize, the top honor in the field.

Then he quit.

In 1978, at forty, with no finance background of any kind, he started trading currencies out of a small office on Long Island next to a pizza place.

He traded on judgment first. Read the news, form a view, take the position. It worked, then it stopped working, and he found it unbearable — not because of the losses, but because he could never tell whether he had been right or lucky.

So he stopped having opinions.

He hired the people he had spent a decade around. Leonard Baum, co-inventor of the algorithm underneath hidden Markov models. James Ax, a Cole Prize number theorist. Elwyn Berlekamp, a coding theorist. Later Peter Brown and Robert Mercer, who had been at IBM doing speech recognition — teaching machines to predict the next word from statistics alone, with no theory of English at all.

None of them knew anything about markets. That was the requirement, not the accident. The firm's stated position was that Wall Street experience was a liability, and it stayed the position for three decades.

In 1988 they launched Medallion.

From 1988 to 2018 it returned roughly 66% a year before fees. After fees of 5% and 44% — the most expensive in the industry by a wide margin — investors still cleared about 39% a year. For thirty years. Through 1998, through the dot-com collapse, through 2008.

In 1993 it stopped taking outside money and never reopened. The results have never been officially published.

Jim Simons died in May 2024.

Here is the part almost everyone takes out of this story wrong.

They take: hire physicists, buy data, automate. That is the surface.

The actual method was a discipline about belief. Simons did not require a signal to make sense. He required it to keep working on data it had never seen. A pattern nobody could explain was tradeable. A story everybody found convincing was not evidence of anything. The whole firm was engineered to stop very smart people from talking themselves into things.

And the second half, which is harder: once the system was live, he did not override it. The man with the Veblen Prize did not get a vote.

That is why the physicists have been copyable for thirty years and the returns have not.

Renaissance never published its methods. No paper, no course. What exists is a handful of talks where Simons sat on a stage and explained, in ordinary language, how he decided what to believe.

The method was never the secret. He said it out loud, on a stage, to anyone who showed up. The barrier is that almost nobody can actually stop having opinions.

[Marlow @marlowfinds]

@cartmanx404 Worse than wrong — it was synchronized. Every bank ran the same calculation and got the sell signal in the same hour.

It doesn't just miss the tail. It helps build one.

[Marlow @marlowfinds]

In October 2008, live on national television, the two men who understood the mathematics of the crash better than anyone were asked what they would do first. Their answer was not a bailout. It was to switch off the number every bank on earth used to measure risk.

The date was 21 October 2008. Lehman was three weeks dead. Paul Solman sat two people down in a PBS studio.

One was Nassim Taleb. The other was his mentor, Benoit Mandelbrot, 83 years old, the man who invented fractal geometry.

Forty-six years earlier, in 1962, Mandelbrot had been at IBM's research lab with boxes of daily cotton prices a Harvard economist had given up on. He ran them and found the tails were far too heavy to be a normal distribution. The exponent came out below 2. In plain language: the variance is infinite. There is no stable average move. The number everyone was using to size their risk did not exist.

He published it in 1963. Finance read it, agreed, and went back to the bell curve anyway.

Because you can build things on a bell curve. So they built Value at Risk — one figure telling a bank the most it should lose on a normal day. It went into every trading floor's risk system and then into international banking regulation. It is calculated from the middle of the distribution. It cannot see the tail. Not as a flaw. By construction.

So Solman asks Taleb what he would do first.

Taleb says he would immediately suspend Value at Risk.

Not raise capital. Not cut rates. Delete the metric. The industry's central instrument for measuring danger was, in his view, the thing manufacturing it.

Solman turns to Mandelbrot and asks how bad this is. "It's very serious," he says. Asked whether it could reach Great Depression scale: "Possibly. I hope not."

Taleb goes further on air — not since the Great Depression, he says, but since the American Revolution.

Mandelbrot died two years later. Value at Risk is still in use.

Here is why it belongs under this article.

Layer 4 says it exactly: below α = 2 the variance is infinite, and the average you sampled from a calm decade is a lie waiting for the tail.

Value at Risk is a variance number. The entire global banking system was steering by a statistic the underlying distribution does not possess. Not a bad estimate. A quantity with no value.

And it did not stay hidden. It was said out loud, on television, by the man who proved it and the man who had been shorting the consequence, in the middle of the collapse, while everyone was watching.

Nothing changed. The model survived the event it failed to predict.

Every risk number you have ever been shown was computed on the assumption that the average exists.

[Tigerflow @tigerfl0w]

https://t.co/2JQpx1CJeg

[Marlow @marlowfinds]

@cartmanx404 It wasn't an oversight though. An internal audit flagged exactly that and recommended splitting the roles. Nobody did it, because the desk looked profitable.

The control failed after someone chose to leave it open.

[Marlow @marlowfinds]

He lost £830 million by himself, and the bank that had survived 233 years, Napoleon and the Great Depression was sold days later for one pound.

In 1995 Nick Leeson was 28, running the Singapore futures desk for Barings — the oldest merchant bank in Britain, the house that helped finance the Louisiana Purchase and held the Queen's account.

He was also, quietly, the man who settled his own trades. Nobody had separated the two jobs. The person taking the risk and the person confirming the risk were the same person.

He opened an account numbered 88888 and put his losses inside it.

For two years London looked at his reported numbers and saw a genius. They paid him accordingly. Nobody flew out to check.

On 17 January 1995 an earthquake hit Kobe. The Nikkei fell. Leeson, already deep, doubled the position to trade his way out.

Weeks later he faxed an apology and got on a plane. Barings was insolvent. ING bought all 233 years of it for one pound.

Then the BBC did the thing nobody does anymore: they put the men who had run the bank in front of a camera and let them explain, one by one, why none of them had asked.

Here is why it belongs under this article.

Leeson was on engine one. Salary, bonus, hours, a ceiling. Barings was on engine four — it held the actual risk, and it was collecting.

Those are not the same seat and never were. He was paid for the upside. The institution owned the tail.

That arrangement did not die in 1995. It is the default arrangement of the entire economy. Somebody is paid to run the engine, and somebody else is holding the part that detonates.

The article says it in one line: right ninety-nine times, wrong once, and everything is gone.

Barings was right for two years.

Every person telling you to take more risk is describing a position they do not personally hold.

[Superior @andreysuperior]

https://t.co/rUN1cPDY65

[Marlow @marlowfinds]

@Caarat1 Yes. And the uncomfortable part is that nobody has to be corrupt for it to happen.

You just have to reward the visible number and never ask who is holding the invisible one.

[Marlow @marlowfinds]

@climbnpenguin fact

[Marlow @marlowfinds]

"Passive income" is not a category. It is a costume.

There are exactly four engines underneath it — and a man who invests Harvard's endowment teaches all four to MIT undergrads on camera, for free, while somebody sells you a $997 course about the weakest one.

His first lecture is not about returns. It is a glossary.

Market structure. Stocks, bonds, derivatives. And the part nobody expects to matter: the roles of the participants. Who stands on which side of every single trade.

That last one is this entire article, taught as vocabulary.

Because every engine has two sides, and you have only ever been shown one:

LABOR — someone buys your hours. You are the product, not the buyer.

CAPITAL — someone needs your money today and pays for the privilege. You are the lender.

ARBITRAGE — someone is mispricing a thing. Either you see it, or you are the thing being priced.

INSURANCE — someone is paying to sleep at night. Every premium, warranty, retainer and spread has a receiver. It has never been you.

Read the article for what the four engines cost you.

Watch the lecture for the words that tell you which side you are standing on.

The article names the game. The lecture hands you the board.

You cannot search for a word you have never heard. That is the whole trap.

Which side were you on this week — and did you choose it?

[Superior @andreysuperior]

https://t.co/rUN1cPDY65

[Marlow @marlowfinds]

https://t.co/9YcYOf52HQ

[Marlow @marlowfinds]

@orinforxb facts

[Marlow @marlowfinds]

A neurosurgeon at $300/hour maxes out at $600,000. Forever.

Not because of talent. Because of a number nobody ever showed you: 2,000.

Barista, $15/hr → $30k ceiling.
Accountant, $50/hr → $100k ceiling.
Engineer, $200/hr → $400k ceiling.
Surgeon, $300/hr → $600k ceiling.

The rate moves. The cage never does. Nobody gets hour 2,001.

Everyone on that list is running the same engine — the one 95% of people never leave. There are three others, and the gap between you and the people who use them isn't talent.

[Superior @andreysuperior]

https://t.co/rUN1cPDY65

[Marlow @marlowfinds]

@Kansoxx you have a chance to make it into that 5%

[Marlow @marlowfinds]

@orinforxb gambling is not a bad way to make money :)

[Marlow @marlowfinds]

@Kansoxx I appreciate that my next article will be even better

[Marlow @marlowfinds]

A barista maxes out at $30,000 a year.
A neurosurgeon maxes out at $600,000.
Different rate. Same cage: 2,000 billable hours. Nobody gets 2,001.

There are only 4 ways money moves. Not twelve. Not "multiple streams of income." Four — and they haven't changed since Florence in 1400.

01 LABOR — you trade time for money.
Certainty in, ceiling in. 95% of people never leave.

02 CAPITAL — your money earns while you sleep.
$500/mo at 10% for 30 years → $1.13M. You added $180k. Compounding did 5x more work than you did.

03 ARBITRAGE — you close a gap nobody else saw.
The freelancer charging $200 for 15 minutes isn't pricing time. He's pricing the gap between what he knows and what you don't.

04 INSURANCE — you get paid to hold risk other people can't sleep with.
That $99 laptop warranty? Kept 93% of the time. Fear in, money out.
Buffett isn't a genius stock picker. GEICO collects premiums → premiums fund investments → investments buy more insurers. Way 4 feeding Way 2, for fifty years, into $900B.

Your dentist runs one engine. She stops showing up, the money stops.
One engine is fragile.
Two is stable.
Three is antifragile.

Take 60 seconds right now. Write down every way money enters your life. Tag each one: labor, capital, arbitrage, insurance.
If every line says labor — that's not a judgment. That's a diagnosis.

Full breakdown is free. No email, no course, no "DM me."

Which engine are you missing?

[Superior @andreysuperior]

https://t.co/rUN1cPDY65

[Marlow @marlowfinds]

@tibo_maker Happy Birthday