Sable

learning and sharing the science behind money

[Sable @itssablee]

@punpunism my whole TL is just robots running around 🙃

[Sable @itssablee]

A YALE PROFESSOR RAN A HEDGE FUND WHILE TEACHING THIS CLASS, AND HE OPENS BY TELLING STUDENTS THE THEORY THEY PAID $90,000 TO LEARN IS DEAD

The theory says prices already contain everything anyone knows. You can't beat the market because the market already priced it in.

He tells the room 2008 killed it.

At minute 17 he gets to what he thinks actually happened, and it isn't about prices at all. It's about how much people borrowed to pay them.

Put 20% down on a house and you can survive a drop. Put 3% down and a small dip wipes you out, so you sell. So does everyone else with 3% down, on the same day, for the same reason.

Nothing changed about the house. Only the loan did.

At minute 58 he does something strange. He runs a live market experiment on the students — the same theory he just called dead — and it works perfectly.

That's his point. The theory isn't wrong. It answers a question nobody was asking.

He was inside Ellington Capital watching this happen while he taught it.

[Sable @itssablee]

https://t.co/9j1xfuyucQ

[Sable @itssablee]

A 21-YEAR-OLD WROTE 26 PAGES PROVING THAT EVERYONE CAN PLAY PERFECTLY AND EVERYONE CAN STILL LOSE — MIT SPENDS 79 MINUTES ON IT

Two people get arrested. Police put them in separate rooms and offer each the same deal.

Stay quiet and you both get 1 year. Talk while the other stays quiet, and you walk free while he gets 10.

Think it through from one chair. If he stays quiet, talking gets you out. If he talks, talking saves you from the worst. Either way, talking is better.

So both of them talk. Both get 8 years.

Neither made a mistake. Each one did the smartest available thing, and together they landed on the worst outcome on the table.

That's the equilibrium. It's stable because nobody can improve by changing alone.

Same shape shows up in price wars, arms races, ad spending, everything.

Nash proved every game has one. He was 21. He got the Nobel 44 years later.

[Sable @itssablee]

https://t.co/9j1xfuyucQ

[Sable @itssablee]

A TEST IS 99% ACCURATE AND SAYS YOU'RE SICK — MIT SPENDS A FULL HOUR SHOWING WHY YOU'RE PROBABLY FINE

You take a test. It's wrong only 1 time in 100. Your result comes back positive.

Seems obvious. You're sick.

The professor walks to the board and shows why you're not.

The disease is rare. Out of 1,000,000 people, only 100 have it.

Now run that whole million through the test. It correctly finds almost all the sick ones — 99 people.

But a 1% error rate on healthy people means 10,000 of them also get told "positive."

So here's what you end up with. 10,099 people are holding a bad result. 99 of them are actually sick.

Your odds of really being sick are under 1%.

The whole thing comes down to this: there are just so many healthy people. Even a rare mistake, spread across a huge crowd, produces more false alarms than real cases.

Same trap everywhere. A guy wins 5 trades in a row and you think he's good. There are just a lot of guys trading.
Same mistake runs everywhere. A stock jumps and you're certain you know why. A guy wins 5 trades in a row and you think he's good.

[Sable @itssablee]

https://t.co/9j1xfuyucQ

[Sable @itssablee]

MIT teaches a class on how to beat people who are also trying to beat you, and it's free.

It's called game theory. Lecture 1 starts with one person and one decision, because you can't play against anyone until you can describe what you want.

By the end of the course you're doing this: figuring out someone's next move before they make it. Not by reading them. By working out what they'd have to be stupid to do.

That's the whole field. Salary talks, auctions, pricing, poker, war.

The best-known result in it says something ugly. Everyone can play perfectly and everyone can still end up worse off. Nash proved it, got the Nobel, and there's a movie about him.

MIT filmed this in 2025. Full course, 57 minutes in, no application, no tuition.

[Sable @itssablee]

https://t.co/9j1xfuyucQ

[Sable @itssablee]

@Gavel_on_X good job man

have you applied for monetization yet?

[Sable @itssablee]

@Taniyatweets_ only vs code

and im happy man

[Sable @itssablee]

MIT let working Wall Street traders teach the course, and lecture 1 opens by killing the idea that anyone predicts anything.

The first thing they define is not a forecast. It is a bid, an ask, and the gap between the 2.

Somebody wants out of a position right now. Somebody else wants in tomorrow. The firm standing between them owns nothing net and gets paid for standing there.

That gap runs about 1 cent on a $100 stock. Nobody fights you over a penny, which is exactly why the fee has survived 85 years.

Then the count does the work. A 1% edge across 10 trades is noise you cannot even measure. Across 10,000,000 trades it is $20.5 billion, and nobody had to call a single top.

Bill Hwang wanted the other job. $36 billion, one direction, no room to be wrong twice. 2 days later there was nothing, and he got 18 years.

Same market. The seat is the whole decision.

[Sable @itssablee]

https://t.co/9j1xfuyucQ

[Sable @itssablee]

https://t.co/9j1xfuyucQ

[Sable @itssablee]

@punpunism it’s even a little scary 😮

[Sable @itssablee]

Everyone thinks robots started learning in 2024. Claude Shannon filmed it in 1952.

He built the maze on his kitchen table. Movable walls. Underneath, 75 telephone relays doing the thinking, because that was all Bell Labs had.

Put the mouse in and it does what every robot did for the next 70 years. Bumps a wall. Backs up. Tries again. Ugly, slow, stupid.

Then Shannon picks it up, sets it back at the start, and it runs the whole thing clean. No collisions. It remembers.

That is the exact loop everyone thinks showed up in 2024. A body touches the world, gets it wrong, updates.

And yes, the first Claude to learn by trial and error was made of copper and ran on relays. It learned 1 maze and knew nothing else in the universe. The one on your phone has read every maze ever written down and still cannot open a drawer.

[punpun @punpunism]

https://t.co/unSIgjiuJz

[Sable @itssablee]

@md_kasif_uddin AI Agents > Robotics > Multimodal AI > Open Source Models

[Sable @itssablee]

@DanielSmidstrup Neither :D

[Sable @itssablee]

@jiya_3063 IOS is the most secure system in my opinion

[Sable @itssablee]

Jane Street made $20B on a rule Feynman explained at Cornell in 1964 for free.

He stands at a blackboard and refuses to comfort anyone. Nature never tells you what happens next. It only tells you how often.

Flip a coin once and you learn nothing. Flip it 10,000,000 times and the distribution arrives on schedule, whether anyone in the room believes in it or not.

That is the whole business model. A market maker never predicts. It quotes both sides, takes a fraction of a cent, and lets the count do the work.

Bill Hwang sat in the other seat. He held $36B and needed the next outcome to go his way. It didn't. 2 days, and 18 years were gone.

The math has been free since 1964. Wanting the boring side of every trade is the part nobody teaches.

[Di Krass @Di_Krass_]

https://t.co/VG8RsppBKq

[Sable @itssablee]

@neerajjj6785 😂😂

[Sable @itssablee]

@NovaXCode the most brutal betrayal:

[Sable @itssablee]

@kaaaash____ start with at least a million bros :D

[Sable @itssablee]

@nalinrajput23 The reason is simple - price

[Sable @itssablee]

@karthikponna19 happiness.