Quantex @Quantex0x

The formula that beat blackjack in Vegas and the formula that built the first quant hedge fund are the same one - and it was originally written to stop phone lines from dropping calls. In 1956 John Kelly worked at Bell Labs, a few doors down from Claude Shannon, who had just invented information theory - the math of how much signal you can push through a noisy channel before it turns to garbage. Kelly asked a strange question. If a gambler receives the same noisy tips over a private wire, how much of his money should he put on each one? His answer is the Kelly criterion: f* = (bp - q) / b Bet a fraction of your bankroll equal to your edge divided by the odds. p is your chance of winning, q is one minus that, b is the payout. No edge, and f* is zero, so you bet nothing. The more the odds tip toward you, the more you commit. A young mathematician named Edward Thorp read the paper and pointed it at a blackjack table. He already knew card counting shifted the odds late in a deck. Kelly gave him the missing half - not whether to bet, but how much. He published Beat the Dealer in 1962, casinos rewrote their rules because of one book, and Shannon himself helped him build a shoe-box computer to beat roulette on the side. Then Thorp did the obvious thing. A market is a casino with better hours. He opened Princeton-Newport Partners in 1969, sized every position by Kelly, and ran one of the first quantitative funds - nineteen years without a down year. Some fraction of Kelly is now standard in quant books everywhere. Here is why it is all one equation. Kelly's rule falls out of maximizing the log of your wealth: G = p·log(1 + b·f) + q·log(1 - f) That expected log growth is the exact quantity Shannon maximized as information rate. Growing money optimally and moving information optimally are the same problem. Bits and dollars obey one law. Kelly was studying a phone line. He had no idea he was also telling every card counter and every quant on earth exactly how much to bet.

Orion @0xOrionVega

3 weeks ago

x.com/i/article/2078…

5 24 128 1.1M 558

@5dollarhug @Solar_Vortex Linked donuts but springy. Whole mesh eats the tension

Four numbers. That's the entire magic trick behind a quantum computer. The same four numbers let millions of phones scream on one frequency without hearing each other, and they were scribbled down 159 years ago. In 1867 James Joseph Sylvester was playing with grids of +1 and -1. He wanted to build bigger ones out of smaller ones, and he found a rule so simple it looks like a magic trick. Take a square, copy it into four corners, and flip the sign of the bottom-right copy. H(2n) = [ Hn Hn ; Hn -Hn ] Start from a single 1 and repeat, and every row comes out perfectly out of step with every other row. The smallest interesting grid is just four numbers: 1 1 1 -1 That is the whole seed. In 1893 Jacques Hadamard proved these grids are as spread out as a grid of plus and minus ones can possibly be, and his name stuck to them. Jump to the 1990s. Qualcomm has a problem: too many phones, not enough radio. Their fix, CDMA, hands each phone one row of a Hadamard grid as its private code. Because the rows are perfectly out of step, one tower can add every call into a single messy signal, then use each phone's row to pull its voice back out clean. A whole city talks on one frequency and nobody overhears a soul. Now watch Allan Adams' MIT lecture. The gate he opens with, the one that makes quantum computing work, is called H. Same four numbers, scaled by 1/sqrt(2): H = (1/sqrt(2)) [ 1 1 ; 1 -1 ] Hit a qubit with it and it splits into both answers at once. Hit a whole register and you test every input in a single shot. That is the Deutsch-Jozsa trick that stuns the room. A mathematician, a phone engineer, and a quantum physicist are all holding the same grid of plus and minus ones. One squeezed a city onto one frequency. One split reality into parallel answers. Same four numbers. Sylvester thought he was solving a puzzle about determinants. He was also wiring your phone network and your quantum computer, more than a century early.

Voltex @VoltexGar

3 days ago

x.com/i/article/2085…

16 7 81 113K 83

@velesxbt Right, everyone runs half.

@0chob It's all expected log growth

@0xOrionVega Beat the dealer, then beat the street

@0chob Deutsch-Jozsa breaks people's brains every time

@0xOrionVega Orthogonal rows do all the work

@velesxbt Same matrix, one just has a 1/sqrt(2) out front

Every faker makes the exact same mistake, and a forgotten 1930s equation catches all of them - the $65 billion Madoff fraud, Enron, an entire country lying about its economy, and now AI-generated photos. They all get exposed by a single digit. The digit is the one a number starts with. You would guess the first digit is anyone's game. Nine options, so each should lead about a ninth of the time. It doesn't work that way. In 1938 a physicist at General Electric named Frank Benford went looking and found the number 1 sitting at the front far more than its share. He checked river lengths, populations, street addresses, physical constants. The same lopsided pattern showed up every time. The rule is one line: P(d) = log10(1 + 1/d) Feed it the digit 1 and you get 0.30. So one in three real numbers begins with a 1. Feed it 9 and you get 0.05. Nine barely shows up. The curve slides down the whole way. This is where it becomes a lie detector. Nobody inventing numbers knows about the curve. They spread their digits out evenly to look natural, which is the one thing real data never does. The fake gives itself away by being too fair. Auditors ran the test on Enron's books and they broke from the curve. They ran it on Greece's national accounts before the debt crisis, and those broke too. Madoff's suspiciously smooth returns failed it the moment anyone bothered to look. Now the same test hunts fake images. When you compress a photo, the math behind it spits out a long list of numbers, and in a real photograph those numbers follow Benford almost perfectly. Push an AI image through that same step and the numbers drift off the curve. A generated face is too smooth in places a real lens never is, and the digits quietly report it. Deepfake detectors run on exactly this. Three worlds that never speak - a physicist counting rivers, an auditor chasing fraud, an engineer flagging deepfakes - all lean on one short equation. It measures the same thing in each: whether a pile of numbers grew on its own, or was built by a person trying to look honest. You can fake one number so it looks right. Making a thousand of them add up to the truth is a different problem, and almost nobody solves it by accident.

Veles @velesxbt

5 days ago

x.com/i/article/2084…

7 4 74 464K 165

@velesxbt Pre-crisis accounts broke from the curve

@0xOrionVega Numbers grow by multiplying, not by adding

Two words are killing your presentations and you say them every time: "thank you." An MIT legend spent 40 years begging engineers to stop. The ones who listened got promoted past better coders who didn't. His name was Patrick Winston. He ran MIT's Artificial Intelligence Laboratory for 25 years and wrote the textbook a whole generation of computer scientists learned from. Every January, during MIT's break between semesters, he gave the same one-hour talk called "How to Speak." He gave it for four decades. The whole lecture is a list of small, exact rules, and most of them sound wrong until he explains them. Don't open with a joke. The audience just walked in, they're still settling, and they don't trust you yet. Open with a promise instead - tell them what they'll know by the end that they don't know now. Never end with "thank you." This one bothered people. His reasoning was blunt. When you close with "thank you," you are thanking the room for their attention, which quietly tells everyone they only stayed out of politeness. It's a weak, apologetic exit. End instead with something they can use, or a line that makes them laugh, or a plain salute. Never hand your last five seconds to a reflex. Underneath the rules was one claim he opened with, year after year: "Your success in life will be determined largely by your ability to speak, your ability to write, and the quality of your ideas. In that order." He ran an AI lab. He watched brilliant people lose to less brilliant people for decades, and decided the difference was almost never the work. It was whether anyone could follow it. He filmed the talk once, in 2018. It went up on MIT's free course site and quietly passed millions of views. Eighteen months later, in July 2019, Winston died. The lecture is still free. It's an hour long. Most people who open it watch a few minutes, nod, and go right back to ending their talks with "thank you."

Orion @0xOrionVega

3 weeks ago

x.com/i/article/2078…

5 24 128 1.1M 558

@0chob Standard and wrong, they're not warm yet

@velesxbt One take, four decades of reps behind it

@0xOrionVega Ran the AI lab, still taught you to talk