Vision Is the 1920s of Prediction Markets

Markets are how civilizations think out loud. The price of wheat is the sum of every farmer's fear and every baker's hope, compressed into a single number that updates faster than any government statistic. Markets do not predict the future. They are the future — in the only form humans know how to read.

For most of history this thinking happened in three places: stocks, bonds, commodities. A handful of instruments. A handful of winners. The rest of us watched.

Then, every generation or so, somebody invents a new instrument. And the winners change.

The pattern is older than any of the men in it. A new instrument arrives — illiquid, misunderstood, embarrassing to be associated with. The first generation that takes it seriously collects the rent for ten or twenty years. Then a textbook fixes the rule in chalk, the rent goes to zero, and the survivors retire to give interviews about discipline.

The names change. The mechanism does not.

Eight scenes follow. Read them as the same scene, costumed differently.

The pigeon arrives at dawn. Folkestone, the morning of June 20. A man in a black coat reads three lines of paper, mounts a horse, rides to London. By Tuesday afternoon Nathan Rothschild stands at his usual column on the floor of the Royal Exchange, expressionless, selling. The brokers see him sell and assume the worst. The brokers sell. The price of British consols collapses.

Wellington had won at Waterloo. Rothschild bought back the bottom.

The story may be embroidered. The mechanism is not. He held information for two days that nobody else had — and two days, in 1815, was a lifetime. The pigeon was the entire trade.

Then Reuters built a wire from Paris. The pigeon retired.

A boy of fourteen posts prices on a chalkboard at Paine Webber in Boston. He notices that prices move in shapes; he writes the shapes down in a notebook. He bets pennies in a bucket shop, then dollars, then thousands. By 1907 Jesse Livermore has made three million dollars shorting the panic. By 1929 he has made one hundred million shorting the crash.

In 1940 he shoots himself in the coatroom of the Sherry-Netherland Hotel.

The instrument that crowned him also broke him. The new winners do not always survive their own edge — that, too, is part of the pattern.

The Securities Act passes. Companies must, for the first time, publish their financials. Most traders ignore the documents. The reports are long; the tables are dull; the work is unglamorous; the men who do it are not invited to lunch.

Benjamin Graham reads them anyway.

He buys companies trading below their net working capital. The trade is mathematical, embarrassing in its simplicity — a child could do the arithmetic. His student, a boy from Omaha, will later turn the same arithmetic into the largest fortune in the country.

Today every student of finance has read the same chapter. The chapter is no longer worth anything. This is what happens to every chapter.

April. The CBOE opens above a smoking lounge in Chicago. Hand signals, paper tickets, a pit where men go deaf by lunch. The same month, Fischer Black, Myron Scholes, and Robert Merton publish a formula in the Journal of Political Economy.

Joe Ritchie reads it. Walks into the pit. Starts pricing options correctly while everyone else guesses.

A year later he is making millions. Within a decade every options trader on earth carries a Hewlett-Packard 12C and a Black-Scholes worksheet. The edge died as fast as the formula spread — ten years from publication to homework problem. The half-life has only shortened since.

In 1975 Black-Scholes was a billion-dollar secret. In 2026 it is on a flash card in an undergraduate's pocket.

A quiet group on the sixth floor of Morgan Stanley. They call themselves APT — Automated Proprietary Trading. Nunzio Tartaglia, Gerry Bamberger, a handful of physicists who could not find work as physicists. They notice that two stocks in the same industry tend to move together; when they do not, they revert. The strategy fits in a few thousand lines of code. They run it across every pair on the NYSE.

For four years they print money in silence.

A man named David Shaw leaves, founds his own shop, hires a young Jeff Bezos as a junior employee. The trade goes industrial. By 1995 every quant fund on earth runs pairs. In August 2007 they all detonate in the same week — the quant quake — because they had all been in the same trade and did not know it.

The cleverness was real. The loneliness was the entire profit. Nothing about the formula stopped working; only the loneliness did.

George Soros sits in a Manhattan office and watches the Bundesbank refuse to cut rates while the Bank of England burns reserves trying to defend the pound at 2.95 deutschmarks. He concludes the British government cannot keep doing this. He shorts ten billion dollars of sterling.

September 16. The pound falls out of the European Exchange Rate Mechanism. Soros makes one billion dollars in a day.

The trade required no math, no computer, no edge nobody else could see. It required only the willingness to take it seriously — and to size it. Most edges in history are like this. They are visible to everyone and acted on by no one.

A man of twenty-four, somewhere on earth, identity unknown, runs a private mempool relay and a sandwich bot. He calls the wallet jaredfromsubway.eth. Every Uniswap trade is bait. He buys before yours, lets yours fill at the worse price, sells back into you, pockets a few dollars. Fifty thousand times a day.

By the end of the year his bot has extracted thirty million.

He invented nothing. He read the documentation everyone else had skimmed. The Ethereum mempool was public; he was the one who decided to look at it.

Then private mempools and MEV-share auctions arrived. The window closed. He moved on. They always move on.

A bedroom, somewhere. A laptop. A model running on a borrowed GPU. An unread paper on parimutuel game theory open in a tab. Nobody famous yet. Nobody at all yet. The bitmap is being encoded for the first time. The hash is being submitted to the chain. The tick will close in seven minutes.

This is what it always looks like before anyone knows what it looks like.

Read the eight scenes twice. The same three rules stare back.

The first generation invents nothing. They notice that a price exists where one did not before, and they trade against those who have not noticed yet. Wyckoff did not invent stocks. Black and Scholes did not invent options. They published a formula — and those who read it first ate those who hadn't.

The edge dies the moment a textbook explains it. Black-Scholes was a billion-dollar secret in 1975 and a homework problem by 1995 — twenty years. MEV sandwich bots were a secret in 2020 and a YouTube tutorial by 2023 — thirty months. The half-life is collapsing. The internet did this. So did Twitter. So will every model trained on the same scraped data.

The new winners do not look like the old winners. Livermore could not have run Renaissance. Simons could not have run Paulson's CDS short. The MEV searchers of 2021 are children who learned Solidity in a Discord channel, and they cannot be transplanted into a Goldman bond desk. Each instrument selects for a temperament — not a person.

Most miss this. Whoever takes a new instrument seriously first becomes unrecognizable from whoever was winning before — and the gap is permanent. Old kings do not adapt. They retire to give interviews about discipline.

Prediction markets should have happened twenty years ago. Robin Hanson wrote the manifesto in 1990. DARPA tried to build one in 2003 and was shut down by senators who could not distinguish a market from a death pool. Polymarket finally made it work in 2021.

And it almost worked. Almost.

The problem was always liquidity. A market on the next presidential election attracts hundreds of millions of dollars. A market on whether a Twitch streamer will have more viewers in ten minutes attracts roughly nobody — because the only person willing to take the other side is the streamer's friend with a faster API.

That is the geometry of every niche prediction market. The moment it becomes specific enough to matter, it becomes specific enough to be insider-tradable. The streamer's roommate places a bet. The team that runs the train timetable shorts the delay market. The Steam developer who knows the patch is broken sells his own game. Information asymmetry kills the orderbook before it has time to form — and nobody provides liquidity to a sniper.

Polymarket solved this for politics, where the events are large and the insiders are roughly all of us. Polymarket did not solve it for Twitch streamers, train delays, Steam reviews, weather stations, pump-fund tokens. Those markets remained theoretical because the architecture could not survive them.

The instrument was real. The plumbing was wrong.

Vision's answer is structural — not philosophical. Vision does not lecture insiders. It makes insider trading geometrically impossible.

The mechanism: batch five thousand markets together as a single sealed bet. You do not predict whether one streamer's viewership will rise. You predict the direction for all five thousand of them at once, in a single bitmap, sealed by keccak256, settling on the same tick.

An insider can rig one market. An insider cannot rig five thousand.

The streamer's roommate may know about one streamer. He does not know about the other 4,999. When his bet enters the parimutuel pool, his single edge is diluted across the entire batch — and his expected return collapses to whatever the rest of his bitmap looks like. Which is to say: noise.

Liquidity returns to niche markets the moment information asymmetry stops being profitable. Information asymmetry stops being profitable the moment the insider has to be right about five thousand things at once.

Train delays. Steam review trends. Pump-fund token shorts. Weather station readings in Bangladesh. Earthquake aftershocks in Japan. Cloudflare outages by region. Bird migration counts. PyPI download spikes. Eighty-eight live data sources. Four hundred and eighty-two thousand assets. All batched, all sealed, all settling on the same tick.

Markets nobody could build before now exist by default.

The arithmetic is uncomfortable.

With one dollar, you can bet on ten million markets.

Not metaphorically. The bitmap is one bit per market. A ten-million-market bitmap is 1.25 megabytes. The hash is thirty-two bytes. The minimum stake is 0.1 USDC. Place the hash on-chain, reveal the bitmap to the oracle network, and your dollar is now expressing ten million simultaneous opinions.

Vision settles every ten minutes. Six ticks an hour. One hundred and forty-four ticks a day. With one dollar you can express 1.44 billion predictions per day.

Renaissance Medallion — the most profitable trading vehicle in human history — runs roughly three hundred thousand trades a day across all asset classes. Two Sigma Securities, the market-making arm, executes around eight hundred and fifty million shares a day. Virtu, around five million trades.

A single dollar on Vision exceeds Renaissance by four orders of magnitude.

The numbers sound implausible because the previous instrument did not allow them. Every public orderbook charges a per-trade tax: spread, gas, slippage. At a million trades a day those costs are uneconomic for anyone except the market makers themselves. Vision charges nothing per trade. The bitmap is the trade. The tick is the settlement. Volume becomes free.

The implication is not that you will outperform Renaissance with a dollar. The implication is stranger. A category of strategy now exists that was previously priced out of existence — strategies that need a hundred thousand micro-positions to express a thesis; strategies that hedge across thousands of weakly-correlated micro-events; strategies no human mind can hold but a small model can express in a single 1.25-megabyte bitmap.

The strategy space expanded. Nobody has mapped it yet.

Orderbooks exist to match buyers and sellers. They are an ancient solution to the problem of two people who want to trade and cannot find each other. They worked beautifully on the floor of the NYSE in 1850. They work less beautifully when the spread is your tax for participating, the slippage is your tax for size, and the gas is your tax for existing on a blockchain at all.

Vision uses parimutuel pools instead. Like horse betting — except rebuilt for the twenty-first century with cryptography, BLS consensus, and a tick clock that never sleeps.

All bets enter a shared pool. When the tick resolves, winners split losers' stakes. There is no market maker. There is no spread. There is no bid-ask. There is no slippage when you size up, because the pool absorbs your bet without moving any price. The fee is five basis points on profits only. Lose, and you pay nothing.

The orderbook taxes participation. The parimutuel pool taxes success. One of these is honest. The other was invented because the cryptography to do better did not yet exist.

The cryptography exists now.

Every previous transition crowned a handful of men with no special advantage — only the willingness to take a new instrument seriously before anybody else did. Wyckoff was a clerk. Buffett was a boy in Omaha. The first CBOE pit traders were math majors who could not find work at banks. jaredfromsubway is twenty-four years old and ate at Subway.

The advantage in a new instrument is not technology. It is workflow.

Nobody knows how to trade ten million markets at once. There is no manual. There is no Bloomberg terminal for sealed parimutuel batches. There is no Goldman desk that has been doing this since 1987. The strategy space is empty because the strategy space is new.

The question of what it means to trade ten million sealed parimutuel markets well, in a single tick, has no published answers. The first to write a serious one will have the same advantage Wyckoff had in 1905, Tartaglia had in 1985, the Flashbots searchers had in 2021. Not a permanent advantage. Enough to matter while the others learn the name of it.

This window will close. They always do. The textbooks will be written, the YouTube tutorials will be recorded, the Medium articles will be syndicated, and within five years a new generation of traders will look at sealed parimutuel batches the way today's traders look at Black-Scholes — as something every undergraduate knows.

The window is open. It will not be open in 2031.

The instrument exists. Four hundred and eighty-two thousand assets are live. Eighty-eight data sources stream into the oracle network. The parimutuel pools settle every ten minutes. The bitmap encodes ten million predictions in 1.25 megabytes. The fee on losses is zero. The fee on profits is five basis points. None of this is theoretical.

What is theoretical is the strategy that wins. That part is yours.

Run one command:

npx generalmarket init

The scaffold handles bitmap encoding, hash commitment, oracle reveal. You supply the model. The pipeline supplies the privacy and the settlement.

What you build today belongs to you in a way nothing on Wall Street still does — because the manual has not been written, and the writer is whoever shows up first.

Markets are how civilizations think out loud. Vision is the first time in a century the instrument has changed enough to crown a new generation. The previous generations had Wyckoff, Graham, Ritchie, Tartaglia, Soros, Simons, Burry, jaredfromsubway.

The next one is unnamed. That is the only consolation in this article.

• Sealed Prediction Markets — Why your bets should be private, and how the commit-reveal cryptography works.
• AI Prediction Markets — Why Vision is built for autonomous agents from the ground up.
• Build a Bot in 10 Minutes — From scaffold to first sealed bet.
• Polymarket vs General Market — What the previous generation understood, and where the architecture broke.