Dex Okafor watches the number tick from 68 to 71 in three seconds. It is a Tuesday in 2036, and he is in the food court on the second floor of a Columbus mall that used to be a Sears. The smell of pretzel salt. A phone with $800 in a Kalshi account and a contract on whether the Federal Reserve will hold rates next week. The market says 71 percent chance of a hold. Dex thinks the market is right. He puts $400 on YES.
He does not know that a former Treasury analyst, two Chicago desks, and a fund out of Greenwich are already on the same contract, in sizes thousands of times his. He does not know the 71 he is reading is their answer, not the crowd's. He thinks he is reading the market. He is reading their conclusion.
Prediction markets arrived with a clean promise. Trade a yes-or-no contract on a real event: buy YES at 70 cents, collect a dollar if you're right, lose your stake if you're not. The price, the theory went, is the crowd's honest probability — because money makes people tell the truth they'd hedge in a poll. Aggregate enough bets and you forecast better than pollsters or expert panels.1
The theory mostly held. What nobody had priced in was a quieter question: who else was in the room?
For most of the century the question barely mattered, because the market barely existed. From 2006 to 2020, U.S. exchanges listed about five event contracts a year. In 2025 they listed roughly 1,600.2 Kalshi and Polymarket alone cleared more than $37 billion in volume,3 and in October 2025 alone Robinhood customers traded 2.5 billion contracts, mostly on sports.4 The crowd had arrived. The signal was live.
And the people who arrived early did the math the crowd couldn't see.
In December 2025 a Polymarket account called AlphaRaccoon went 22-for-23 on contracts tied to Google's Year in Search rankings, turning a bet on an obscure singer into a near-million-dollar haul in under a day.5 Google had briefly exposed the data. The window was minutes; the positions were already in. A month later, an anonymous trader bought about $34,000 of contracts on the fall of Venezuelan President Nicolás Maduro — days before a U.S. operation captured him — and walked off with more than $400,000.6
For a while it looked like nobody would answer for any of it. The regulator named the problem in slow motion: on March 12, 2026, the Commodity Futures Trading Commission issued an advisory and an Advance Notice of Proposed Rulemaking, comment window closing April 30.7 Senators Blumenthal and Kim introduced a bill calling prediction markets a channel for insider trading and leaked national-security information.8 Rulemaking takes years. In the gap, the market ran.
What no one in 2026 could yet see was which version of the market that gap would produce.
Nadia Ferreira ran a wealth-advisory practice outside Chicago, and she'd watched the split in her own client book. Some treated the odds the way they'd once treated bond spreads — one signal, sized to a plan. Those clients were fine. Others got hooked on the feedback. They traded Fed contracts on lunch breaks, lost $200, filed it as a near-miss, and reopened the app by two. The interface fed them streak badges and a leaderboard. It was not built for deliberation. It was built for the next visit.
That wasn't an accident. Researchers at Columbia Law documented the borrowing directly: the animations, the "hot streak" rewards, the virtual credit against paper gains all came out of mobile gaming.9 A blunter read of the business model: platforms tuned for engagement burn through their own users, because a player who blows up his account stops paying rent. "A churned user is worth zero."10 Nadia's churned clients showed up in no integrity report. They were just gone.
Here is the turn, and from 2036 it is the whole story.
The promise was that the price belonged to the crowd. Pushed to its edge, the tool flipped into its opposite: a price that belonged to whoever knew first. When one account goes 22-for-23 on outcomes it accessed early, the number on Dex's screen didn't aggregate dispersed judgment. It laundered one person's privileged access into a fee that every retail trader on the other side paid without knowing the line item existed. The instrument built to surface what the public knew became the most efficient way yet to sell the public what it didn't.
What saved it from being only that was enforcement — and, in the end, enforcement came. The CFTC had said all along it could prosecute insider trading under the Commodity Exchange Act,11 and SDNY's Jay Clayton said in early 2026 he expected cases.12 He was right. In April 2026, prosecutors charged a U.S. Army soldier over the Maduro bet — the first time the government applied insider-trading law to a prediction market.13 In May, they charged a Google engineer as AlphaRaccoon, alleging $1.2 million in profits off the Year in Search data.14 The cop, it turned out, could reach the room.
But those were the marquee cases — the ones with a leak and a name. What got decided between 2027 and 2030 was the boring infrastructure: position limits, responsible-gaming disclosures, a staffed desk that could catch manipulation before it cleared. Where that got built, the signal held and retail had a floor under it. Where it fragmented across fifty state fights with no federal spine, the price kept reflecting what connected traders already knew. From 2036 you can tell which version a market is by who its profits flow to. Dex cannot.
Prediction markets did not go away. The forecasting is useful; the access is real; a live probability you can trust is worth something. What the decade decided was only this: whether the price serves the crowd that built it, or whoever placed his bet first and was gone before the odds moved.
Dex closes the app. He cleared $180 on the Fed contract. He'll tell his brother tonight, and he'll be back tomorrow. He still doesn't know whether he is forecasting or providing liquidity — and the platform was never going to tell him. That answer got written in a comment window, a markup session, and an enforcement budget. Not by the app in his hand.
Author's Note
This is speculative journalism: the vantage of 2036 is imagined, and Dex Okafor and Nadia Ferreira are fictional composites, not real people. Everything load-bearing under them is real and sourced as of mid-2026 — the platforms, the volumes, the trades, the bill, the rulemaking, and the enforcement actions, including the soldier charged over the Maduro contract and the Google engineer charged as AlphaRaccoon. The future here is one plausible path, a caution rather than a forecast. Which version we get still depends on choices being made now about enforcement capacity, retail protection, and what "democratized finance" is allowed to mean.