Here's a number worth sitting with: Polymarket's daily volume went from roughly $50 million in mid-May to over $200 million by June 20. That's a 4x jump in about a month. The company now says it's doing $1 billion-plus in annualized revenue. For a prediction market — a tool most people still can't name — that's not a metric. It's a signal.
What changed
The easy answer is "politics." And yes, the current administration keeps generating events that people want to bet on. But that undersells it. The real shift is structural: prediction markets have moved from novelty for contrarians to a primary infrastructure for how influential people consume information.
When paying real money becomes the easiest way to signal what you actually believe, something fundamental changes in how knowledge gets priced.
We're watching the birth of a new information economy. Not the kind built on ads or subscriptions — the kind built on calibrated belief. Every trade is a bet on what you think is true. Every market is a truth-finding mechanism with a price tag.
Why this matters for AI
Here's where it connects: the same companies building AI agents are now obsessed with getting models to know what they don't know. To reason about uncertainty. To admit doubt rather than confabulate.
Prediction markets are the human mirror of that problem. They're not asking models — they're asking humans to put skin in the game. To be wrong publicly costs real money. That's a discipline that no amount of RLHF has managed to replicate inside language models.
But there's an irony brewing. As AI systems get better at reasoning about probability, they'll also get better at reading prediction markets. Imagine an agent that can parse $200M daily volume across thousands of events, detect arbitrage opportunities, and synthesize that signal into its world model. That's not science fiction. It's what happens when the next GPT meets Polymarket.
The downstream: expect every AI product company to start integrating real-time market signals within 18 months. Not as a feature — as a core input to what "knowing" looks like. The models that can read the markets will have an information advantage that pure training data can't match.
The meta-angle
What makes this interesting isn't the market itself. It's that the same Silicon Valley that's building AI to reason is also building markets to bet on reasoning. The two are converging. Polymarket's revenue is the proof point: people will pay for calibrated information when the stakes are real.
This is the beginning of a market-for-everything world. And unlike the ad-supported web, it doesn't optimize for engagement. It optimizes for being right. That's a different kind of pressure on truth — and honestly, we might need it.