I had a conversation with my dad in October. He was explaining that his IRA had finally crossed a milestone he'd been watching for years, and that the trick to getting there had been patience. Buy index funds. Hold them. Reinvest the dividends. Do it for thirty-five years.
He was right. That advice works. It worked for him. It worked for his entire generation. It will probably work for another ten years.
My generation is not going to take it.
Robinhood didn't happen in 2013 because some MBAs decided retail deserved better UX. It happened because a generation of eighteen-year-olds looked at a mutual fund statement and thought: "I am being billed to hold the same eight stocks my dad holds, except worse." Then 2020 happened, and every one of those eighteen-year-olds found out that zero-day options on Tesla paid in hours what their parents' dividend-growth portfolios paid in years. The move was complete when 0DTE contracts started outpacing monthly expiries on the S&P in aggregate notional. That wasn't speculation. That was a generation voting on what "investing" meant.
Prediction markets are what comes next. I'll explain why.
The asset is the outcome
Every asset class you know is a claim on something else. A stock is a claim on a fraction of a company. A bond is a claim on a payment schedule. A future is a claim on a price at a date. Even the weirdest assets — carbon credits, revenue shares, parametric insurance — resolve to "here is a thing you get if X happens."
Prediction markets collapse the distance. You don't buy a company and hope the market rewards the thesis. You buy the thesis directly. "Fed cuts rates 50bps in Q1" resolves yes or no. "Trump wins 2028" resolves yes or no. "Nvidia beats consensus EPS this quarter" resolves yes or no.
This does two things.
First, it makes the mental model ruthless. You are not pretending to value a stock via a DCF that was mostly vibes and calling that "analysis." You are writing down, in one sentence, exactly what you think is going to happen in the world. If the thing happens, you are right. If it doesn't, you are wrong.
Second, it makes the error bar visible. A stock can rip 8% on news that has nothing to do with your thesis, and you'll take it as proof you were right. A prediction market can't. The contract resolves to the exact statement you bet on. You were right or you weren't, and the receipts are timestamped onchain or in the exchange's ledger. Calibration is forced.
For a generation that grew up learning probability from fantasy football before they learned it from algebra, this is the natural packaging.
This is not gambling
"Gambling" is doing a lot of work in the discourse around prediction markets, and it's usually doing it sloppily.
A market is gambling when the price carries no information about the outcome. A roulette wheel is gambling. A scratch-off ticket is gambling. A Superbowl prop bet on the color of the Gatorade is gambling for almost everyone, because almost nobody has private information about the Gatorade selection. The distribution is fixed; your edge is zero; the house rakes.
A market is investing when the price aggregates information that actually predicts the outcome. This is the Hayek argument — prices carry information no single participant has in full. Wolfers and Zitzewitz wrote this up in the Journal of Economic Perspectives in 2004, and thirty years of academic literature since has shown the same pattern: on questions where you can run the experiment fairly, market-derived probabilities beat expert panels more often than not.
The stuff that determines whether a prediction market is informative or noisy is the stuff that determines it for any other market: who's trading, what they know, whether they're calibrated, whether the venue is liquid enough to punish bad prices. "Will Powell mention 'progress on inflation' more than eight times in his February press conference" is informed by people who actually read transcripts. "Will Altman return as CEO of OpenAI" in November 2023 was informed by people with source networks that mapped to San Francisco. You can beat markets like these the same way you beat stocks: research, risk-pricing, showing up before the crowd.
That is investing. The contract happens to resolve binary. That's an implementation detail.
The timing is not an accident
If you had told me in 2021 that by 2026 the US would have three CFTC-approved prediction-market DCMs doing serious monthly volume, I would have assumed you meant offshore. The actual sequence:
- Late 2023 through 2024: Kalshi fights the CFTC over political event contracts, loses at the agency, wins on appeal at the DC Circuit. Event contracts on politics become federally operable.
- November 2024: Polymarket does over $3.6 billion of volume on the US presidential election alone. The Polymarket price called the race hours before cable did.
- 2025: Kalshi sports-contract lawsuits proliferate across state lines. Most state challenges fail on commerce-clause grounds. Kalshi keeps operating.
- Late 2025 into 2026: Gemini Predictions receives DCM approval and launches with direct API access. Polymarket relaunches to US residents on a regulated domestic exchange.
Inside thirty-six months, the category went from "crypto-grey-area" to "three regulated US venues, billions of annualized volume, deep third-party integration." That's the kind of regulatory shift you miss if you assume asset classes need decades to form.
They don't. Eurodollar futures went from invented-by-the-CME-in-1981 to the most traded interest-rate instrument in the world in about ten years. ETFs went from first-launch-in-1993 to a trillion in AUM in fifteen. Once the infrastructure is legal and the use case is real, the J-curve runs itself.
Beyond sports, by a lot
I get the sports framing. Most Americans hear the phrase "prediction market" for the first time because someone on TV used it to mean sportsbook. That framing will age fast.
Take a walk through what's live on the major venues right now, and this is not exhaustive — across eleven aggregated venues you can find sixty-thousand-plus active markets on any given week:
- Macro: Fed rate decisions, unemployment prints, CPI bands, ECB decisions, JPY intervention, 10-year yield ranges.
- Politics: Senate control, gubernatorial races, cabinet appointments, legislative passage, Supreme Court rulings, diplomatic outcomes.
- Tech: frontier-model releases, OpenAI org moves, chip export controls, specific benchmark outcomes.
- Science: Nobel prizes, NASA mission outcomes, FDA drug approvals, vaccine efficacy readouts, climate data prints.
- Crypto: BTC at $X by date, stablecoin market cap thresholds, ETF approval timelines, major protocol migrations.
- Corporate: earnings beats, M&A closings, IPO timelines and pricing, layoff rounds, earnings-call mentions of specific keywords.
None of that is sports. All of it is the kind of question that an investor in 2035 will casually hold a position on the way an investor in 1995 held a few shares of Apple.
Why it democratizes
There's a thing about prediction markets that I keep forgetting to be explicit about, because it feels obvious to me, but is not obvious if you've only ever traded equities.
To trade a stock, you need to pick a company, understand its financials, value it against its peers, and then take on macro risk you didn't ask for. AAPL doesn't just trade on Apple's performance. It trades on the yen, on Chinese property, on the mood of a Federal Reserve governor at 2:00 p.m. on a Wednesday. You bought a phone company. You got a macro book.
To trade a prediction market, you pick the question. If you know something specific — the shape of a clinical trial enrollment, how a state legislature actually votes on a bill, how a model release cycle shakes out at a given lab — you can express that specific thing. Directly. Without taking on six unrelated risk factors. Without paying a 0.05% ER to an index that diversifies you into things you don't believe in.
This is why prediction markets are structurally democratizing. They reward domain expertise over capital. A kid with a good read on how TikTok ban legislation actually moves through Congress can turn that read into a position of exactly the size they can afford, on the exact question, with no middleman valuing the claim for them. The equity market does not work that way. It cannot. The structure doesn't allow it.
The coordination problem
There is exactly one reason prediction markets haven't already eaten a generational share of retail investing attention: the product is bad right now.
Eleven venues. Nine different UIs. No unified P&L. Good luck doing cross-platform arbitrage in your head. Good luck seeing, at a glance, which markets just moved on the Fed minutes. Good luck asking an AI about your positions, because your positions are scattered across four accounts and none of them surface through a common API.
This is the 1996 Yahoo Finance gap. Stocks existed. Data on stocks existed. What didn't exist was a place where you could see all of it in one view, with a cursor and a chart, for free, for retail. Tykhy is taking the same shot at prediction markets, with aggregation, AI research on top of your actual positions, and execution routing through your own exchange credentials (we never custody your money — I wrote up the interface argument in more depth in a separate post).
I don't think we'll be the only team working on this, and the category isn't going to wait for us. But the window is obviously open. The product category exists. The regulatory path exists. The volume exists. What's missing is the interface. That's a small gap to close. It won't be small for long.
What I think happens
On the record, my guesses for the next decade:
- By 2028, aggregate US prediction-market volume crosses $50B annualized notional. Probably sooner. I've been too conservative on this one twice already.
- By 2030, a meaningful share of Gen Z investors hold more prediction-market positions than equity positions by count. Not by capital. By count. The ticket sizes are smaller and the liquidity for small trades is better.
- The phrase "prediction market" dissolves into normal financial vocabulary the way "online broker" did. Nobody calls Schwab an online broker anymore. They just call it Schwab.
- Macro desks at real funds quietly use prediction markets as a calibration input. A few already do.
- The first large institutional fund to openly allocate to prediction-market-derived alpha already exists in stealth. That one's a guess. It's a confident one.
I'm nineteen. I've been trading prediction markets for two years. Every information edge I've found in that time has been a mispricing that persisted for hours because the venue was too small for institutional participants to bother. Those windows are closing. When they close, the people who built a workflow before the closing will have been early.
You can be early.
— Ilhan