Machines Are Eating the Prediction Market

Every market is a prediction market. Stocks predict earnings. Bonds predict default. Housing prices predict where people want to live next year. The only difference is that prediction markets are honest about it — they price beliefs directly, skip the derivative layers, and settle on outcomes instead of proxies.

We're early. Not early like "this could be big someday" — early like the infrastructure is being built right now, the regulatory lines are being drawn this month, and the people who show up in the next 12 months are the ones who shape what this becomes. Open interest just crossed $1 billion for the first time. The NYSE is invested. The CFTC is fighting state governments over jurisdiction. None of this existed two years ago.

Reality Managers is an SF community for the people in the room while this happens. Traders, builders, researchers, people who think probabilistically about the world and want to be around others who do the same. We host events, we write about what's happening, and we're learning in public — because nobody has this figured out yet, and pretending otherwise would be dishonest.

This is the first issue of the newsletter. It won't be perfect. It'll get sharper as we go. What it will be, every time: curated, opinionated, and written by people with skin in the game.

Machines are eating the prediction market.

Last Saturday, CoinDesk reported that algorithmic traders extracted $40M from Polymarket over the past 12 months. One account made $2.2M at a 74% win rate using a neural network. The strategy: find moments when Yes + No contracts sum to less than $1, then buy both legs. Free money, as long as you're fast and systematic enough to find it first.

If you've been trading on instinct, this is a useful update on who you're playing against.

This is what prediction markets look like in 2026. Not a hobbyist playground. Contested financial infrastructure. Institutional money has arrived, the CFTC is fighting to own the space, and retail algorithms are running. And the jurisdictional question that determines whether all of this is legal, nationwide, just got very interesting.

That's what this newsletter covers. Curated, editorialized PM intel every two weeks. Skin in the game, not observation from outside.

The CFTC Is Going to War. States Are Fighting Back.

The most important regulatory story in prediction markets happened over six weeks and most people missed it.

Here's the sequence: CFTC Chairman Brian Selig (confirmed December 2025) started moving fast. January 29: withdrew the 2024 proposed ban on political event contracts. February 4: withdrew staff guidance against sports contracts. February 17: filed an amicus brief in federal court asserting federal preemption. The legal argument: CFTC jurisdiction supersedes state law.

The states didn't sit still. Massachusetts got a court injunction blocking Kalshi's sports markets on January 23. Nevada sued Kalshi. Tennessee is pressing enforcement actions. Twenty-three Senate Democrats sent a letter demanding Selig reverse course.

This fight comes down to one jurisdictional question: are prediction markets gambling (state jurisdiction) or derivatives (CFTC jurisdiction)?

It matters more than it sounds. CFTC wins: federal framework, national liquidity pools, institutional-grade legitimacy. States win: every market fragmented by state, subject to gambling regulations written for casinos, Kalshi fighting 50 separate battles.

The CFTC's amicus brief is the aggressive play. Selig isn't waiting for Congress. He's asking courts to declare federal preemption now, betting the judiciary sides with derivatives classification. Recent case law supports it, but the outcome isn't settled.

The current CFTC posture is the most market-friendly it's ever been. If you're building or trading in this space, that brief is worth reading. It's the best argument for why prediction markets are derivatives, not gambling, and why federal law governs. Whether it wins is a different question.

$1B in aggregate open interest. The sector crossed $1 billion around February 7, driven partly by Super Bowl LX ($795M in Polymarket volume for that event). Polymarket hit $12B in January trading volume, a monthly record. The number is big enough now that it gets cited in investor decks and regulatory briefs. That's what legitimacy looks like: your market shows up in other people's arguments.

ICE/NYSE put in $2B at a $9B valuation. The Intercontinental Exchange, which owns the New York Stock Exchange, invested in Polymarket in October 2025. That's not a venture bet. That's incumbents hedging their exposure to something that might replace parts of their business. When the NYSE has financial upside on Polymarket's success, the "niche" framing stops being credible.

Polymarket launched real estate markets with Parcl. January 5: housing price index contracts for NYC, LA, Miami, SF, and Austin. This is the product pivot to watch. Polymarket has been an election novelty. Housing markets are persistent, liquid, and affect every adult in the country. If these gain traction, prediction markets become financial data infrastructure, not just political entertainment.

UMA tightened oracle governance. UMA Protocol whitelisted 37 approved resolvers for its oracle system. No hack, despite what circulates. The move to an approved-resolver architecture is the right call. Oracle manipulation is the attack surface that breaks PM legitimacy. Stricter governance makes these systems harder to exploit and easier to trust.

Prediction markets don't just predict reality. They shape it.

The tariff feedback loop is running right now.

The Supreme Court partially struck down Trump's tariffs on February 21. The next day, Trump announced a new 15% global baseline rate. Polymarket markets are pricing these events in real time. Trump has publicly referenced Polymarket odds. His team watches them.

So the loop goes: markets price the probability of a policy action. Policymakers read the market. That influences their positioning and timing. Which changes the underlying probabilities. Which changes the odds.

This is Soros's reflexivity thesis running live. Markets don't passively reflect reality. They participate in creating it. When the president uses your price signal as political intelligence, you're not predicting policy. You're inside the feedback loop that generates it.

We're not observing. We're managing.

Reality Managers is the SF prediction market community. IRL meetups, builder demos, market autopsy sessions. The newsletter runs on the same logic: curated, editorialized, no hedging.

If someone should be reading this, forward it. They can subscribe at realitymanagers.beehiiv.com

See you Thursday.