Meta is building its own prediction market. The company is internally testing an app called Takes, in which employees forecast the outcomes of real-world questions and events, from who will run Nvidia next to who wins that night's episode of Jeopardy. What separates Meta's version from Kalshi or Polymarket is the single feature regulators keep going to court over: there is no money in it.
Players on Takes stake virtual points called cred, short for credibility. A correct call adds cred and a wrong one subtracts it. No cash changes hands, nothing resolves into cash, and Meta has disclosed no way to convert cred into anything of value. The result is a prediction market stripped down to the part its defenders say matters most, the forecasting, with the part its critics attack, the payout, removed entirely.
The project is internal for now. Meta declined to comment, has set no launch date, and has made no decision on whether Takes ever reaches the public. The company treats the exercise as dogfooding, the industry term for making staff live on an unfinished product before anyone outside sees it, and the names and features on screen today can change before any release.
A market that pays in credibility, not cash
Takes works the way the paid exchanges do, minus the wallet. Employees see a question, take a side, and are scored when reality resolves it. Get it right and your cred climbs; get it wrong and it falls. The score is the whole reward, a running tally of who in the building actually calls things correctly.
That design quietly answers the attack that has dogged Kalshi and Polymarket throughout 2026. When the stake is points rather than dollars, a position looks like a forecast rather than a bankroll at risk, and the activity reads as a game of judgment rather than a casino floor. Whether that distinction survives contact with a billion-user platform is the open question, but on an internal build it is doing real work: Meta can study how people forecast, how markets of opinion move, and how to make the format compelling without touching the money that invites a regulator.
The sample questions that leaked out show the range. One, posted November 12, asked whether former Apple chief Tim Cook would become Nvidia's next chief executive; most staff forecast no. Another, that same week, asked who would win that night's Jeopardy. The mix of a market-moving succession rumor and a television game show is the point: Takes is being tuned for everyday curiosity, not just finance.
Ρενέ Χάας25%ΝαιΌχι
Ίαν Μπακ19%ΝαιΌχιThe Nvidia question is live in the real prediction markets too, where traders price the odds on the chipmaker's leadership rather than score points for a guess.
From 'Arena' to 'Takes'
The app did not start under its current name. An earlier internal version reportedly carried the codename Arena, and the effort traces back to Mark Zuckerberg himself, who is said to have directed a team to build it. That lineage matters: a prediction market inside Meta is not a skunkworks experiment some product manager slipped through, it is a project with the founder's name attached, which is usually how things at Meta graduate from test to launch.
It also fits a pattern. Meta has a long history of watching a format prove itself elsewhere, Stories, Reels, standalone messaging, and then building its own version at platform scale. Prediction markets have spent 2026 proving themselves, in audience and in headlines. A company that owns Facebook, Instagram and WhatsApp deciding the format is worth a prototype is, on its own, the most serious signal yet that forecasting-as-a-feed is going mainstream.
Why the audience skews young, and why Meta cares
The demographics explain the interest. In a survey by Kagan, the research arm of S&P Global Market Intelligence, 61% of US prediction-market traders were under 35. The same work found the activity heavily male, about 66%, and heavily sports-driven, with roughly 70% of traders active on sports questions, while about 10% of American internet adults had traded a prediction market in the past year.
That is precisely the cohort Meta worries about aging out of Facebook and Instagram. A format that already pulls in people under 35, that rewards being right in public, and that generates a constant stream of fresh, shareable questions is close to a designed-in-a-lab answer to the platform's oldest problem, keeping young users engaged. Points instead of dollars make it something Meta can roll out to that audience without the licensing fight a real-money product would trigger.
The legal heat Meta is stepping around
The timing sharpens the contrast. The same week Takes surfaced, the real-money side of the industry was in court. On November 12 the US government and the Commodity Futures Trading Commission, the federal regulator that claims exclusive authority over event-contract markets, sued the state of Arkansas to stop it enforcing its gaming laws against Kalshi. It was the 11th time Washington has moved to preempt a state this way, after Arizona, Connecticut, Illinois, Kentucky, Minnesota, New Mexico, New York, Rhode Island, Wisconsin and Utah.
Arkansas had moved first, filing on November 4 and casting Kalshi as an unlicensed casino operating in the state. That framing is the fight in miniature: states see a game of chance, the federal regulator sees a financial market, and Kalshi and Polymarket are caught paying lawyers in the gap between the two readings. Every one of those cases turns on the presence of real money.
Meta has not been sued and Takes is not part of any of this litigation. That is not an accident. By launching, if it launches, with cred instead of cash, Meta would walk into the prediction-market business through the one door no state gaming regulator has a key to.
Apple97%ΝαιΌχι
Tesla3%ΝαιΌχιWhat a no-money market really changes
The easy read is that a points-only market is a toy. The more useful read is that Meta may have found the version of this product that scales to a billion people precisely because it is a toy. Kalshi and Polymarket built real-money exchanges and spent their growth years in regulatory trench warfare. Meta is proposing to take the habit those platforms created, the reflex to turn any uncertain event into a position, and hand it to a mass audience with the legal liability engineered out.
If that works, the competitive threat to the paid platforms is not that Meta takes their traders. It is that Meta trains the next generation to forecast for free, so that by the time those users are old enough and inclined to put money down, the exchange they reach for is a reflex Facebook built. The question that should worry Kalshi and Polymarket is not who wins Jeopardy. It is whether the most valuable prediction market ever built turns out to be the one that never pays out.
