Analysis
Meta is testing “Takes”, a prediction app where employees bet “cred” points instead of money, the Wall Street Journal reports
Meta employees are testing Takes, a prediction-market app where bets use "cred" points, not cash, the WSJ reports. No launch date; Meta declined to comment.
Meta employees have started testing a prediction-market app called Takes, Meghan Bobrowsky reported in the Wall Street Journal on November 13, citing people familiar with the matter (Techmeme summary, Reuters via Investing.com). Users wager on the outcome of questions or events with "cred", short for credibility, not real money. A correct prediction wins cred, a wrong one loses it. The app is not public, and it is unclear when, if ever, it will be released. Meta declined to comment (The Block).
Key points
- A points game, not a market with stakes: unlike Kalshi and Polymarket, Takes does not allow money wagers, according to the WSJ.
- Internal test only: Meta often has employees test apps before a public release ("dogfooding"), but not every product gets out of that phase, and names and features can change, the paper notes (Sulat, summarizing the WSJ).
- Example questions: whether former Apple CEO Tim Cook will become Nvidia's next CEO (most voters bet he will not) and who would win an episode of Jeopardy (Investing.com, Sulat).
- Continuation of "Arena": the project follows the app Mark Zuckerberg asked a team to build in June, then called Arena internally, according to The Block and Investing.com.
From Arena to Takes
On June 23, the New York Times reported that Zuckerberg had tasked a small team with building a standalone app similar to Polymarket and Kalshi, called Arena internally. It would "probably rely" on points rather than money, although Meta had not ruled out real-money betting later (The Verge, CoinDesk). People inside the company described it as both experimental and a "top priority".
The next day, NPR published internal documents describing "a daily virtual allotment" of "play money" for betting on "the outcome of future events". Meta has tried this before: Forecast, launched in 2020, was shut down two years later. According to the documents seen by NPR, the reason was "the operational cost of manual question curation", and the new project is a "rebuild" of it.
The WSJ report, as relayed by The Block, does not say whether real-money betting is still on the table.
The AI angle: questions written and settled by a model
This is where the project concerns AI products. According to the June documents described by NPR, the app was to use Llama to generate questions automatically from trending topics, make "personalized market recommendations", and resolve markets in "near real-time". In other words, a model would have the final say on whether an event did or did not happen.
That answers the problem that killed Forecast, the cost of human curation, but it raises another: on a prediction market, settlement is the critical moment. A model error on a contested event decides who wins. With points, a mistake costs no one money, which makes it a low-risk test bed. Neither the WSJ nor its relays say whether these AI features are in the version employees are testing: they are described in June documents about Arena, not in the Takes coverage.
Why points rather than money
| Point | Takes (WSJ, November 2026) | Kalshi / Polymarket |
|---|---|---|
| Stake | "cred" points, no money | Real money |
| Access | Meta employees only | Public |
| Legal status | Outside the event contract model | Lawsuits from US states that call them illegal gambling; the platforms say they fall under federal regulation |
In June, gaming lawyer Daniel Wallach told NPR that launching without money could give Meta time to seek a license from federal regulators while the legal landscape settles: "We're clearly in legal limbo," he said, citing more than 30 pending lawsuits. That is a lawyer's analysis, not a stated Meta strategy. The sector is moving fast: Bloomberg reported this month that Kalshi is in talks to raise at a $65 billion valuation, weeks after a $40 billion round (our November 8 article).
What is not established
- Whether Takes will be released, and when: the WSJ says it is unclear, and Meta gives no timetable.
- Whether Takes uses the Llama-based features described in June for Arena.
- Whether Meta still plans to allow real-money betting later.
- How many employees are testing it, and how cred is distributed.
Method and limits
Sources checked on November 14, 2026. The WSJ article is paywalled and was not read in full: its content comes from Reuters, The Block, Investing.com, Techmeme and Sulat. The details on Arena come from the June reports of the New York Times (via The Verge and CoinDesk) and NPR, which read internal documents. This is not investment advice.
Sources
- Techmeme: summary of Meghan Bobrowsky's article (Wall Street Journal) (verified 2026-11-14)
- Investing.com, carrying Reuters: Meta employees test prediction-market app, WSJ reports (verified 2026-11-14)
- Investing.com: Meta tests prediction-market app for employees, WSJ reports (verified 2026-11-14)
- The Block: Meta tests prediction market app using 'cred' points instead of cash: WSJ (verified 2026-11-14)
- Sulat: Meta employees test points-based prediction app Takes, WSJ says (verified 2026-11-14)
- RG: Meta Tests Takes Prediction Market App Using 'Cred' Points (verified 2026-11-14)
- NPR: Meta plans to release AI-powered prediction market app (verified 2026-11-14)
- The Verge: Zuckerberg reportedly wants a Polymarket clone — but without real money (verified 2026-11-14)
- CoinDesk: Meta is developing a prediction market app called 'Arena' as sector booms: NYT (verified 2026-11-14)
Spotted an error in this article? Report it · Our corrections policy