Meta Tests Prediction-Market App 'Takes' With Employees

Meta Platforms (META) is testing an unreleased prediction-market app called Takes with its employees, asking them to predict Jeopardy! winners using a points currency rather than real money, according to a 24/7 Wall St. report published Nov. 14.

The test appeared the afternoon of Nov. 13, and Meta declined to comment and gave no launch date. Testers commit a points currency called cred, gained when a prediction is right and lost when it is wrong. No money changes hands, which separates Takes from Kalshi and Polymarket, where users trade real dollars.

Meta closed at $641.69, up $18.08 or 2.9%, then fell to $637.03 after hours, down 0.73%, as of 7:59 p.m. ET on Nov. 13. Nothing links the gain to the report, and shares remain down 13% over the past month.

The project has changed names three times without shipping. On June 23, 2026, reporting said Mark Zuckerberg directed a small team to build a points-based app called Arena; a day later internal documents described it under the codenames Antwerp and FBForecast. Those documents laid out a daily allocation of play money, with Meta's Llama AI model writing questions from trending topics, recommending markets and resolving them in near real time, and employees testing a prototype before a public launch on iPhone and Android.

Earlier reporting cited a goal of at least 100 million monthly participants aged 18 to 34, figures Meta itself has never stated. Meta launched a crowdsourced prediction app called Forecast in 2020 and shut it down about two years later, with internal documents citing the cost of selecting questions by hand. Llama writing and settling questions targets that expense.

Play money keeps Meta clear of commodities and state gaming regulators, although players risking nothing tend to guess casually, which weakens the signal. Listed peers moved without a stated cause on Nov. 13: Robinhood Markets (HOOD) fell 4.29% to $77.86, DraftKings (DKNG) slipped 1.20% to $18.89, and CME Group (CME) gained 0.78% to $317.97. A free Meta product would compete with them for attention and daily habit, threatening Robinhood's retail engagement more than CME's institutional futures franchise.

Like other Meta products tested by its own staff, Takes could change names and features or never ship. Set against fourth-quarter revenue guidance of $73-76 billion and 2026 capital spending of $135-145 billion, it is an option with no disclosed revenue. The investment case rests on advertising and AI: third-quarter revenue rose 26.07% to $64.60 billion, more than 10 million people set up Muse personal agents within about a month and a half of launch, and GAAP diluted EPS of $3.83 absorbed a $9.89 billion settlement charge. Shares trade near 21 times earnings, with 55 Buy ratings against 1 Sell and an average target of $850.10. Free cash flow was negative $2.1 billion in the quarter, and fourth-quarter revenue should land inside the $73-76 billion range. A close below the one-year low of $520.26 would signal the market has rejected the spending plan; a public Takes listing would add upside on top of that thesis.