Meta’s Arena: Here’s What We Know About the New A.I.-Driven Prediction Market App

Picture the scene that opened The Social Network: a Harvard sophomore in a hoodie, hunched over a laptop at two in the morning, building something nobody asked for because he is convinced it will make him matter. Two decades and roughly three billion daily users later, that same founder is reportedly doing it again… but this time the late-night project is not ranking classmates, it is ranking the future itself.

According to internal documents reviewed by NPR, building on an earlier report from The New York Times, Meta CEO Mark Zuckerberg has quietly assigned a small team to build a stand-alone app, internally nicknamed Arena, that lets people forecast how real-world events will unfold.

What Are Prediction Markets, and Why Is Silicon Valley Suddenly Obsessed?

For anyone arriving late to this conversation, a prediction market works a bit like a stock exchange for opinions. Instead of buying a share of a company, you buy a position on a yes-or-no question: will a film clear two hundred million dollars, will a central bank cut rates, will a hurricane make landfall by Friday. The price of that position drifts up or down as collective sentiment shifts, then pays out once the real outcome is known.

What used to be a curiosity for political junkies has grown into a category some analysts believe could process a trillion dollars a year before the decade is out. Kalshi and Polymarket alone moved a combined $50 billion in 2025, and that figure has already topped one hundred thirty billion this year: a hint at why a company with Facebook’s reach would want a seat at that table.

Arena, Explained: How Meta’s AI Would Actually Run It

Per the documents NPR reviewed, Arena, also referred to internally by the codenames Antwerp and FBForecast, would hand users a daily ration of virtual currency rather than real money, at least to start.

The more striking detail is who is minding the store.

Meta’s Llama model would generate questions out of whatever is trending, suggest markets tailored to each user, and then act as the judge, deciding in near real time whether an event actually happened. It is a tidy piece of automation, and also a reminder that an algorithm with no stake in the outcome would be the one ruling on events that, elsewhere, real money rides on.

Why Shares Dropped the Moment Arena Leaked

News rarely gets louder than moving other companies’ share prices before a product even exists.

Shares of DraftKings slipped once the report spread, Flutter Entertainment, the parent of FanDuel, wobbled, and Robinhood, which already lists event contracts, dipped too. Canadian fintech Wealthsimple had only just unveiled its own prediction market app, and now finds itself with considerably bigger company.

That kind of market reaction to a project still confined to internal testing says less about Arena’s prototype than about Meta’s gravitational pull. More than three and a half billion people open one of its apps every day (over forty percent of the planet), an audience no rival in this space has ever had on tap.

Not Meta’s First Rodeo: From Forecast to the Metaverse Hangover

This is not even Mark Zuckeberg’s first attempt.

A points-based app called Forecast briefly let users predict pandemic milestones back in 2020, before getting quietly shut down in 2022 over the cost of manually curating questions. Arena, per the documents, is explicitly framed as its rebuild.

Standalone apps have not always been Meta’s strong suit, either: Threads broke records by signing up one hundred million people in five days, then spent the following months bleeding many of them back out. And looming over all of it is the roughly seventy-seven billion dollars the company poured into the Metaverse, a venture still searching for a convincing return.

The legal terrain is not settled either. Gaming attorney Daniel Wallach points to more than thirty pending lawsuits testing whether prediction markets fall under federal or state authority, and admits plainly, “We’re clearly in legal limbo.”

So will Arena actually launch, or quietly join Forecast in the graveyard of Meta’s standalone experiments? Nobody, perhaps not even Zuckerberg, can resolve that market yet. But if the hoodie-clad coder of 2004 taught Silicon Valley anything, it is that he tends to keep building until somebody, eventually, downloads it.