Meta is preparing to enter the rapidly growing prediction market industry with a new standalone application called Arena, according to internal documents obtained by NPR. The move places the tech giant in direct competition with established platforms like Kalshi and Polymarket in an industry that market analysts project could soon reach $1 trillion in value. Under instructions from CEO Mark Zuckerberg, Meta's development team is building a platform designed to let users leverage collective intelligence to forecast real-world events.

The project, which has been developed under the internal codenames "Antwerp" and "FBForecast," represents a highly strategic response to the shifting legal and economic realities of the digital age. Prediction markets have faced intense scrutiny from federal regulators, who have historically sought to restrict Americans' freedom to engage in speculative forecasting. By utilizing a "play money" system rather than real-world currency, Meta is positioning itself to navigate these regulatory hurdles while delivering a service that consumers clearly demand.

According to the leaked documents, Arena users will receive a "daily virtual allotment" of play money to wager on yes-or-no questions regarding future events. This structure allows Meta to establish its market infrastructure without running afoul of the complex financial and gaming regulations that currently bind real-money competitors. It offers a practical path forward for innovation in a sector currently stifled by government intervention and bureaucratic delay.

A key driver of Arena’s design is the integration of advanced artificial intelligence. Meta plans to use its proprietary large language model, Llama, to automate the creation, recommendation, and resolution of prediction markets. The AI will automatically generate questions from trending topics and deliver personalized recommendations to users. Most notably, the AI will also be responsible for resolving markets in "near real-time," meaning that technology, rather than human administrators, will verify whether an event actually occurred.

This reliance on AI represents a major step forward in operational efficiency. Meta's previous attempt at a prediction platform, an app called "Forecast" launched in 2020 to track events like the pandemic, was shut down in 2022. Internal documents reveal that the shutdown was driven entirely by "the operational cost of manual question curation." By replacing expensive human curation with automated Llama AI processes, Meta is applying free-market efficiency principles to solve the cost constraints that previously derailed its forecasting initiatives.