Meta's New Prediction Game Runs on 'Cred' Points and Zero Real Money
Meta is testing an internal forecasting app called Takes, where employees win or lose a points balance called cred. No cash, no settlement, no public launch. That's the whole point, and also the reason it tells us less than the headlines suggest.
Meta built a prediction market and then took the money out.
That's the short version. Employees are testing an internal app called Takes, which was previously called Arena, where you make forecasts and your payout lands in a balance called cred. Points. Not dollars. No cash settlement, no custody, no payout rails, nothing that resembles a traded event contract. The reporting traces back to the Wall Street Journal by way of The Block, and it describes a staff experiment, not a consumer launch.
Which is exactly why it's worth two minutes of your attention.
What Takes Actually Does
Per the reporting, Takes lets Meta employees stake their read on future outcomes and get graded on accuracy. Win, your cred goes up. Lose, it goes down. There's a ranking layer implied by the design, because a points balance nobody can see is a points balance nobody cares about.
And yes, an internal rename happened. Arena became Takes. If you've worked anywhere near product at a big company, you know what a rename means. It means someone cares enough to keep iterating, and it also means the thing is still soft enough to change names without telling customers.
Here's the part that makes this more interesting than a random internal beta. Meta has tried this before. In 2020 the company shipped Facebook Forecasts, a points-based prediction feature that graded your calls with a Brier score, the same scoring rule used by professional forecasters. It was shut down by mid-2021. Roughly one year of life. No money, no market, no exit.
So Takes isn't a first attempt. It's a third or fourth pass at the same question. Do people like forecasting? Do leaderboards make them better at it, or just louder?
Why Zero Dollars Is The Whole Story
Real-money prediction markets drag in everything. Customer eligibility. Settlement disputes. Market manipulation. Gambling and derivatives law. Federal and state regulators who can't agree on what an event contract even is. Kalshi has been fighting that fight in court. Coinbase got pulled into a CFTC dispute over single-stock perpetuals. The legal bill for a real-money product is measured in years, not sprints.
Points sidestep almost all of it. A cred balance isn't a security. It isn't a derivative. It isn't a gambling stake. It's a number in a database with a leaderboard attached.
So my read: the money is the easy part. The hard part is engagement. Anyone can wire up a payout. Getting a person to come back on a Tuesday to forecast something they don't care about is the actual product problem. Meta is solving for retention first, and that's the correct order of operations whether or not anything public ever ships.
But there's a problem points don't fix. An employee who forecasts "we ship X in Q3" while sitting three desks from the roadmap is trading on material nonpublic information with extra steps. Nobody's going to jail over cred. But if Takes ever feeds into performance reviews, or if cred starts showing up in promotion conversations, you've built an insider signal generator and called it a game. That's the risk I'd be watching.
Ask yourself this. Do you want a leaderboard that rewards the person with the best information, or the person best positioned to already know the answer?
Now zoom out. Prediction markets are having a moment. Polymarket cleared a billion dollars in volume in a single month around the 2024 election. Kalshi went from a niche idea to a courtroom fixture. The primitive works. The question was never demand, it was plumbing.
And this is where my Solana bias shows. On Solana, this stuff ships as code and finds out in public. Markets go live on mainnet, get farmed, get drained, get fixed, and the whole cycle happens before a committee finishes its second meeting. Blinks turned a market into a link you can tap. Jupiter routes the liquidity. Priority fees sort out who actually wants the trade. Solana doesn't wait for permission. The speed difference isn't theoretical. You feel it.
Meta, by contrast, is running a points game behind a login wall for its own staff. Different sport entirely. That's not a knock, it's just honest about what's happening. One group is doing discovery in a sandbox. The other is doing discovery in production with real money on the line.
Here's my hot take. A cash-free prediction game is a hedge, not a strategy. It lets Meta learn about forecasting behavior without committing to the compliance scaffolding, and it lets the company walk away clean if the signal is weak. That's smart management. It's also the cheapest possible version of the idea, and cheap versions rarely become flagship products.
The other thing people keep missing. Meta doesn't need a prediction market to monetize attention. It already has north of 3.4 billion people opening its apps daily. Forecasting is an engagement surface, not a new business line. If Takes ever goes public, it'll be a feature inside an existing feed, not a destination. That changes the economics in a way a standalone exchange can't replicate, and it also caps how serious the product will ever get.
What Would Make Me Believe It
Watch the spec, not the headlines. Three things matter. First, the scoring rules. If Meta publishes how cred is earned and whether it converts into anything, that tells you the intent. Second, opt-in design. If forecasting is voluntary and invisible to managers, it's a research tool. If it's tracked and visible to managers, it's a performance metric with a gambling skin. Third, whether any of this escapes the building.
Until one of those shows up, Takes is a reported internal test. Nothing more. I've watched enough of these die quietly to know that an internal rename isn't a launch. Facebook Forecasts had a scoring system and a year of runway and it still went nowhere.
So here's the takeaway. The interesting part of this story isn't that Meta is curious about prediction markets. Everyone is curious about prediction markets right now. The interesting part is that a company with infinite engineering resources chose to build the version with no money in it, while anonymous builders on Solana shipped the version with real money and real losses before the week was out.
A game with no money has no losers. That's the point. It's also the reason it tells you almost nothing yet.