A Forecasting Bot Just Beat 10,000 Human Predictors. Then It Raised $25 Million.

Last week, a London startup called Mantic entered a prediction tournament against 10,000 human forecasters. These weren't random guesses. The Metaculus Cup is one of the most respected forecasting competitions in the world. Participants assign probabilities to real events: elections, economic shifts, cultural milestones.

Mantic's AI won. It beat every single human competitor. Only one other bot scored higher.

Three days later, the company closed a $25 million seed round led by Radical Ventures, with Microsoft's M12 fund and Balderton Capital joining in. Hedge funds and trading firms are already lining up to use the technology.

Why forecasting matters more than you think

Most businesses treat prediction like guesswork dressed up in a spreadsheet. You look at last quarter's numbers, add a growth assumption, and call it a forecast. The gap between that process and what Mantic just demonstrated is enormous.

Their model doesn't just crunch numbers. It reads news, weighs conflicting evidence, updates its confidence in real time, and assigns probabilities to events that don't have clean data sets. That's the part worth paying attention to.

Think about the decisions you make every quarter. Should you hire two more people or wait? Is that new market worth entering before Q2? Will your biggest client's industry face a downturn? You're already forecasting when you make those calls. You're just doing it with gut feelings and last month's spreadsheet.

The hedge fund signal

When hedge funds line up for a product, it tells you something. These aren't companies that buy software because it sounds impressive. They buy it because it makes them money. The fact that financial firms want Mantic's forecasting layer means the predictions are accurate enough to bet real capital on.

That signal matters for every business owner, not just the ones moving billions. If a prediction engine can price risk better than a room full of analysts, it can probably help you figure out whether to sign that lease or delay your product launch.

What this means for small businesses and agencies

The technology is cool, but the price point is what I keep thinking about.

A $25 million seed round is small by AI startup standards. That means Mantic is building something that needs to scale broadly, not just serve a handful of Wall Street clients. If this kind of forecasting becomes accessible at $50 or $100 a month, suddenly every operations manager has access to prediction quality that used to be reserved for quantitative trading desks.

For AI agencies, this opens a real service line. You could bundle forecasting into the decision-support packages you sell to clients. Not vague "we'll help you use your data better" promises. Actual probability scores on outcomes that matter to their business.

The honest limitation

Prediction AI isn't magic. Mantic beat human forecasters on the kinds of questions the Metaculus Cup asks: well-defined events with clear resolution criteria. "Will the Fed cut rates by December?" is a solid question for an AI forecaster. "Should I fire my underperforming sales manager?" is not.

The tool works best for decisions with measurable outcomes and enough public information to train on. Strategic decisions that depend on your specific team, culture, or relationships still need a human.

Most businesses don't even use forecasting tools at all, though. They rely on instinct and whatever their accountant put together last month. The bar isn't "perfect prediction." The bar is "better than what you're doing now." For most companies, that bar is very low.

What to watch

Keep an eye on Mantic's go-to-market over the next six months. If they launch a self-serve product under $200/month, that's a signal that quality forecasting is about to become a standard business tool rather than a luxury.

For now, the takeaway is simple. AI can predict real-world events more accurately than expert humans. The companies that figure out how to use that for everyday decisions will have a real advantage. The ones that ignore it will keep guessing.

— Mark Garza, Laimen AI