Mantic, a London company that applies AI to judgmental forecasting, has raised $25 million in seed funding. Radical Ventures led the round at a valuation the startup is not disclosing; Microsoft’s M12, Thinking Machines Lab and Balderton Capital also took part, and the board gains Aaron Rosenberg of Radical. Sky News broke the story, and Reuters confirmed the amount and the terms.
The news lands weeks after a strong showing at the summer 2026 Metaculus Cup, a crowded forecasting contest. Entrants stake probabilities on questions spanning politics, economics and culture, and Mantic beat every human player, finishing behind only one other machine, a bot named laertes. Reuters read the result as a first: machines had never so thoroughly overshadowed the human field.
Founded in 2024 by Toby Shevlane and Ben Day, Mantic chases questions that resist standard statistical treatment – business, technology, geopolitics, policy and culture, usually weeks or months ahead. It does not train a frontier model of its own; instead it tunes other labs’ models for forecasting, scores them against past events and refines them from there.
Shevlane spent time as a research scientist at Google DeepMind, where he wanted sharper forecasts about global developments that touch AI. He describes the goal as upgrading the level at which humans can understand the future.
The tournament also exposed a flaw familiar to human forecasters: herding. Would Shakira’s “Dai Dai” overtake “Waka Waka” on the Billboard Hot 100? That was one of the questions put to the field, and the crowd settled hard on the consensus answer, which turned out wrong. Mantic, assembling its own evidence base instead of following the pack, avoided that miss. Interest, Rosenberg said, has come from companies and government agencies, some of which already use the system, with hedge funds and trading firms especially keen.
