London-based Mantic, founded by ex-DeepMind and Citadel researchers, claims its AI forecaster rivals top human experts.
Mantic, an AI research and product company, has come out of stealth with $4 million in pre-seed funding to develop automated forecasting systems for complex global events. The round was led by Episode 1 Ventures, with participation from DRW and several AI researchers from Anthropic and Google DeepMind.
The London-based startup is led by co-founders Toby Shevlane and Ben Day, who previously worked at Google DeepMind, Citadel, and the Universities of Cambridge and Oxford. Their mission: to push AI forecasting systems beyond human limits in areas such as geopolitics, business, and policy.
Mantic’s technology is designed to replicate the flexible reasoning of human “superforecasters” but at digital speed and scale. According to the company, its latest system set a new state of the art in forecasting accuracy when backtested against 348 questions from the Q2 2025 Metaculus benchmark. The system is also competing in the ongoing Metaculus Cup, where it has performed on par with top human forecasters.
“Forecasting requires deep research and reasoning, so traditional computer models don’t beat the top humans” Shevlane said. “Our system thinks like a human forecaster, but with digital speed and scale.”
The company argues that most automated models fall short in data-poor or fast-moving domains. By contrast, Mantic’s backtesting approach allows its AI to iterate rapidly, training on thousands of historical forecasting questions and reducing the feedback cycle from months to milliseconds.
Mantic says its technology can deliver predictions at a speed and volume beyond human capacity. Potential applications include helping private equity firms, insurers, trading houses, and corporates make better-informed decisions.
The startup now plans to collaborate with an early client willing to integrate AI-driven forecasting into decision-making processes. Longer term, Mantic’s ambition is to achieve “superhuman accuracy” in forecasting—transforming judgmental forecasting in the way automation reshaped weather prediction.