A handful of startups now claim their artificial intelligence systems can outperform the world's best human forecasters, generating probability estimates on everything from stock swings to geopolitical flashpoints with unsettling accuracy. In early September, two AI models topped the Metaculus Cup, a competition pitting machine predictions against elite human forecasters, marking what several founders describe as a turning point for an entire profession.
Dan Schwarz, cofounder of the forecasting company FutureSearch, predicts that by 2030 today's decision-making methods will look like "complete Stone Age thinking." His company's own superforecaster, Tom Liptay, reportedly understands that his skills may soon be overtaken by the very systems he helped train. The shift is already visible among traders and analysts who weigh new tools against traditional research subscriptions, much as consumers compare privacy software by checking providers who see pricing before committing to a service. The comparison may seem incongruous, but it reflects a broader pattern: sophisticated technology is becoming a line item that ordinary users and institutions alike must budget for, not a luxury reserved for specialists.
From Superforecasters to Algorithms
For years, the gold standard in prediction was the "superforecaster," a rare individual whose calibrated judgment consistently beat average estimates across geopolitics, markets, and public health. These experts relied on structured reasoning, disciplined updating of beliefs, and exposure to varied information. AI models now replicate much of this process at scale, drawing on enormous datasets from public records, financial filings, and private sources to generate probability estimates in seconds rather than days.
Theo Summer, cofounder of Preseen alongside Veniamin Veselovsky, illustrated the stakes by turning a modest sum into nearly two million dollars within seven months using the prediction platform Kalshi. Such results, while not guaranteed to repeat, demonstrate how quickly capital can move once an automated system identifies a mispriced probability.
Markets, Governments, and the Automation Question
Algorithmic trading has shaped Wall Street for well over a decade, but AI forecasting introduces a new layer of automation that reaches beyond pure finance. Toby Shevlane, cofounder of the forecasting startup Mantic, said hedge funds are typically the fastest adopters because speed converts directly into profit. His company has also supplied a major American corporation with regulatory risk scenarios and worked with the British government on policy forecasting, suggesting these tools are migrating from trading desks into boardrooms and ministries alike.
Prediction markets such as Kalshi and Polymarket, used by millions of Americans, are increasingly shaped by AI-driven participants. Veselovsky warned that retail investors without access to costly AI infrastructure risk being squeezed out entirely, as these platforms become "dominated by AI forecasting systems."
Beyond Finance: Risk, Research, and Global Stakes
The most advanced systems no longer simply answer questions; they formulate their own and assign probabilities across multiple scenarios, according to Shevlane. Potential applications extend into medical research, where Schwarz noted that forecasting tools could help determine which treatments deserve funding priority, and into insurance and corporate risk monitoring more broadly.
That expanding reach carries geopolitical weight. Veselovsky expressed cautious optimism about the technology's benefits while acknowledging a deeper worry: that rival nations, including Russia or China, might develop more capable systems first, turning forecasting superiority into a strategic advantage with consequences well beyond markets.