A handful of startups in New York and beyond are wagering that artificial intelligence can outpredict even the most seasoned human analysts, whether the question concerns stock market swings, the reopening of a strategic shipping lane, or how a pop song will fare on the charts. What was once the domain of so-called superforecasters - rare individuals whose probability estimates consistently beat the average - is now being replicated, and in some cases surpassed, by machine learning systems trained on vast stores of public and private data.
Dan Schwarz, cofounder of the forecasting company FutureSearch, believes the shift will feel seismic in retrospect. By 2030, he said, current decision-making methods will look like "complete Stone Age thinking." That confidence is not without basis: in early September, two AI models took the top spots in the Metaculus Cup, a competition pitting human and machine forecasters against one another, marking a milestone many in the field saw coming. As these tools proliferate and become commercially available, users navigating the growing marketplace of AI-driven platforms would do well to understand why the cheapest option is rarely the best, since the quality of underlying data access and model rigor often separates genuinely useful forecasts from superficial ones.
From Superforecasters to Algorithms
Tom Liptay, FutureSearch's resident superforecaster, is candid about the implications for his own profession. Schwarz said Liptay "knows that these things are going to surpass him" in the medium term. The trajectory mirrors what happened to chess and later to complex strategy games, where machines eventually eclipsed even the most gifted human practitioners. The difference now is scale: forecasting models can cross-reference interconnected variables, such as oil prices and geopolitical tensions, far faster than any individual analyst could.
That speed has already attracted attention from hedge funds, who Veniamin Veselovsky, cofounder of Preseen, describes as typically the fastest movers once a profitable signal emerges. Preseen's own results illustrate the stakes: Theo Summer, Veselovsky's cofounder, reportedly turned $35 into $1.9 million over seven months trading on the prediction platform Kalshi. Algorithmic trading has shaped financial markets for more than a decade, but AI forecasting threatens to accelerate automation further, potentially sidelining retail investors who lack access to sophisticated tools. Veselovsky warned that consumer-facing prediction markets like Kalshi and Polymarket "will soon just be dominated by AI forecasting systems."
Beyond Wall Street
The reach of these systems extends well past finance. Toby Shevlane, cofounder of the startup Mantic, said his company has supplied a major American corporation with regulatory forecasts and has worked with the British government on policy-related risk assessment. The most advanced models no longer simply answer questions posed to them; they can generate their own lines of inquiry and assign probabilities across multiple possible outcomes, Shevlane noted. Applications are emerging in medical research as well, where Schwarz said forecasting tools could help determine which treatments deserve funding priority.
Weighing the Risks
Wider access to predictive AI carries real tradeoffs. Veselovsky acknowledged the potential benefits while voicing unease about an uneven global playing field, saying he worries that rival nations such as Russia or China might develop superior systems. As forecasting tools migrate from niche trading desks into government policy, scientific research, and insurance risk modeling, questions about oversight, data provenance, and equitable access will likely intensify alongside the technology's growing influence.