On 5 March 2026, the day after Hypermind and beBartlet announced their partnership at Mobile World Congress, Agenda Pública published a long interview with Émile Servan-Schreiber in Spanish and English. Its title asked: “Can a citizen network predict the future of war better than the CIA?” Marc López Plana, the outlet’s editor and director, conducted it at beBartlet’s Madrid office. Servan-Schreiber, Hypermind’s CEO and a cognitive scientist by training, shared it as a discussion of prediction markets, AI-powered forecasting and the future of journalism. He added that journalism had “totally missed out” on an opportunity that could have saved its business model.
Below are the main arguments, in the order he made them.
The media make readers smarter, then waste it
Servan-Schreiber described himself as a scientist from a family of journalists. His father co-founded L’Express and his grandfather co-founded Les Echos. His starting point is that newspapers spend their effort making readers better informed, then get little back beyond letters to the editor. “If you create something valuable and you don’t use that value, it’s a waste,” he said. He quoted a 1990s Hewlett-Packard chief executive: “If only HP knew what HP knows.” A forecasting platform reverses the flow. The paper provides the news, and readers return a collective view of what happens next.
Prediction markets are older than polls
Asked how prediction markets differ from polls, he pointed to history. Polls, he noted, date only from Gallup in 1936. Before that, according to Servan-Schreiber, records survive of betting on Wall Street for fifteen US presidential elections, and the betting favourite won fourteen times. He said newspapers such as the New York Post printed the odds the way they now print polls. He also argued that polls did not prove more accurate than those markets, and that bad polls later made betting markets less efficient.
The deeper difference, he said, is what each method counts. A poll needs a representative sample, because it measures preferences. A forecasting market needs knowledge, so it recruits people who follow a subject closely, whatever their politics. In his words, “here you’re not asked to express preference; you’re asked to express a prediction.” He also called polls a weak technology, because each pollster’s weighting formula is “as secret as Coca-Cola’s recipe”.
Why would people take part? He listed four incentives: rewards, recognition (such as a T-shirt declaring the best forecaster inside Google), relationships within a community of forecasters, and relevance, meaning a channel for what you already think about a subject you care about.
AI can now forecast, so why keep humans?
“We just created a purely artificial forecasting machine,” he said, referring to The Forecasting Machine. A year earlier this had not been feasible, he said, but “now AI can make predictions as well as a diverse crowd of smart humans”. He put AI at about 100 times faster than human forecasters, and much cheaper. Machines also take a question a century away as seriously as one about tomorrow. For horizon scanning around a company, AI is enough. Human forecasters remain necessary when the goal is to engage people, and to organise them around a cause.
Glimt and the intelligence agencies
Glimt is his example of that engagement. It is the platform run by FOI, the Swedish Defence Research Agency, on Hypermind’s technology. Ukrainian analysts pose the questions that worry them, and European volunteers answer. He said the questions range from the war economy (Russian assets, Russian inflation, oil moved by the shadow fleet) to how many missiles will hit Kyiv, which town falls next, and elections in Poland and Hungary. “It’s not about asking AI,” he said. The point is to keep Europeans informed and engaged.
Intelligence agencies, he said, were the first natural customers. He recounted the IARPA forecasting programme that began about fifteen years earlier. IARPA is the US intelligence community’s research agency. In his account, 10,000 amateurs matched the accuracy of professional analysts at much lower cost. About 2% of them were exceptional, roughly 30% better than the analysts, and were paid around $200 a year. He sees such crowds as complementary to agencies, not a threat. A “superforecaster” badge rewards talent, and the Good Judgment Project showed how such people think.
What journalism missed
Nobody would read a financial newspaper if there were no stock market: it would be boring.
Émile Servan-Schreiber, Agenda Pública
The financial press, in his view, thrives because readers act on markets, and he expects “the future of journalism will rely on gamifying the news”. He cited Polymarket’s partnership with The Wall Street Journal. He contrasted Polymarket’s model, which earns money from bettors, with the one Hypermind uses with publishers. There, a publication’s readers forecast as a community, and companies pay for the insights instead of buying advertising alone. “The readership is an intelligence asset,” he concluded. “If we can aggregate it, we can monetise it.”
