Models can predict future events and make money on Polymarket now?
Summary
Researchers at the Max Planck Institute introduced FutureSim, an environment where AI agents predict real-world future events by replaying historical web data. GPT 5.5 running in Codex achieved near-perfect Brier skill scores on some Polymarket markets like Super Bowl LX, outperforming human aggregate markets, though it struggled on others like UK elections and the Grammys.
Similar Articles
FutureSim: Replaying World Events to Evaluate Adaptive Agents
FutureSim replays chronological world events to benchmark AI agents' long-term predictive abilities, finding that even the best agent achieves only 25% accuracy.
How AI Will Save Prediction Markets (10 minute read)
The article examines the failure of prediction markets to achieve their utopian vision and argues that AI can transform them by enhancing market creation, analysis, and accuracy, shifting focus from sports and crypto to high-impact questions.
Prediction markets are breaking the news and becoming their own beat
Prediction markets are increasingly influencing news coverage and becoming a subject of journalism in their own right, as platforms like Polymarket gain mainstream attention for forecasting real-world events.
Looking at the data behind prediction markets
An analysis of prediction markets like Polymarket and Kalshi, examining whether their massive trading volume actually produces valuable forecasting information or merely serves as gambling, referencing historical academic support and current data.
@nicos_ai: GOOGLE HAS SILENTLY RELEASED AN AI THAT PREDICTS PATTERNS Sales. Market prices. Web traffic. Energy demand. Crypto vola…
Google has released TimesFM, an AI model for zero-shot time series forecasting, trained on 100 billion real data points, free and open-source.