The AI Superforecasters Are Here

· Source: Astral Codex Ten · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Finance & Economics · Depth: Intermediate, extended

Summary

AI superforecasters, typically frontier models like ChatGPT or Claude enhanced with "scaffolds" for research and subagent deployment, are now achieving performance levels comparable to top human superforecasters. Startups like FutureSearch and Preseen demonstrate this capability, with one founder claiming to turn \$35 into \$2 million on Kalshi in seven months, and another beating the stock market by 25%. While out-of-the-box AIs like Gemini 3.1 were approaching the Metaculus Community Prediction in May 2026, scaffolded AIs are estimated to be nine months ahead, suggesting human-AI parity in about six months. Recent Metaculus Cup tournaments show AIs like Preseen's bot placing third, indicating a statistical dead heat with humans. AIs offer advantages in speed, diligence, and potentially reduced bias, especially in data-heavy domains like finance, where Preseen's bot recently won Metaculus's Market Pulse competition.

Key takeaway

For strategic decision-makers evaluating future market trends or project viability, AI superforecasters now offer highly accurate, scalable predictions comparable to human experts. You should explore integrating these tools, such as FutureSearch or Preseen, into your analytical workflows to gain rapid, data-driven insights. Prepare for a future where off-the-shelf AI chatbots provide superior probabilistic opinions, transforming financial analysis and strategic planning by reducing the cost and time of high-quality forecasting.

Key insights

AI superforecasters, enhanced with "scaffolds," now rival top human experts, democratizing access to high-quality, unbiased predictions.

Principles

Method

An AI superforecaster uses a "scaffold" program to guide frontier models (e.g., ChatGPT, Claude) through a research process, employing prompts, tools, and subagents to generate precise probabilistic forecasts.

In practice

Topics

Best for: Entrepreneur, Director of AI/ML, AI Product Manager, Investor

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Astral Codex Ten.