A regularly updated forecasting dashboard that uses an ensemble of frontier AI models to estimate probabilities of consequential AI-related outcomes.

The project also investigates how forecasts change under different assumptions about AI capabilities and policy.
Explore AIRO

A platform for collecting and evaluating expert forecasts of social science experimental results. Researchers register their studies, experts submit quantitative predictions, and the platform tracks forecast accuracy over time via a public leaderboard.

Developed as part of a broader research agenda on forecasting in social science. The platform has collected predictions for hundreds of studies and provides data for understanding how well researchers can anticipate experimental findings. Related paper: "Predict Science to Improve Science" (Science, 2019).
Check out the Social Science Prediction Platform

AI-powered feedback on pre-analysis plans and registered reports. Upload a document and get structured comments on research design issues — statistical power, identification strategy, outcome specification, and more — before sending it for peer review.

Built to address a gap in the research pipeline: pre-analysis plans are increasingly required by journals and registries, but researchers rarely get feedback on them before submission. EarlyReview uses AI to provide immediate, actionable comments on power calculations, identification strategies, outcome specification, and other common methodological issues.
Visit EarlyReview