First Forecasting Dojo Group Meetup

The LessWrong community has launched its first Forecasting Dojo Group meetup, marking a new initiative to systematically improve prediction skills within the AI and rationality communities. Organized by Vojtech Brynych, the inaugural session on March 1, 2026 (11:00-12:00 CET) will be conducted via Discord video call and follows the structure outlined in Brynych's original group invitation post from earlier this year.

The meetup employs a structured practice methodology combining calibration exercises with 'pastcasting' sessions using the Sage forecasting platform. During pastcasting, participants make predictions about historical events where outcomes are initially hidden, then immediately review results and discuss reasoning. This approach provides rapid feedback loops essential for skill development. The format requires no preparation and welcomes participants at all skill levels, from beginners to experienced forecasters.

This initiative emerges from growing recognition within AI safety and forecasting communities that systematic practice is necessary to improve prediction accuracy on critical questions about AI development timelines, capabilities, and impacts. The Forecasting Dojo represents a practical implementation of evidence-based learning techniques for probabilistic reasoning. Future sessions may expand to include live forecasting on current events, prediction market analysis, and specialized drills for common cognitive biases affecting AI-related predictions.

The group's formation reflects increasing professionalization of forecasting within tech circles, particularly as AI companies and researchers seek better methods to anticipate technological developments. Similar forecasting practice groups have demonstrated measurable improvements in participants' accuracy in other domains, suggesting this approach could enhance the community's ability to make reliable predictions about AI trajectories, safety challenges, and societal impacts.