In October I got an ACX grant to build The Metascience Observatory, a database of replications reported in the scientific literature. I developed an AI pipeline to help with the collection, and the database currently contains over 7,000 reported replications.
The project was founded to get a better idea of reproducibility across all fields of science, with the secondary goal of creating a "reproducibility ranking" for select journals.
I'll discuss what has been accomplished so far, what the challenges have been, and how far we've gotten in realizing these goals.
The Metascience Observatory has two other exciting metascience initiatives - a forensic meta-analysis AI agent to sanity check paper's calculations (based off this book) and data, and Bird's Eye Reviews. I'll be discussing those briefly as well.
Expect a 20-30 minute overview talk followed by open discussion. I'd love to get feedback and ideas on where to take the project next!