The Replication Engine is a proposal by Abel Brodeur, chair of the Institute for Replication, and Bruno Barbarioli, a research scientist at the institute, for an AI system that would automatically reproduce the computational findings of a paper when it is published. Four AI agents would work in sequence, one parsing the paper’s results, one checking that the authors’ code runs and gives the reported output, one looking for coding errors and data irregularities, and one testing whether the main results hold under reasonable robustness checks, with each component given a green, amber or red badge. The Institute for Progress published the proposal on 11 August 2025, and it sets out a three-year first phase costing about $10 million, with targets of 2,500 papers in the first year and 15,000 verified papers by the third.
Notes & quotes
- The proposal says results would go to editors and authors before publication and be visible to readers afterwards, and that two public knowledge graphs would track unverified claims and fragile research clusters.
Publisher / company: Institute for Replication
Date of case study: 2025
Status: Not stated
Built by: Research, non-profit or individual
Sources: