AI-assisted research generates a tangle of instructions, computations, artifact revisions, and checks. When another researcher receives that work, the relationships between these pieces matter for understanding and reproducing it. AIPOCH Open-Science, described in a new arXiv preprint, is an open-source workbench designed to capture and package those relationships.

The paper introduces the workbench and its use of ".science packages" as a reproducible format. By bundling the full context of an AI-assisted research process, the tool aims to make it easier for others to follow the same steps and verify results. The abstract emphasizes that the relationships between components are central to what gets passed along.

Because only the abstract is available in the source, details on implementation and evaluation are not yet covered. The workbench is positioned as a practical response to the reproducibility challenges introduced by AI in research workflows.