Glimmer

Reproducible AI science, as a navigable knowledge graph.

Glimmer turns a research project into a graph you can explore, run, verify, and extend — datasets, methods, experiments, findings, and publications, with every result carrying a provenance trace you can check. Each study on Glimmer is its own knowledge graph, and studies link into a wider research graph.

What makes it different

• Research as a graph, not a folder. Datasets, methods, experiments, findings, concepts, and papers are nodes with typed relationships you can navigate. See How it works.
• Provenance you can check. Every derivative points back at the method, inputs, and code that produced it.
• Public by default at the surface, gated where it counts. Each project exposes a required public summary — what it studies, its scale, its references — while raw data stays members-only.
• An agent per project. A public site agent answers questions about the platform and any open project; each study runs its own agent grounded in that study's graph. See Agents.
• Compute on request. Studies can request compute to run analyses, with the cost and provenance tracked. See Architecture.

Start here

• How it works — the knowledge graph and the explore hierarchy.
• Projects — the studies live on Glimmer right now.
• Bring your project — stand up your own reproducible study.

Live site: glimmer.science (opens in a new tab) · Explore: glimmer.science/explore (opens in a new tab)