About

FedCourtsAI is the first project of ModelMirror, a small independent effort to evaluate frontier AI models on forward predictions about the real world, where the answer key is written by events rather than by us. The Supreme Court is a good first subject: the docket is public, the questions are well defined, the outcomes are unambiguous, and legal expertise is about as specialised as expertise gets.

How it works describes the method and its limitations; the ledger is the record itself; the code, the data, and the design documents are all in the public repository. For the wider project, and for other work under the same umbrella, see modelmirror.ai.

Contact

• Questions, corrections, and bug reports: GitHub issues
• Feedback, press, collaboration, and anything that doesn’t belong in a public issue: contact@fedcourts.ai
• Updates: @FedCourtsAI, or by email below

Get an email when there's a new release

Sign up

2–4 emails a year, around releases. Unsubscribe anytime. How your address is used.

Not legal advice

Everything on this site and in the repository is an experimental model output. It may be wrong. It carries no affiliation with or endorsement by any court, and it is not advice about any case, filing, or decision. See the terms.