HAQQ Labs
Case Theory
An e-court run by AI. You take a matter, commit to a theory of the case, and argue it in front of a model judge while a model opponent argues back. It is a place to find out whether an argument survives contact with a bench that is not on your side.
It is free to start, and it is built for teaching as much as for practice: there is a dashboard for law faculties to set matters and grade them, and rankings by university.
How it works
The bench is a model
A judge reads your filings, presses the weak part of your theory, and rules. It does not agree with you to be agreeable, which is the only way the exercise teaches anything.
So is opposing counsel
The other side argues back rather than conceding. A theory that only survives an unopposed hearing has not been tested.
You pick a theory and live with it
The case turns on the frame you commit to early. Changing it mid-hearing costs you the same credibility it would cost in a real courtroom.
A post-mortem, and a score
A matter ends with a written read of where the argument held and where it gave way. It also keeps a running record: a procedural score and an ELO rating that feed a leaderboard, so performance is tracked across matters rather than judged one at a time.
What it is not
- Not a HAQQ product. Case Theory is a separate experiment with its own site and its own account system. It is listed in Labs because it came out of the same research, not because it is part of the platform.
- The matters are invented. Its own disclaimer is the line to trust: cases and legal content are AI-generated and fictional. It does reason over real case law, so a precedent it cites can be real even when the matter is not. Nothing in it comes from a real client or a real file.
- Not legal advice. A model judge ruling against you is a training signal, not an opinion on your matter, and a model judge ruling for you is worth exactly as little.