How caselaw works

Every brief starts with a primary judgment. Here’s the pipeline that turns 80 pages of court prose into something a law student can actually use in 30 seconds.

1. Step 1

   Source: The National Archives

   Every full judgment lives on caselaw.nationalarchives.gov.uk, the official UK government repository, free under the Open Government Licence v3.0. We pull the structured XML for each judgment we cover — raw paragraphs, headers, and citations — never anyone’s headnote or paywalled annotation.

2. Step 2

   Extraction: structured brief generation

   We pass the judgment to Anthropic’s Claude with a UK-law-specific prompt and a strict output schema. The model fills in:

   • Material facts — only the ones the decision turns on, no biography.
   • Issues — the legal questions in dispute.
   • Held— the operative answer.
   • Reasoning — the path the judges took.
   • Significance — why this case matters.
   • Obiter dicta — flagged separately (the obiter / ratio distinction is non-trivial in UK law).
   • Cited cases — with the relationship (followed, distinguished, applied, overruled).
   • OSCOLA citation — ready to copy.
3. Step 3

   Verification & iteration

   Briefs are spot-checked against textbooks and headnotes before a case ships. Every brief on the live site has an error-flag link — corrections feed straight back into the prompt for the affected case category. The full primary judgment is one click away on every brief, so students can verify any claim themselves.

4. Step 4

   AI tutor: not a search engine

   The tutor uses Claude Sonnet with a system prompt aimed at UK exam technique. Citations get auto-linkified to our case library so a tutor reply is always one click from the underlying brief. Daily caps prevent runaway use. The tutor is told explicitly: simplified explanations are flagged, uncertainty is flagged, and it never claims to be a regulated legal-advice service.

5. Step 5

   What we don’t do

   • We don’t reproduce textbook content or paywalled commentary.
   • We don’t use your private notes to train AI models. They’re yours.
   • We don’t hide the source — every brief links to the full judgment on The National Archives.
   • We don’t offer legal advice. This is a study platform.
   • We don’t sell, share, or monetise your data.