Ask a justice question.
Ask in plain language. The Matrix returns cited cases, campaigns and evidence, with the source, limits and review status visible.
Keyword and semantic search for people who want to inspect every result.
Issue pages explain the question, linked law, campaigns and evidence.
Review-ready OHCHR material, status notes and source matrices.
Send a missing case, source link, campaign, correction or note.
Open the records, maps, guides and source queues behind the AI response.
The opening question box is the primary interface. These are the deeper layers when a user needs to inspect, browse, export, correct or add information.
Strategic litigation across courts worldwide, with outcome, holding and precedent strength.
OpenAdvocacy work and tactics, with the organisations leading them.
OpenAustralian youth-justice research and evaluations of what works, with consent respected.
OpenSearch the corpus geographically: recorded coordinates first, then labelled jurisdiction centroids.
OpenCompare youth remand across law, systems, campaigns, funders, countries, and consented stories.
OpenPublic NJP / OHCHR review route with the status brief, UI plan, background paper, and source matrices.
OpenCurated issue pages for the questions people ask again and again.
The LLM can answer in simple language; the wiki layer gives reviewers a stable page with cases, campaigns, evidence and source context gathered under one question.
Landmark records the answer engine can cite.
Curated entry points for each surface: refugee and asylum law across borders, and the Australian youth-justice record. These records are the evidence layer beneath plain-language answers.
Or start from a question: browse all issuesLLM first, evidence underneath.
A visitor should not have to know whether they need a case, campaign, treaty, report, source or guide. They can ask a normal question first, then follow the citations into the deeper Matrix.
The Justice Matrix maps, classifies and connects strategic litigation, advocacy, and the Australian evidence base. The ask layer retrieves Matrix records first, generates a cited answer second, and keeps the boundary clear: research support, not legal advice.
Sources and corrections flow through a review queue before they become answerable knowledge.
29 active feeds: court databases we scan (HUDOC, CourtListener, CJEU), the EDAL aggregator, and civil-society networks, via cron + a Playwright/LLM CLI. BAILII, AustLII, and UNHCR Refworld are cited as link-outs, not scanned feeds.
Every scanned item lands in a curator queue, not the live matrix. An admin sees the candidate beside its source, edits fields, and approves, rejects, or marks duplicate. Nothing publishes without approval.
Approved items become case, campaign, or evidence profiles, connected by semantic similarity so any profile surfaces the related cases, campaigns and evidence around it.