A public test release · search refinement prepared 27 October 2026
Find source guidance or a passage from the paper
Read or print source guidance and paper passages. Optional on-device AI can help select cards from the checked library.
You can use this guide yourself, or with someone you choose. You do not need an account or subscription.
Read or print all source guidance · Read or print all lessons · Simple start · Accessibility information · Transparency journal
Portable website copy prepared 25 October 2026. The source guidance, review dates, paper versions and release holds are unchanged. Download the guide and public review documents for offline use (ZIP). Optional AI still needs its model download and a compatible browser.
What is happening?
Ask in your own words. We start with practical checks drawn from experience, then help you find relevant paper sections, crosswalks, templates and source guidance. Search covers the full white paper. Current-guidance cards cover selected WA and Victoria police/legal routes and national hospital information. This guide has no live web search.
Browse the paper’s headings · Browse all crosswalks · Go directly to help routes
Need a person, another state, or a situation not covered?
You can still find a human support route and prepare what to ask next. You do not need to get a matching answer from this guide first.
Find a person or support route · A service cannot help, or no one has accepted
Optional free AI on this device
The AI helps choose relevant cards from this checked library. It does not write new legal or clinical instructions. The source wording and limits stay visible.
Loading downloads a model from Hugging Face and a WebLLM runtime from its published GitHub release. Expect hundreds of megabytes of data and about 1 GB of graphics memory. Your browser may keep the model in its cache. Loading can take several minutes and may not work on older devices.
The model runs on your device. Your question is not sent to a model provider by this app. Network services used to deliver this website and download the model receive ordinary connection information. The guide works without loading AI.
AI has not been loaded. Source guidance is available now.
Test model: Qwen2.5 0.5B Instruct, MLC quantisation · WebLLM 0.2.85. Model information and licence. A small model can select poorly; check the source cards and try the ordinary guide too.
Learning across the organisation
Eight draft lessons
Eight short draft modules include scenarios and questions for workers, supervisors, managers, executives, health/emergency staff and legal decision makers. These are learning activities, not accredited qualifications, clinical protocols or CPD certification.
No learner account, score record or certificate is created. You can print the lessons. Download the module source.
Sources, check dates and review limits
Official guidance and the white paper are different source types. Editorial checking records what a source says; it does not establish that a particular legal or clinical action is right for you.
Public reference library
33 white-paper crosswalks have been indexed into 417 passages. These are reform proposals awaiting validation. Their stated review date was 20 December 2025. They are kept separate from the current-guidance answer cards.
Search and read the full white-paper library
You can save the new library page for offline text reading and search. The older portable guide ZIP does not include this new library; original PDF and archive downloads need a connection unless saved separately.
Read the original 2025 core by page, the 33 crosswalks by paragraph, and the deposited files and template archive. These are historical reform proposals. The 2026 corrections and review status remain visible; source dates are unchanged.
Privacy, funding and model choices
No advertising, paid priority, referral commission, sale of personal data or visitor subscription is part of this test release. No private case records are included. This app has no upload or microphone feature and no analytics. This app does not write a saved question or quiz history. It keeps current entries in page memory. Clear question and results removes the question and results. Your browser may restore page or form state; the model may remain cached.
Existing publishing infrastructure is used. No new paid AI subscription or hosted model API was purchased. A reliable shared model for less capable devices still needs an agreed operator and funding. Public testing, ongoing human review and staffing are not funded or booked.
We use a small on-device model as an experiment. Its selections are checked against the displayed, jurisdiction-filtered card IDs. It cannot add a service, source or factual instruction to the answer. This reduces some risks; it does not prove the recommendation is suitable.
Earlier release record · Source availability check · Plugin readiness review
New papers and a connected project.
24 October 2026: QIP and Rails for Consequence have published 2026.10 companions. QI is still under review and is not yet public. The disability white paper remains v3. These papers inform proposed designs; they are not added to the official-guidance answer cards.
Read the releases and their limits Meet Future You · Pain Card project
Meet Future You is a separate financial-learning project. Check the current project page for its release, language and review status.
Open the white-paper reading library
The white paper: start with the corrections.
Published review release v3 · 23 October 2026. This assistant now uses the new Zenodo release (DOI 10.5281/zenodo.23594319), including audit and addendum v0.3. The original core PDF remains v1.0, dated 25 September 2025. Its 16 historical files are preserved unchanged.
The review records 15 findings and three verified bibliography corrections. It is a targeted editorial and source review. Complete legal, clinical, economic, citation and accessibility validation remains outstanding. These reform proposals are separate from the official guidance used in answers.
Editable review sources · ZIP · Zenodo release record · File checksums · Assistant source provenance · Previous repository edition
The same 33 historical crosswalks and their original review dates are retained. A new repository date does not make them current law or prove that the proposed safeguards work. Daily source monitoring does not silently alter a deposited edition.
Proposed pilot and testing
The proposed first pilot combines three practical pathways, eight learning modules and a transparent evidence review process. People with disability should help choose and change it now, with payment and access costs agreed before participation.
- Agree an independent host, a named lead, participant support, scope and funding.
- Test fictional tasks, source accuracy, accessibility, worker burden and what happens when the system is wrong or unavailable.
- Only consider a live service after an explicit, independently reviewed operating decision.
A proposed 12-week period starts after written acceptance of a funded scope. No host, organisation, Attorney-General or government has approved this pilot. Nothing here is a ministerial order or legal representation.
Independent review and paid testing: start here
Possible first discussions: Council for Intellectual Disability for co-design and usability testing, and People with Disability Australia for broader participation/governance. Neither is an agreed partner.
What we will check in a test session
- Can the person find a useful route and understand its limits?
- Does each claim have a source that supports it, in the correct jurisdiction?
- Can the person distinguish existing guidance from a proposed reform?
- Can someone use keyboard, screen reader, zoom, print and a low-bandwidth/no-AI route?
- Do the lessons improve a scenario response without adding unnecessary worker burden?
- Are errors corrected, dated and checked before a new release?
Sessions, participants, payment and professional reviewers are not yet arranged. The test guide is preparation, not an approved research protocol or a signed consent.