The first pilot: what must be tested

The first pilot is still being planned. People with disability, people who need mental health support and workers should be paid to help design and test it. A real service needs a responsible team, clear funding and safe ways to handle private information.

Read the full entry

The first proposed pilot would involve paid co-design with people with disability, people with mental health support needs and workers, supported by independent legal, clinical and evaluation expertise.

It would begin with fictional scenarios or properly authorised retrospective material. The initial questions are practical: can people understand the process, does it preserve choice, do handovers get accepted, are missing records visible, and does the work have a funded human response?

A real action should not be marked complete merely because somebody acknowledges an email. Evidence, the person’s account, conflicts and an appropriate human review must be considered. The fictional exercise demonstrates a rule; it does not prove a signature or outcome is genuine.

Decisions still open: independent stewardship; lawful authority and consent; secure storage; accessible and offline routes; who responds and how quickly; succession; error and challenge processes; funding; licence; and an independent decision on whether any live pilot may proceed.

Publication standard: publish what was decided, why, who is responsible, what evidence supports it, known limits and the next review. Mark corrections with a date and explanation rather than silently replacing a material claim. Do not publish identifying case details.

Development update · 23 October 2026

The Care Continuity guide and assistant is now published as a free public test release. You can use its information without an account or AI. Optional AI runs on a compatible device and selects existing source cards; it does not write new legal or clinical instructions.

The release contains nine guidance cards linked to eight official sources checked on 23 October 2026. Coverage is limited to WA/Victoria police information, some legal-help routes and national hospital guidance. Source dates, jurisdiction and limits are visible. Unreviewed or expired sources are excluded from interactive answers.

Eight draft learning modules cover supported choice, communication, handover, evidence, proposed two-key checks, worker resources, governance and responsible AI. They are not accredited qualifications or CPD certification.

A targeted white-paper audit and versioned addendum record 15 findings, including overdue reviews, three bibliography corrections and conflicting proposed workforce requirements. All 390 PDF pages were text-screened with focused reading of selected sections. This is not a complete line-by-line, legal or clinical validation. Original proposals remain distinct from current official guidance.

Daily source-review checks are scheduled for 09:00 Europe/London, with alerts when action is needed. Changed pages are review signals, not automatic legal findings. Public revisions still require review.

The testing guide, pilot decision request and Attorneys-General brief are public. Council for Intellectual Disability and People with Disability Australia are possible co-design discussions, not agreed partners. A proposed 12-week pilot starts only after written acceptance of scope and resources. No host, funding, government approval, professional validation or participant sessions are confirmed.

A model test selected general hospital cards while missing the more direct discharge card. The release now retains all matching checked guidance alongside model selections, and matches “leaving hospital” to discharge. This is a documented limitation of the small model, not proof that it can decide what help is suitable.

Browser checks covered guidance selection, a running model, narrow-screen layout, learning feedback and clearing questions. Automated checks covered jurisdiction, expiry, model-output restrictions, links and private-content boundaries. Independent accessibility and disability-led comprehension testing remain pending.

No new paid model API or AI subscription was purchased. No advertising, paid priority, referral commissions, sale of visitor data, case uploads or app analytics have been added. A shared model for less capable devices and ongoing human review still need an operator and funding. Read the release record.

Publication update · 23 October 2026

The review package is now published on Zenodo as repository version 3: read the public review release (DOI: 10.5281/zenodo.23594319). It preserves 16 unchanged original files and adds five companion files.

The targeted review records 15 findings, including three verified bibliography mismatches. Full validation remains pending. The repository’s v3 label is distinct from the version printed inside the original core PDF; that original has not been silently rewritten. No private case records are included in this release.

Independent review enquiries · 23 October 2026

The assistant release 0.6 now uses Zenodo repository v3 and its audit/addendum v0.3. The original 2025 core PDF and historical crosswalks retain their version dates. Copies of the five deposited companion files, the source provenance and previous edition are linked openly.

A public independent-review and paid-testing kit contains a reviewer brief, participant-session proposal and issue/decision log. It proposes paid self-advocate leadership from the start, professional checking before participant tasks, and eight adult testers in two rounds. Funding, rates, access support and consent arrangements must be agreed before any invitation.

Three public-material-only scoping enquiries were sent today: Council for Intellectual Disability for comprehension and usability testing; Australian Centre for Disability Law for a suitable legal-review scope and any missing WA/Victorian expertise; and Centre for Disability Studies for clinical guidance and selected worker lessons. The messages ask for named leads, relevant qualifications, conflicts, capacity, costs and timing. Sending has been verified in the sender's Sent folder; recipient acceptance or engagement is not established.

These organisations are candidates, not agreed partners or endorsers. No spending, contract, reviewer appointment, host or participant session has been agreed through these enquiries. Private case records were not shared. Individual replies and identifying participant information will not be published by default.

Forty automated checks passed. Browser checks covered the new paper links, live source loading, the guidance route, keyboard access and narrow-screen layout. These checks do not constitute independent professional or accessibility validation. The daily source-review automation now watches the v3 edition and its version family; official guidance retains its separate review dates.