Mission

The mission

To give the people who run 24-hour care the instruments to govern the machines that join their teams — published open access, so that “we couldn’t govern it ourselves” stops being a reason to rent. Get the human team in order first, because the team is what the machine integrates into, and whatever it integrates into, it amplifies.

Why culture is the term that matters

Every current regime for governing AI — conformity assessment under the EU AI Act, device regulation under the MHRA, clinical risk management under DCB0129 — assesses the capability of the system. None assesses the culture of the organisation deploying it. An agentic system does not install into an organisation: it amplifies the organisation, at machine speed and with perfect fidelity, after first deleting the informal human layer that has always damped organisational error. So deployment risk is not capability plus culture. It is capability multiplied by culture.

A licence to deploy is a licence to amplify. It should be granted on evidence about what will be amplified. That is the case made in The Cultural Safety Case, and it is the reason the rest of the estate exists.

Where it starts: the team

The estate is written by a registered mental health nurse from sixteen years of clinical practice, for the people who run 24-hour care. The 24-Hour Team argues that a 24-hour service is one organism with two phases, not two teams, and that the shift handover is at once its highest-risk safety event and the main place its culture is passed on. The Mixed Shift follows that team as it acquires members that do not sleep, do not gossip, do not take breaks, and do not die. The Long Handover makes the largest claim: care is not lagging the future of work but running its first full-scale trial, and the failures documented in care are previews of failures now visible at national scale.

What the estate does

The papers argue; the instruments do. The Reading Map sets out the order of operations: build the team → clear the gate → hire the mind → seat the overseers → measure the floor — and for the person in care, assess → measure → record. The instruments are bound in the Instrument Deck in the order a deployment actually meets them — gate, screen, specify, oversee, trial, audit, measure, record — or, in the Deck’s own words: assess at the gate, measure over time, record beyond amendment. Each section opens with a divider naming its parent document and DOI.

Five standing rules, as the Reading Map states them, travel with all of it:

• Lenses, not labels.
• Machines report observations; humans assign meaning.
• No accountability without authority.
• Every machine behaviour has a return address.
• Signals, never targets.

Who it is for

Anyone who is about to deploy AI, or is being sold it. The manager running a home, ward or 24-hour unit. The person who has just been told they now “oversee” an AI system. Whoever is buying or governing AI deployments, and whoever signs a clinical safety case. The organisation choosing, renting or leaving a care records system. The people building AI for care or crisis support, or building an agent’s memory or audit record. Whoever is responsible for a person in care who uses AI, or needs an audit trail that survives inspection. The Reading Map gives each of them a place to start.

Why it is open

For twenty years, renting was the rational choice for a care organisation, because three moats stood between it and owning its own systems: build cost, operations, and governance. Governance was the quietest and the strongest — we could never write the safety case ourselves. The instruments a care organisation needs to deploy and govern its own system are now written, published and open access. Free to take, adapt, cite, and deploy against. The last respectable excuse for tenancy — we couldn’t govern it ourselves — is now a download.

That is not a claim that ownership is free. The operations moat still stands: running a system of record safely for years is real work, and any paper telling you ownership is free is selling you something. Nor is it a claim that every subscription is a tenancy. Rent the sockets; own the gravity. The argument is in The Tenant’s Record, and what it means in practice is on Sovereignty. Why this approach holds, and why the gate comes first, is set out in The Moat.

How it stays honest

Many of the papers state the tier of evidence behind their claims, and say plainly when a claim is reasoned rather than validated: The Damping Audit describes itself as “internally coherent, unvalidated”. In The Floor Test’s words, the frameworks “are argument and synthesis until they meet a floor”. To make the argument falsifiable, the estate registers predictions in advance — four in The Cultural Safety Case, four in The Second Pair of Hands and three in After the Machine. The Floor Test is a ninety-day protocol that tests six of the series’ claims on a real unit, with the predicted movements, and what would falsify each, stated in advance. A claim is not a control until it can fail. If a claim dies on enough floors, it dies in print, with thanks.

What I commit to

• Every record stays open access under CC BY 4.0, with the Glass Ledger’s code under MIT, each archived on Zenodo under its own DOI.
• Field returns, corrections and null results are welcomed. The series improves by versioned revision, and contradicting results are published with thanks.
• Nothing on this site is for sale. Enquiries are read and answered by me, not a pipeline.

Start with the argument, or go straight to the instruments. If you want to talk about a deployment, write via the contact page.