Purposing · P1 Diagnose · AI assistant
AI Diagnose — the pandemic assumption radar
Paste a pandemic preparedness plan (or an excerpt). The assistant surfaces the hidden assumptions the plan treats as certain — especially the structural silences, the things it never says — and returns them as a draft Assumptions Register ready to be windtunnelled. It does not predict the next pandemic and it does not build scenarios; it only does the Diagnose step of Purposing.
The eight pandemic assumption categories
Nature of the threat
what the next pathogen will be like (respiratory, familiar, flu-shaped)
Detection & surveillance
that we will see it early enough to act
Community trust & behaviour
that people will comply as they did before
Pharmaceutical sovereignty & supply
that vaccines, drugs and supply chains will be there
Governance & authority
that decision rights and escalation triggers are clear
Health-system capacity
that surge, staffing and beds will hold
One Health integration
that human, animal and environmental signals will join up
Security framing
that a biological event is a health event, not a security one
When an ANTHROPIC_API_KEY is configured on the deployment, this calls Claude (model claude-opus-4-8) with strict tool use to return a validated register. Without a key it runs a transparent keyword-based demo, clearly labelled, so the feature still works. Surfaced assumptions are a starting point for human review, not a verdict.