The Journal
Research notes.
Notes from Healthattica's programme — on the substrate, the governance model, and the research that grounds them. Entries are published as the work progresses.
Methodology · 25 May 2026
Why the knowledge layer belongs underneath AI, not alongside it
On the difference between attaching knowledge to a model and grounding inference in a substrate the model is bounded by - and on the discipline that makes the substrate itself inspectable.
Methodology · 24 May 2026
Provenance at generation time: what an auditable AI output actually looks like
Reconstructing why an inference happened after the fact is forensics, not provenance. Real provenance is captured at generation time, in the same call that produces the output - and the difference between the two is what makes an AI defensible under audit.
Technical · 23 May 2026
SHACL conformance: what we ship and why
A substrate's claims about its own structure are only as trustworthy as the consumer's ability to test them. Conformance shapes are how we make those claims testable - by the consumer, on demand, against the release they are actually consuming.
Regulatory · 20 May 2026
The regulator graph: resolving MoHAP, DHA, and the same body under three names
A regulator's name appears half a dozen ways across documents, vendor decks, and submissions. Resolving them to a single canonical entity - with country, scope, and authority as first-class - is what makes a regulatory landscape queryable instead of just searchable.
Clinical · 6 May 2026
Cross-mapping ICD-10 to SNOMED CT: the 17% that needs a rationale, not a rule
Most clinically useful mappings between ICD-10 and SNOMED CT can be expressed as deterministic rules. The remainder is where the engineering ends and the curation begins - and where the substrate stops being a lookup table and starts being a defensible artefact.
Technical · 15 April 2026
Ontology vs taxonomy: why the difference is load-bearing in healthcare
A taxonomy classifies. An ontology classifies and relates. In healthcare that gap is where contraindications, drug interactions, and pathway logic live - and it is what most 'knowledge graphs' silently miss.