A record a model can read and a person can trust.
Legible data is structured data with its story attached: what it says in words, what it said before, and who changed it.
Plain language on the record. Records written through the platform carry a readable version generated from their structured content, so a model and a clinician read the same statement of the same fact.
History is kept, not overwritten. Every version of every record is retained. Ask what the record said at any past moment. The second storage engine also answers what was believed to be true for a period, which is the question a retrospective review or a corrected result actually asks.
Every change is attributed. A change on the clinical interface must name the clinician responsible for it. A device or service credential without one is refused; a patient's own session is the one exception. That rule is what makes acting on someone's behalf meaningful when agents arrive.
The trace below is the shape the accountability work is building toward: one line per read, draft, proposal, approval and applied change, each naming who or what acted and for whom.
Structured and readable at once.
Every record on the platform is structured: codes carry the system they come from, quantities carry their units, references point at the records they mean. The plain-language version is generated from that structure, not written separately, so it cannot say something the structured record does not. A model reading the narrative, a clinician reading the screen and a query over the structured fields are reading one fact three ways.
Two questions about time.
What did the record say on a given day? Every version of every record is kept, so the answer is exact. A corrected result shows both what was first recorded and the correction, each with its moment.
What was believed to be true for a period? A retrospective review or a corrected result asks this second question, which is different from the first: a fact can be recorded on Tuesday about Monday. The second storage engine answers both questions; the current one answers the first. A store asked for an axis it does not support refuses rather than approximating.
Who did this.
A change on the clinical interface must name the clinician responsible for it. A device or a service credential without one is refused. A patient acting on their own record through the portal is the one exception, and their changes to identity details are staged for staff review rather than applied. When agents arrive, this is the hook that makes acting on someone's behalf meaningful: the person is already named on every change, and the agent is recorded as acting for them.
Codes with meaning.
Codes on the record are checked against the loaded code systems and value sets, and each carries its display text. Picklists a practice curates are normalized as a whole whenever they change. A model that reads the record therefore sees a consistent vocabulary with meanings attached, and a rule that asks whether a diagnosis is in a value set gets a definite answer.
Toward one trace.
Version history says what changed and attribution says who; together they are accountability for changes. The uniform trace the platform is building toward also records reads, with the purpose they were made for, and agent actions, with what the agent was given and what it relied on, all in one attachable service with pluggable destinations. Some parts of the platform already emit such records; the single uniform service is direction. The example trace at the top of this page shows its intended shape.