Glossary
The words behind the context layer
Short definitions, written to stand alone. Each links to where the idea is explained in full.
C
- Clinical context layer
- A persistent, continuously maintained model of how a patient's facts relate to one another, from which applications draw the context they need. It proposes; a clinician decides. The definition →
- Concerns and constraints
- The objectives, thresholds, policies and judgments that clinicians and other responsible people author. In a context layer they sit inside the model and shape what counts as significant. The whitepaper →
- Context graph
- A single queryable model of the entities in a patient's care and the relationships between them, rather than a set of disconnected reports. What it is →
- Context layer
- A persistent, semantically integrated information substrate that maintains the evolving relationships needed to build meaningful context windows. It exists independently of any one inference. The whitepaper →
- Context window
- The information made available to an AI model for one reasoning task. It is assembled for a single inference and is temporary. Context window vs context layer →
- Continuous observation
- Monitoring that does not stop between visits, for example through home devices or remote monitoring. It produces a steady stream of signals instead of isolated episodes of data. Why now →
D
- Domain ontology
- A representation of the main entities, relationships and distinctions through which a domain is understood. It gives a context layer its semantic foundation. The whitepaper →
E
- Event-driven contextualization
- Updating the context model each time a new event or observation arrives, interpreting it against what is already known. The whitepaper →
L
- Living context model
- The evolving representation inside a context layer. It is updated as new signals arrive, so each cycle builds on what earlier cycles established. How it works →
P
- Proposes; a clinician decides
- The rule that the system may surface and propose, but never asserts a clinical fact on its own and never automates a clinical action. The clinician decides. How it works →
R
- Retrieval-augmented generation (RAG)
- A technique that retrieves information fragments similar to a query and gives them to an AI model. It finds similar information, but does not itself establish how the fragments relate. Context layer vs RAG →
S
- Subclinical signal
- An observation too small or too tentative to count as a finding on its own, such as a slow drift in a measurement. Alone it may mean little; together with others it may matter. Article: when the danger sits below every threshold →
- System of information
- A system that holds the settled record, subclinical signals and the connections between them, and keeps uncertainty and nuance instead of resolving it away. What it is →
- System of observation
- A system that answers one specific question and then goes quiet, such as imaging, labs, diagnostics or monitoring. What it is →
- System of record
- A system built to keep a structured, auditable record of what happened, such as the electronic health record. It preserves what was decided, not how the facts relate. Systems of record and context →
Last reviewed
Want to go deeper?
Talk to the founders about where the context layer fits in your program.
Talk to the founders