Identity is not a name match
Canonical entities require compatible, attributable support; ambiguity can remain unresolved.
Glossary
Precise memory language helps an agent—and the people reviewing it—distinguish evidence from inference, current context from history, and retrieval from authority.
Why terminology matters
A polished sentence can conceal whether context was explicit, inferred, stale, disputed, or out of scope. These definitions describe the intended product and trust model—not a claim that every beta journey is complete.
Canonical entities require compatible, attributable support; ambiguity can remain unresolved.
Current, historical, superseded, inferred, conflicting, and uncertain are not interchangeable states.
Provenance explains support and lineage without pretending that a citation guarantees truth.
Definitions are written for product and implementation discussions. Availability, integration details, retention, and customer-facing correction behavior still depend on the verified beta deployment.
Terms we avoid as shortcuts
“Complete,” “infallible,” “secure,” “private,” “compliant,” and “autonomous” each imply evidence the current beta does not universally establish. Name the source, state, control, scope, and limit instead.
Use uncertain, conflicting, unavailable, or insufficient context rather than confident filler.
Describe isolation, authentication, retention, provider, or deletion facts without blanket assurance.
Distinguish an implemented mechanism, a beta outcome under evaluation, and a future product direction.
Put the terms into context
See how authorization, replaceable contributions, identity resolution, qualification, publication, scoped retrieval, and correction fit together.