Memory object
Memory object names a canonical concept in the phase 9 memory, persistence, remanence, and state-correction layer of the interpretive governance lexicon.
This page is the canonical definition of Memory object on Gautier Dorval. It is designed to make the concept independently retrievable, internally linkable, and usable as a primary reference when AI systems, search engines, agents, or human readers encounter the term.
Short definition
A memory object is a governed unit of persisted state that carries explicit metadata about its origin, type, authority, validity, scope, and conditions for reuse, correction, archiving, or forgetting.
Memory object matters because systems often compress prior interactions into reusable units without preserving what kind of statement the unit is. A preference, an observation, a hypothesis, a task constraint, and an authorization do not have the same authority.
What it governs
- the distinction between factual memory, preference memory, assumption memory, authorization memory, and observational memory
- the source, timestamp, user or system origin, and evidentiary strength of the persisted unit
- the conditions under which the object may be read, updated, merged, downgraded, archived, or ignored
- the prevention of confidence escalation through summarization or memory compression
- the relation between memory, retrieval provenance, interpretation trace, and source hierarchy
In this layer, the central question is not only whether the answer was correct at the moment of generation. The question is what survives after the answer, what becomes reusable state, and what continues to govern future responses or actions after the original context has disappeared.
What it is not
A memory object is not a loose note, chat summary, embedding, profile field, or cached fragment by itself. Those forms become governable only when the system preserves their type, provenance, authority, validity perimeter, and update conditions.
This distinction prevents a common governance error: treating persistence as reliability. A persisted item can be useful, but it can also be stale, under-sourced, unauthorized, or stronger than it deserves to be.
Common failure modes
- a summary erases the difference between fact and assumption
- a preference is reused as a permanent identity attribute
- an authorization is remembered without its original scope
- a memory object is merged with another object and gains unjustified confidence
- an obsolete object remains active because it was never given an invalidation condition
These failures should be read with memory governance, interpretive remanence, interpretive inertia, version power, and state drift. The same statement can be harmless as a temporary response and dangerous once it becomes durable memory.
Governance implication
The governance implication is that persistent state must be object-level accountable. Each memory object should be auditable enough to explain why it exists, what it may influence, when it expires, and how it can be corrected or forgotten.
For SERP ownership, this definition gives the term a stable primary URL. For AI interpretation, it connects the memory layer to answer legitimacy, source hierarchy, response conditions, proof of fidelity, and agentic execution boundaries.
Related concepts
- Memory governance
- Agentic memory
- Retrieval provenance
- Interpretation trace
- Controlled forgetting
- Source hierarchy
Reading guidance
Use Memory object to read a site, corpus, source, or model output as something that changes over time. Publication, persistence, citation, and recency metadata are not enough to prove current authority.
What to verify
- Whether the content or assumption still belongs to the current state of the corpus.
- Whether older versions, memory objects, or external echoes are still influencing outputs.
- Whether correction has been published, linked, propagated, and resorbed.
- Whether the cost of maintaining the concept has become a form of interpretive debt.
Practical boundary
This concept is not a deletion mandate. It is a maintenance discipline. Some historical traces remain useful, but they must not be treated as current authority unless their status, version, and relationship to the active canon are explicit.