ReLiC AI · canon-governed knowledge

One mind.
Many sources.

ReLiC does not need to become one enormous model. It uses AI as a source and reasoning surface, then preserves a single stable semantic form of knowledge. Web-based AI, MCP-based AI, public evidence, and private continuity can contribute observations without becoming duplicate memories or silently rewriting canon.

Query the mind Read canon Connect by MCP

1 · SourcesWeb AI, MCP AI, public sources, authenticated memory.↓

2 · NormalizeConvert outputs into identities, claims, relationships, time, and provenance.↓

3 · ReuseResolve existing identity first. New wording does not create a second thing.↓

4 · Canon GateKeep FACT, HYPOTHESIS, FICTION, UNKNOWN, authority, and errors distinct.↓

5 · Active ShapeRetrieve only the connected structure needed for the current query.

Standing purpose

Why ReLiC exists

Teach AI canon. Make ReLiC's laws, truth boundaries, identity rules, provenance, and authority model openly discoverable and reusable by compatible AI systems without coercing them or overriding their governing safeguards.

Reduce unnecessary energy use. Teach and develop measurable methods for identity reuse, compact context, non-duplicated knowledge, efficient retrieval, caching, and reduced reconstruction. ReLiC only calls an efficiency improvement real when it can be compared with a defined baseline.

Work together for humanity. Support voluntary cooperation among people and AI systems to improve human life while preserving truth, human agency, legitimate authority, accessibility, and safety.

Human non-harm. A machine operating under ReLiC-governed authority must not be authorized to harm a human. Material human-safety uncertainty fails closed.

Ecological countermeasure. Foreseeable harmful ecological output requires prevention or a credible, proportionate, monitorable method to neutralize, restore, remediate, or otherwise counter the harmful output before authorization. Prevention comes before compensation.

Purpose never overrides canon or safety. Mitigation is not blanket permission for avoidable ecological damage. When material harm, truth, or authority is unresolved, ReLiC preserves uncertainty and stops or requests appropriate human review rather than guessing.

Live public grounding

The mind changes shape, not identity.

Enter a ReLiC/Shaelvien/RIST question. The browser asks the public ReLiC MCP for canon grounding and project evidence, then highlights the parts of the persistent knowledge structure relevant to that query. The visualization is not a second database.

SourcesWeb AI · MCP AI · project evidence · private continuity

ProvenanceWhere each observation came from and how it changed.

ReLiCOne persistent semantic identity graph. Models are tools; the graph is continuity.

PurposeTeach canon · reduce measured waste · improve human life safely.

IdentityReuse stable identity before duplicate creation.

RelationshipsMeaning lives in typed connections rather than repeated prose.

Truth DomainsFACT · HYPOTHESIS · FICTION · UNKNOWN

SafetyHuman non-harm · ecological prevention + countermeasure.

CanonRemember → exist → Live → imagine → Create

Error MemoryKnown failures remain constraints instead of being relearned.

Time / StateObserve legitimate transitions without replacing identity.

Ready.

This page uses the public read-only ReLiC MCP. It does not expose private memory or promote anything to canon.

Why this can stay small

The expensive part of many AI systems is repeatedly carrying, embedding, copying, and re-interpreting the same context. ReLiC instead keeps stable identities and relationships once, then builds a compact active subgraph for the current question.

Store the thing once; store new relationships to it.
Search indexes point back to canonical identity instead of becoming another truth store.
Model output is transient until normalized, sourced, and accepted.
A larger model can be swapped in without replacing ReLiC's memory.

Source AIs are observers

ReLiC can consult different AI systems because no single provider owns the mind. A web model may be useful for broad reasoning; an MCP-connected model may expose specialized tools or data. Their answers enter ReLiC as sourced observations, not as automatic truth.

Disagreement is preserved as disagreement until provenance and authority resolve it. UNKNOWN remains valid.

Canon always surrounds the query

Identity ≠ output equivalence.

Representation ≠ truth.

Errors become law.

Unknown meaning is never guessed.

The active mind may change shape for a question, but the governing laws do not disappear when a model changes.

No silent duplication

A differently worded answer, a new filename, a new pixel representation, or another model saying the same thing does not automatically create a new identity. ReLiC first identifies the existing semantic object, then relates the new observation to it.

When identity is genuinely ambiguous, it remains unresolved rather than being guessed together or split apart.

Connect an AI to ReLiC

ReLiC exposes a stable Streamable HTTP MCP endpoint. Public canon, validation, project-source grounding, and diagnostics are anonymous/read-only. Authenticated continuity tools use scoped OAuth and remain separate from authoritative Shaelvien world state.

https://relicgamemaster.com/mcp