In a web interpreted by AI systems, visibility no longer guarantees existence. What “exists” in a response depends on interpretability, the authority being activated, governed limits, and the ability to maintain a canon over time.
This page is a pivot: it links observable phenomena, rules of authority, mechanisms of proof, and operating environments (open web, RAG, agentic systems), then shows how debt accumulates and why versioning becomes a form of power.
The 6 observable phenomena
- Interpretive invisibilization: the information exists, but is not mobilized.
- Interpretive collision: entity fusion and synthesis hallucinations.
- Interpretive capture: saturation of the surrounding signal field and diversion of truth.
- Interpretive inertia: corrections do not “stick.”
- State drift: an outdated state becomes frozen (price, inventory, policy).
- Interpretive smoothing: thought is standardized, boundaries disappear.
Proof, audit, measurement
- Proof of fidelity: why a citation is no longer enough.
- Interpretation trace: making a response auditable without exposing the black box.
- Canon-output gap: measuring distortion rather than debating the “true.”
- Interpretive observability: the minimum metrics to log.
Operating environments: open web, RAG, agentic systems
- Open web vs closed environments: different surfaces of action, different forms of proof.
- Reliable RAG: reliability is a problem of boundaries, not only of retrieval.
- Agentic systems: non-response becomes a safety rule.
Debt, sustainability, versioning
- Interpretive debt: accumulation without spectacular failure.
- Interpretive sustainability: correction budgets and version discipline.
- Version power: versioning correction like software.
Canonical register
Core terms are consolidated in /definitions/. To connect articles systematically to conceptual objects, the entity graph is published in entity-graph.jsonld.
Complete list of the series
See the assembly page: Complete series: interpretive governance.
