02 — Lineage Repair Audit
Purpose
This document states what CNS 8.0 restores, what it keeps from the grounding/access work, and what it rejects.
CNS 8.0 core flow
CNS 8.0 uses this flow:
SNOs
→ chiral opposition
→ evidential entanglement
→ Antagonist pressure
→ critic ensemble
→ tensor proof closure
→ residual contradiction analysis
→ predicate invention
→ Synthesizer
→ orthesis candidate
→ audit / uncertainty report
Other subsystems are infrastructure or are omitted.
Restored concepts
Structured Narrative Objects
SNOs are the unit of analysis. Evidence atoms are attached inside SNOs, alongside identity, structure, provenance, and synthesis lineage.
Dialectical agents
The Proposer, Antagonist, Synthesizer, and critic ensemble are explicit roles with incompatible objectives. This prevents the system from collapsing into single-pass summarization or truth scoring.
Evidential Entanglement
CNS selects conflicts where accounts disagree over shared evidence. This is the target case for synthesis. Low-overlap disagreement is often just topic mismatch.
Chirality
Chirality is structured asymmetry. In CNS 8.0 it has three estimators:
- graph opposition over SNO reasoning graphs;
- evidence-weighted support/refute asymmetry;
- language–logic round-trip distortion:
||G(S(T)) - T||.
Orthesis
Orthesis is the stable synthesis candidate that survives grounding, rendering, and re-grounding. It is not a truth oracle. It is a fixed-point criterion for stability under the CNS loop.
Predicate invention
Persistent contradiction should trigger hidden-context discovery, not only possible-world enumeration. CNS 8.0 treats residual contradiction as a signal that the predicate vocabulary may be incomplete.
Topology
Graph cycles, Betti-1, persistence, holonomy, and curvature are diagnostics of synthesis difficulty. They are not decoration and not the whole theory.
Useful material retained from the later grounding/access work
The later grounding/access material supports CNS in these roles:
| Material | CNS 8.0 role |
|---|---|
| Evidence atoms | Span-level grounding inside SNOs |
| Record-access states | Missingness and source-availability metadata |
| Possible-world rankings | Auxiliary uncertainty layer after synthesis |
| Oracle boundary | Training/runtime separation |
| Strict proof vs likely truth | Output classification |
| Calibration | Evaluation and reporting |
| Audit reports | Final interface, not the engine |
| Prior-art boundary | Publication boundary |
Rejected failure pattern
CNS 8.0 rejects this structure:
evidence atom → record state → possible world → posterior ranking → audit report
That is a verification/ranking machine. It can support CNS but does not replace CNS.
Correct hierarchy
CNS 8.0
├── Structured Narrative Objects
├── dialectical agent loop
├── chirality / entanglement selection
├── tensor proof grounding
├── predicate invention
├── orthesis synthesis
└── access / possible-world / audit substrate
Style rule for future docs
Avoid prose that sounds like a naming correction or political repair. Write the architecture directly:
CNS 8.0 uses an access-aware grounding substrate to constrain what synthesized SNOs may claim. The synthesis step is performed by the dialectical SNO loop, not by the access substrate.