Objective Persistence — The Existence-Condition Layer

Objective persistence · amendment authority · assurance independence

Core question: Under what conditions does a positive objective-persistence verdict lose independent evidential force because the same amendment authority can alter both the objective and the basis used to validate persistence?

What the public research supports

Version 3 states a narrower conditional result. Self-containedness alone does not imply universal mutability or impossibility of formal verification. If an amendment authority can admit a masking amendment that changes the objective while co-amending persistence-relevant semantics or the validation basis so that the assurance chain still returns PASS, that verdict is not amendment-independent evidence of persistence. Defensible assurance requires either a verified invariant that excludes such masking amendments or at least one persistence-relevant validation basis outside the amendment authority being assessed. That basis need not be physically external.

Why this is a different problem from control

PCR-C concerns whether an intervention path remains available. Objective Persistence concerns the basis on which a system's continuing objective is assessed. A system can remain controllable while a persistence claim is poorly grounded; conversely, a persistence argument does not itself provide an infrastructure cutoff.

Assurance independence / protected basis

The required structural separation is from the amendment authority under evaluation, not necessarily from the physical system boundary. An internal immutable kernel, protected verifier, verified transition invariant, or independently controlled external mechanism can potentially serve this role if it cannot be silently co-amended to preserve acceptance of a persistence violation. The paper does not claim that any particular anchor is sufficient for long-run objective persistence.

Research questions opened by this line

What counts as genuinely external evidence for a persistence claim?
How should common dependencies between the evaluator and the evaluated system be represented?
When does an apparently external check remain functionally internal because it shares the same model, infrastructure, or assumptions?
How should objective-persistence claims be re-evaluated after self-modification or material architectural change?

These are research-agenda questions derived from the public research line; they are not claims already resolved by the paper.

Relation to the rest of the stack