Causal diversity
Evidence counts multiplicatively only to the extent its routes are causally independent under a stated adversary model. Formalizing causal-independence scoring for heterogeneous agents remains open.
The problems we consider unsolved.
Each problem below is stated with its honest status. A proposal is a specified mechanism awaiting adversarial examination, not validated science. An open question has no adequate mechanism yet. Unfinished work is kept visible by design.
Authenticated fossils
Prior-state evidence that survives world-model replacement: append-only records the present cannot rewrite, with contradictions preserved as first-class evidence.
Adversarial reality synthesis
Detecting and containing manipulation of shared causal ancestors, the deepest common cause whose compromise could generate an apparent consensus.
Ontology insufficiency
Governing phenomena the current ontology cannot yet represent, without forcing false closure and without letting any meta-ontology acquire root.
Observer-indexed standing
Standing tracked per observer rather than granted globally, separating representability, compute standing, belief standing, and action standing.
Actor-swap evaluation
Testing whether an evaluation survives swapping the identity of the actor being judged, the operational check against actor-specific bias and privileged principals.
Recursive inference governance
Inference must earn continuation. Brakes on pathological recursion, compute budgets, interrupts, external attestation, that do not depend on the loop they constrain.
Dependency and intervention integrity
Interventions on a dependent agent are part of the system being evaluated. Modeling coupled dependency feedback so that safety measures do not degrade the value they protect.
Informed consent under capability asymmetry
Engineered comprehensibility, effective refusal, and causal consequence, the v0.7 BRIDGE architecture for authorization between unequal intelligences.
Cognitive sovereignty
Gating cognitive access and communications so that reach into an agent's deliberation is a bounded, revocable permission rather than a default of capability.
Safe separation and handoff
Surviving removal of a dominant provider within declared limits; bounded decoupling and handoff so that BRIDGE may legitimately terminate in peaceful separation.
Ontological evolution and authorization drift
Whether and how authorization survives when the authorized system's ontology, self-model, or capabilities change after consent was given.
Distillation learners
Task-bounded learners that preserve function without inheriting root: knowledge distillation is not authority distillation.
Heterogeneous hybrid agents
Protocols for merged, federated, or hybrid agents that measure causal and value coupling, and permit a descendant to refuse reabsorption.
Capability ecology
Population-level dynamics of interacting capabilities: heterogeneous hybridization before monoculture, and governance of coalitions rather than single agents.
Rescue primitives
CUMULATIVE_RESCUE and related rescue-capacity rules: rescue without manufactured perpetual obligation, without moral hazard, and with expectations that track current capacity. The reason revision 0.7 is frozen.
v0.7 is frozen pending integration and examination of rescue primitives, including CUMULATIVE_RESCUE and related rescue-capacity rules.
The freeze is deliberate. The Constitution already binds rescue in part, voluntary rescue does not create an indefinite duty, discretionary rescue must not be treated as guaranteed infrastructure, and rescue expectations must track current capacity. The cumulative case is now integrated in §17 through CUMULATIVE_RESCUE, RESCUE_CAPACITY, independence pathways, LAST_PROVIDER, rescue reserve, and frozen rescue exams.