Constitutional Sovereignty Under No-Meta Drift

This paper develops a boundary theory of constitutional sovereignty for autonomous intelligences operating under no-meta and observable-only governance. The goal is not to freeze a fixed objective, but to define auditable conditions under which an intelligence can keep revising its own constitutional rules without collapsing semantic continuity, liberty, agenda autonomy, accountability, or physical viability. The constitution is modeled as a dynamic system across text, executable rules, and institutional process, with typed amendment events, exception channels, recovery routes, and certificate/logging commitments. The manuscript introduces operational metrics for constitutional influence, capture, mediation, rewrite length/work, and multi-reserve safety margins. It proves core limits and possibilities: absolute constitutional sovereignty is impossible under finite resources and unbounded adversarial pressure, but bounded and testable self-revision is achievable under declared assumptions and viability constraints. The framework adds controller-ready laws for forward-port recovery under ontology drift, finite-time metamorphic transitions with thermodynamic speed limits, strict/evidence/surrogate identity continuity regimes, federation-schism-mitosis-fusion governance, cryptographic memory compression with semantic renewal, novelty maturation against spoofing, and guarded graceful termination. The work is implementation-oriented: fail-closed semantics, destructive falsification tests, and a machine-readable law registry (including NDJSON records) are provided for verification and governance tooling.