Jnana II: Constitutional Transformation Substrate

Software assurance usually evaluates programs, artifacts, or executions more directly than the authority and law governing the transformations that produced them. Compilers, build pipelines, configuration, CI, package hooks, deployment systems, and AI-generated rewrites may introduce dependencies, capabilities, effects, and policy changes whose authority is partly ambient or retrospectively narrated. Jnana II proposes a constitutional transformation substrate in which transformation regimes, admission rules, capability grants, effect boundaries, and receipts become explicit governed artifacts. Its strongest thesis is not that program correctness can be replaced by constitutional review. It is that program correctness and transformation admissibility answer different questions: whether a result satisfies a technical property, and whether the transformation was permitted to bind under declared law, authority, capability, and effect limits. The architecture organizes MCIR, MCIR-Law, Constitutional Invariants, PEIR, Domain Effects, and ReceiptLog into a proposal-admission-lowering-effect-witness path. The revision sharply bounds the claims. Canonical representation gives an artifact standing within a declared Jnana profile, not existence in general. A decidable law layer does not make arbitrary execution predictable. Effect membranes govern only mediated paths. Actual effects may be partial or irreversible. Receipts prove recorded relations, not truth or legitimacy. Replay requires explicit closure and may be exact, deterministic, equivalence-class, or observational. The paper adds an authority envelope, negative-power obligations, coverage testing, failure atomicity, recovery, an incremental adoption ladder, GMB/GSM integration, and a full evaluation program. It remains a specification candidate: no implementation, universal containment, deterministic replay, hardware adaptation, or transformation-correctness result is established.