Differentiable Phase-QNode Claim Ledger

Claim	Status	Artefact IDs	Benchmark IDs	Known gaps
framework_overlay_parity	bounded_candidate	diff-qnode-framework-overlay-profile-v1	diff-qnode-framework-overlay-profile-v1	Local machines must still run the manifest install command before parity can execute. The overlay is CPU-only and intentionally excludes the repository jax[cuda12] extra.
ci_benchmark_evidence	bounded_candidate	diff-qnode-ci-evidence-schema-v1	diff-qnode-ci-evidence-schema-v1	GitHub-hosted runners are downgraded to functional_non_isolated. Production isolated_affinity promotion requires a self-hosted isolated-benchmark runner artefact. Accelerator benchmark claims require explicit CUDA/ROCm device metadata; missing visible devices are classified as silent_accelerator_fallback.
external_framework_comparison	bounded_candidate	diff-qnode-external-comparison-schema-v1	diff-qnode-external-comparison-schema-v1	Missing framework dependencies are hard_gap rows. Unconfigured LLVM/Enzyme or Catalyst tooling remains dependency_missing; configured runners must pass strict JSON, timeout, toolchain, and correctness gates.
phase_qnode_claim_boundary	bounded_candidate	diff-qnode-claim-ledger-v1	diff-qnode-claim-ledger-v1	The lane remains bounded_candidate until isolated CI benchmark and external comparison artefacts pass. Provider and QPU execution remain explicitly outside this claim. PyTorch dynamic-shape compile promotion, fullgraph compiled-frame promotion, and AOTAutograd/export persistence remain blocked until artefacts exist.
support_surface_alignment	bounded_candidate	diff-support-surface-alignment-audit-v1	diff-support-surface-alignment-audit-v1	The audit checks committed generated-manifest inventory and does not replace isolated benchmark artefacts. It validates support-surface wiring, not provider hardware execution or production performance.
hardening_slice_gate	bounded_candidate	diff-hardening-slice-gate-v1	diff-hardening-slice-gate-v1	The gate builds an auditable verification checklist but does not execute shell commands. It checks benchmark-classification invariants and does not upgrade functional_non_isolated rows to isolated_affinity evidence.
module_hardening_audit	bounded_candidate	diff-module-hardening-audit-v1	diff-module-hardening-audit-v1	The audit verifies inventory, tests, and declared diagnostics; it does not prove full formal correctness. Provider execution, hardware execution, and isolated benchmark promotion remain separate evidence gates.
external_validation_environment_lock	bounded_candidate	diff-external-validation-environment-lock-20260616	diff-external-validation-environment-lock-20260616	The lock manifest records reproducibility inputs only and does not execute external validation jobs. The artefact remains functional_non_isolated and cannot promote benchmark or hardware claims.
public_claim_table	bounded_candidate	diff-public-claim-table-20260616	diff-public-claim-table-20260616	The table constrains wording but does not execute external validation or isolated benchmarks. Rows remain bounded candidates unless the claim ledger promotes them with passing evidence.
external_validation_artifact_bundle	bounded_candidate	diff-external-validation-artifact-bundle-20260616	diff-external-validation-artifact-bundle-20260616	The manifest records committed evidence checksums but does not bundle private or ignored artefacts. The artefact remains functional_non_isolated and cannot promote benchmark or hardware claims.
external_validation_technical_report	bounded_candidate	diff-external-validation-technical-report-20260616	diff-external-validation-technical-report-20260616	The report is descriptive and does not execute external validation jobs or isolated benchmarks. Rows remain bounded candidates unless the claim ledger promotes them with passing evidence.
differentiable_baseline_scorecard	bounded_candidate	diff-baseline-scorecard-20260620	diff-baseline-scorecard-20260620	The scorecard is governance evidence and does not execute external validation jobs or isolated benchmarks. Rows remain behind-baseline until promoted claim-ledger rows, category-specific external comparisons, and isolated benchmark artefacts exist.
differentiable_rust_python_inventory	bounded_candidate	diff-rust-python-inventory-20260620	diff-rust-python-inventory-20260620	The inventory is governance evidence and does not execute Rust migrations, provider jobs, hardware jobs, LLVM/JIT lowerings, or isolated benchmarks. Rows remain non-promotional until Rust parity, polyglot parity, claim-ledger promotion, provider/hardware evidence, and isolated benchmark artefacts exist where required.
differentiable_architecture_rustification_map	bounded_candidate	diff-architecture-rustification-map-20260627	diff-architecture-rustification-map-20260627	The map is governance evidence and does not execute Rust migrations, provider jobs, hardware jobs, LLVM/JIT lowerings, or isolated benchmarks. Layer readiness remains blocked until Rust parity, polyglot parity, provider/hardware evidence, Enzyme/MLIR evidence, promoted claim-ledger rows, and isolated benchmark artefacts exist where required.
differentiable_dependency_environment_map	bounded_candidate	diff-dependency-environment-map-20260627	diff-dependency-environment-map-20260627	The map is lockfile provenance and does not execute framework, Enzyme, provider, hardware, accelerator, or isolated benchmark jobs. The Enzyme runner remains a hard-gap dependency profile until configured native Enzyme/LLVM/MLIR runner artefacts pass.
differentiable_isolated_benchmark_plan	bounded_candidate	diff-isolated-benchmark-plan-20260627	diff-isolated-benchmark-plan-20260627	The plan records reserved-host rerun commands and readiness blockers; it does not execute benchmark jobs. Current committed source artefacts remain functional_non_isolated or hard_gap until a self-hosted isolated-benchmark runner emits validated isolated_affinity artefacts. The current local host is not used for promotion because observed load and affinity do not satisfy the reserved-host policy.

Bounded language: the differentiable lane remains a promotion candidate unless isolated CI benchmark evidence and external comparison artefacts pass.