Roadmap

Canonical work queue

This file is the single source of truth for active roadmap and TODO selection. Older planning files under private local workspace/ and private internal records/ are retained as historical context only unless an item is copied here.

Current platform programme: reproducible evidence and benchmark infrastructure

• BL-54 DLA and topology-constrained differentiable control. Implemented 2026-07-29: exact finite parity-sector JVP/VJP and leakage gradients, a strict-decrease projected synthetic task, fixed-active-set topology-ledger JVP/VJP rules, explicit non-smooth/discrete blockers, composition evidence for the existing projected optimiser, notebook 51, and digest-bound JSON/Markdown evidence. Full-DLA classification, controllability, persistent-homology derivatives, active budget rescaling, positive connectivity thresholds, error correction, hardware, provider, QPU, deployment, and application claims remain out of scope. S54.5 QGNN wiring is descoped because no typed consumer maps Hilbert-space parity to graph topology.

• BL-60 finite synthetic chimera and multiscale control. Implemented 2026-07-29: exact two-population Kuramoto–Sakaguchi regimes, complete nested hierarchy contracts, multiscale Shanahan observables, composed analytic cluster-order targets, an unapplied backtracking phase proposal, topology- ledger projection reports, notebook 50, and digest-bound deterministic evidence. Thermodynamic-limit, biological/EEG, stability, controllability, provider, QPU, hardware, deployment, and market-efficacy claims remain out of scope. Optional BL-32 F5–F6 catalogue rows are explicitly descoped because the direct public facade has no registry consumer.

• Platform opportunity prioritisation. Recorded 2026-05-18 in <private-internal-record>: first result packs, then benchmark suite, then symmetry/sector-aware mitigation compiler, then Kuramoto/XY DSL, then differentiable control co-design.

• Hardware result-pack manifest and offline verifier. Implemented 2026-05-18: data/hardware_result_packs/manifest.json binds promoted IBM raw-count datasets, summaries, job IDs, SHA-256 digests, byte sizes, reproduction commands, claim scopes, and non-claims; scripts/verify_hardware_result_packs.py and the installable scpn-verify-hardware-packs entry point verify the manifest offline with explicit source-root handling and regression coverage.

• Hardware result-pack release export. Implemented 2026-05-18: scpn-verify-hardware-packs --export-dir ... writes deterministic per-pack tar.gz archives after verification, supports --pack-id filtering, and reports archive byte sizes plus SHA-256 digests for release notes.

• Hardware result-pack release checklist. Implemented 2026-05-18: docs/hardware_result_pack_release_checklist.md defines the evidence packet for verifier output, export digests, and reproduction logs before any tag or paper-facing update cites promoted hardware evidence.

• Hardware result-pack evidence-packet generator. Implemented 2026-05-18: scpn-generate-hardware-pack-evidence verifies selected packs, writes deterministic exports, runs reproduction commands, records logs, computes log digests, and writes the release-audit evidence packet.

• Standardised synchronisation benchmark suite registry. Implemented 2026-05-18: scpn-bench sync-benchmark-registry exports canonical Kuramoto/XY benchmark instances and a stable result schema for classical, exact, simulator, tensor-network, GPU, and hardware-replay rows.

• Standardised synchronisation benchmark runner. Implemented 2026-05-18: scpn-bench sync-benchmark-run exports schema-compatible no-QPU reference rows for kuramoto_ring_n4_linear_omega using a SciPy classical ODE row and dense exact XY row.

• Synchronisation benchmark tolerance gate. Implemented 2026-05-18: scpn-bench sync-benchmark-compare validates schema stability, no-QPU status, observable presence, unexpected rows, and per-observable tolerances for regenerated synchronisation benchmark artefacts.

• Second synchronisation benchmark instance runner. Implemented 2026-05-18: scpn-bench sync-benchmark-run --benchmark-id kuramoto_chain_n8_decay_omega exports no-QPU SciPy ODE and dense exact XY reference rows for the n=8 decaying-chain instance.

• Synchronisation benchmark comparator multi-instance gate. Implemented 2026-05-18: scpn-bench sync-benchmark-compare now checks every committed synchronisation benchmark artefact by default, while retaining focused --expected/--actual comparisons for debugging.

• Synchronisation benchmark regeneration gate. Implemented 2026-05-18: scpn-bench sync-benchmark-gate regenerates the registry, regenerates all committed synchronisation reference rows, and runs the multi-instance comparator as one no-QPU release gate.

• Synchronisation benchmark release integration. Implemented 2026-05-18: docs/release_readiness.md requires scpn-bench sync-benchmark-gate for benchmark-touching releases, and tools/audit_release_readiness.py treats the synchronisation benchmark registry, reference-row artefacts, and public benchmark pages as required release artefacts.

• Symmetry- and sector-aware mitigation compiler scoping. Implemented 2026-05-18: SymmetrySectorProblem, SymmetrySectorPlan, and plan_symmetry_sector_mitigation() define a fail-closed planning contract for parity postselection, symmetry expansion, and GUESS eligibility before execution-path integration.

• Symmetry-sector mitigation fixture gate. Implemented 2026-05-18: scpn-bench symmetry-sector-mitigation-gate regenerates and compares committed planner fixtures for eligible, missing-counts, missing-GUESS, and nonsymmetric-coupling cases before any execution-path integration.

• Symmetry-sector raw-count replay adapter. Implemented 2026-05-18: replay_symmetry_sector_counts() applies eligible parity postselection and symmetry expansion to offline raw counts with shot-accounting diagnostics and explicit GUESS deferral.

• Symmetry-sector replay fixture gate. Implemented 2026-05-18: scpn-bench symmetry-sector-mitigation-gate now locks raw-count replay fixtures for applied postselection/expansion and blocked missing-counts behaviour alongside planner fixtures.

• Native differentiable-programming foundation. Implemented 2026-05-20: scpn_quantum_control.differentiable exposes backend-neutral scalar-objective parameter-shift gradients, explicit parameter metadata, trainable masks, provenance-bearing GradientResult payloads, a native gradient-descent step, optional JAX value/gradient bridging, QSNN trainer integration, and a PennyLane VQE value/gradient adapter.

• MLIR, real-time runtime, and cloud-native deployment foundation. Implemented 2026-05-20: compiler.mlir emits deterministic Kuramoto-XY MLIR-style textual IR with digests and resource counts; control.realtime_runtime adds fixed-period deadline, jitter, and miss-budget accounting for software control loops; and deployment.cloud_native emits secret-free Kubernetes and Docker Compose manifests with non-root, read-only, no-privilege-escalation defaults.

• Provider-neutral hardware abstraction layer. Implemented 2026-05-20: hardware.hal adds immutable backend profiles, workload and job/result payloads, a runtime QuantumBackend protocol, metadata-only discovery for IBM Quantum, IonQ, AWS Braket QPU/simulator routes, Azure Quantum QPU/emulator/preview routes, Quantinuum, Rigetti QCS, QuEra/Bloqade, IQM, Pasqal, OQC, D-Wave, qBraid IonQ, Quandela photonics, and local simulators, plus fail-closed cloud submission and explicit approval tokens before any injected adapter can submit live QPU or managed-simulator work.

• Qiskit HAL adapter integration. Implemented 2026-05-20: hardware.hal_qiskit adds base64-QPY and OpenQASM 3 workload conversion, a local QiskitAerHALAdapter, and an approval-gated QiskitRuntimeHALAdapter for IBM Runtime Sampler execution through the provider-neutral HAL contract.

• IBM Runtime no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_qiskit_runtime_backend adds the direct direct/ibm_quantum capability path for injected Qiskit Runtime backends, including route-bound IR support, basis gates, queue depth, shot/circuit limits, online state, dynamic-circuit feature inference, and calibration timestamp without job submission.

• Direct D-Wave no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_dwave_solver adds the direct direct/dwave route and metadata snapshot path for injected D-Wave solver objects or metadata JSON, including BQM/Ising/QUBO/MLIR declaration translation, qubit count, annealing topology, read limits, queue/load estimate, online state, simulator flag, category, and last update timestamp without sampler calls.

• Direct IQM no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_iqm_backend adds the direct direct/iqm route and metadata snapshot path for injected IQM backend objects or metadata JSON, including QPY/Qiskit/OpenQASM declaration translation, qubit count, native gate set, shot and circuit limits, queue depth, online state, simulator flag, architecture name, and calibration timestamp without circuit execution.

• Direct IonQ no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_ionq_backend adds the direct direct/ionq route and metadata snapshot path for injected IonQ backend objects or REST JSON, including IonQ JSON/OpenQASM/QIR declaration translation, qubit count, basis gates, shot limits, queue depth, online state, simulator flag, gateset, and calibration timestamp without job creation.

• Direct OQC no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_oqc_target adds the direct direct/oqc route and metadata snapshot path for injected OQC target objects or metadata JSON, including OpenQASM/QIR declaration translation, qubit count, native gate set, shot and circuit limits, queue depth, online state, simulator flag, topology, and calibration timestamp without QCAAS submission.

• Direct Pasqal no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_pasqal_target adds the direct direct/pasqal route and metadata snapshot path for injected Pasqal target objects or metadata JSON, including Pulser/Pasqal IR/OpenQASM/MLIR declaration translation, atom count, supported bases, channel declarations, shot and sequence limits, queue depth, online state, simulator flag, lattice geometry, and calibration timestamp without Pasqal submission.

• Direct Quandela no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_quandela_processor adds the direct direct/quandela route and metadata snapshot path for injected Quandela processor objects or metadata JSON, including Perceval/OpenQASM/MLIR declaration translation, mode count, optical component set, photonic feature flags, shot and circuit limits, queue depth, online state, simulator flag, topology, and calibration timestamp without processor or sampler calls.

• Direct Quantinuum no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_quantinuum_backend adds the direct direct/quantinuum route and metadata snapshot path for injected Quantinuum backend objects or metadata JSON, including tket/OpenQASM/QIR declaration translation, qubit count, native gate set, mid-circuit measurement feature inference, shot/batch limits, queue depth, online state, simulator flag, machine id, and calibration timestamp without circuit processing.

• Direct QuEra/Bloqade no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_quera_bloqade adds the direct direct/quera route and metadata snapshot path for injected Bloqade routine objects or metadata JSON, including Bloqade/Braket AHS/MLIR declaration translation, atom count, native analogue operations, shot and circuit limits, queue depth, online state, simulator flag, lattice geometry, and calibration timestamp without routine execution.

• Direct Rigetti QCS no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_rigetti_qcs adds the direct direct/rigetti route and metadata snapshot path for injected Rigetti QCS quantum-computer objects or metadata JSON, including Quil/OpenQASM declaration translation, qubit count, native gate set, shot limits, queue depth, online state, simulator flag, compiler versions, and calibration timestamp without program compilation or execution.

• AWS Braket HAL adapter integration. Implemented 2026-05-20: hardware.hal_braket adds OpenQASM 3 workload conversion for Braket circuits, local SV/DM simulator execution, and approval-gated AWS Braket device/task submission through the provider-neutral HAL contract. Optional cloud broker extras now expose braket, azure, and qbraid install groups.

• AWS Braket no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_braket_device adds route-bound metadata snapshots for injected Braket QPU, AHS, and managed-simulator devices, including action-derived IR support, qubit count, basis operations, shot limits, queue depth, online state, simulator flag, and calibration/update timestamp without task submission.

• qBraid catalogue no-submit IR normalisation. Implemented 2026-05-20: snapshot_from_qbraid_device now consumes qBraid catalogue program_specs, normalises broker program declarations into HAL IR tokens, adds a broker_catalog_target feature marker, and records the resolved broker route without device submission.

• Strangeworks catalogue no-submit IR normalisation. Implemented 2026-05-20: snapshot_from_strangeworks_backend now consumes Strangeworks catalogue program declarations such as available_programs, normalises them into HAL IR tokens, adds a broker_catalog_target feature marker, records the resolved broker route, and treats backend state/availability as no-submit readiness evidence.

• Azure Quantum HAL adapter integration. Implemented 2026-05-20: hardware.hal_azure adds OpenQASM 3 workload construction and an injected Azure target adapter for IonQ, Quantinuum, Rigetti, Pasqal, and preview routes. The adapter uses Azure Target.submit(...), requires explicit approval through HAL, and fails closed when no target or target factory is supplied.

• Azure Quantum no-submit capability snapshot. Implemented 2026-05-20: snapshot_from_azure_target adds route-bound metadata snapshots for injected Azure Quantum targets, including declared IR format translation, qubit count, gate basis, shot/circuit limits, queue depth, availability state, simulator flag, provider id, and calibration timestamp without calling target submission APIs.

• Provider frontier watchlist and certification gate. Maintain SCPN's declared HAL/provider matrix as provider SDKs and broker catalogues evolve. A provider may be publicly claimed only when it has descriptor coverage, route resolution, optional dependency smoke coverage, no-submit capability snapshot coverage, docs, tests, approval-gated live submission semantics where applicable, and a passing provider-conformance gate. New direct adapters should be added only for providers not adequately covered by the existing direct, qBraid, Strangeworks, Braket, or Azure routes.

• Stable-core release/repro gate. Use scpn-bench stable-core-release-gate before release notes, API changes, or public stable-core documentation changes. The bundle runs stable-core contract, stable-core capability, and preflight checks in one no-QPU command.

• Stable-core contract gate component. Implemented: use scpn-bench stable-core-contract-gate to verify only contract fixtures (Problem, Backend, Experiment, Result, and Kuramoto adaptor mapping) when stable-core surface edits are contract-only.

• Stable-core capability gate component. Implemented: use scpn-bench stable-core-capability-gate to verify only capability artifacts when stable-core surface edits are capability-only.

• Stable-core preflight gate component. Implemented: use scpn-bench stable-core-preflight-gate to validate stable-core preflight fixtures before touching stable-core API text, docs, or release-facing claims.

• Symmetry- and sector-aware mitigation compiler. Implemented: plan_symmetry_sector_mitigation() provides a bounded fail-closed planner, replay_symmetry_sector_counts() provides the offline raw-count replay adapter, and scpn-bench symmetry-sector-mitigation-gate locks planner and replay fixtures before execution-path integration.

Current top priority: repository hygiene and release safety

• GitHub Actions history audit. Completed 2026-05-06: current failed and cancelled workflow-run queries are empty after resolved superseded runs were classified and removed.
• Latest CI/link-check failures. Completed 2026-05-06: latest main CI run 25412632454 passed lint, tests, security, Rust audit, optional integration, hardware smoke, and CI gate.
• Security alert audit. Completed 2026-05-06: open CodeQL, Dependabot security, and secret-scanning alert queries returned empty lists.
• PR and branch hygiene. Completed 2026-05-06 for active repository state: no open pull requests were present during the hygiene pass.
• Safe workflow-run cleanup. Completed 2026-05-06: deleted only resolved or superseded failed/cancelled workflow runs after their causes were fixed by later successful commits.
• Private audit record, and Arcane notification. Completed 2026-05-06 with append-only coordination records and a new SNN stimulus.

Next repository hygiene follow-up

• Full historical Actions audit automation. Implemented 2026-05-06: tools/audit_github_actions_history.py classifies gh run list --json ... history into clean successes, in-progress runs, unresolved failures, resolved failures, unresolved cancellations, superseded cancellations, and safe delete candidates without deleting any run.
• Actions audit GitHub workflow integration. Implemented 2026-05-06: .github/workflows/actions-history-audit.yml runs the classifier on a schedule or manually, uploads read-only audit artefacts, and performs no deletions.
• Link-check history automation. Implemented 2026-05-06: tools/audit_link_check_history.py records live link-check failures, resolved historical failures, accepted external-transient failures, and safe delete candidates from workflow history.
• Actions-history dashboard documentation. Implemented 2026-05-06: docs/actions_history_dashboard.md documents audit artefacts, bucket meanings, accepted external-transient Link Check failures, and the manual safe-delete rule.
• Post-push Actions history workflow observation. Completed 2026-05-06: manually triggered Actions History Audit run 25436780329; the classify-history job passed and uploaded the actions-history-audit artefact.

Archived repository hygiene checklist

• GitHub Actions history audit. Classify the full workflow-run history into resolved failures, unresolved failures, and cancelled superseded runs. Do not delete any run until the corresponding failure is demonstrably resolved by a later successful run or superseded by a closed branch/PR.
• Latest CI/link-check failures. Inspect the latest failing Link Check runs on main, fix any live issue, and record whether older link failures are resolved by the fix.
• Security alert audit. Check open CodeQL, Dependabot security, and secret-scanning alerts. Fix true positives; document API-permission limits or accepted false positives internally.
• PR and branch hygiene. Review open Dependabot PRs and stale branches, decide whether to merge, rebase, close, or leave blocked, then document the decision.
• Safe workflow-run cleanup. After classification, delete only cancelled or fully resolved failed runs where deletion does not remove the only evidence for an unresolved defect.
• Private audit record, and Arcane notification. Keep a new timestamped log and private handoff record for the repository-hygiene audit, and emit a factual Arcane stimulus.

Active release tasks

• Coverage and test-quality closure. Implemented 2026-05-18: tools/audit_release_readiness.py composes the release coverage XML gate and behavioural-quality gate into one tag-readiness contract. The gate requires a fresh coverage.xml, aggregate release coverage, no unjustified missing files, explicit reviewed exclusions for generated code, and aggregate behavioural density thresholds. Per-file gaps remain visible and can be promoted to hard blockers with --fail-on-file-gap. The 100 percent coverage target remains a future improvement, not a blocker for a bounded release.
• Scientific gap queue. Implemented 2026-05-18: release safety is closed by hard claim-boundary gates rather than by pretending the open science is solved. K_nm measured-system promotion, TCBO p_h1 promotion, and S2/S5 broad-advantage readiness remain blocked unless their executable gates pass. The former Paper 0 downstream register was intentionally extracted from this repository in 97f9e910, so it is no longer a package release blocker.
• Documentation-surface release gate. Implemented 2026-05-20: tools/audit_documentation_surface.py --allowlist tools/documentation_surface_allowlist.json --fail-on-findings now runs in the primary CI lint job after the repository-wide documentation surface was burned down to zero active findings. The allow-list remains tracked and reasoned; after the Paper 0 extraction it carries no current repository-specific waivers.
• Documentation-surface local preflight mirror. Implemented 2026-05-20: tools/preflight.py and the repository pre-push hook now run the same allow-listed documentation-surface gate as CI. Focused static tests lock both local guard paths so the release gate cannot silently disappear from developer preflight.
• XY Kuramoto trajectory time-grid hardening. Implemented 2026-05-12: QuantumKuramotoSolver.run() now builds explicit time boundaries and evolves a final partial interval when t_max is not an integer multiple of dt, preventing state/label drift in non-divisible horizons. API and performance docs record the exact endpoint contract.
• TCBO coupling-weighted complex reconstruction. Implemented 2026-05-12: tcbo_weighted_complex.py reconstructs the roadmap blocker using K_ij * |cos(theta_j - theta_i)| edge weights, a thresholded flag complex, beta-1 over GF(2), and a threshold scan against the 0.72 target. The audit runner records this reconstruction separately from the legacy delay-embedded observer path; the claim remains unpromoted pending preregistered replay with uncertainty.
• TCBO p_h1 replay-uncertainty gate. Implemented 2026-05-18: tcbo_weighted_uncertainty_replay() now turns the remaining TCBO promotion gate into an executable no-QPU replay contract. It samples phase draws over a preregistered coupling matrix, reports confidence intervals and threshold/error distributions, and refuses promotion unless a named preregistered dataset crosses the 0.72 target within the declared tolerance. scripts/run_tcbo_reproduction_audit.py exposes the replay count, confidence level, promotion tolerance, and dataset-id gate.
• K_nm measured-system promotion hard gate. Implemented 2026-05-18: scripts/run_knm_physical_validation_audit.py now emits a measured_system_promotion_readiness decision that blocks physical validation closure unless the candidate has locked physical-unit normalisation, per-edge uncertainty, a full pairwise matrix, per-edge agreement within uncertainty, magnitude and critical-response errors within tolerance, and node-label plus edge-value null-model wins. IBM remains unnecessary unless the claim is explicitly redefined as a backend calibration/coupling-map claim.
• K_nm measured-system candidate expansion. Implemented 2026-05-18: the power-grid measured-coupling builder now emits an IEEE 14-bus voltage-weighted admittance candidate with branch reactance, voltage, all-pairs uncertainty accounting, and explicit non-promotional limitations. The physical validation gap remains open because public case14 does not provide measured per-bus inertia constants for all load buses and the hard promotion gate must still pass before any claim is strengthened.
• K_nm measured-system unit-class promotion guard. Implemented 2026-05-20: scripts/run_knm_physical_validation_audit.py now rejects association-observable units such as phase-locking value, coherence, correlation, mutual information, and transfer entropy at the strict measured-system promotion gate. Model-derived power-grid coupling units still pass the unit-class screen but remain blocked by the existing uncertainty, magnitude, spectral, and null-model gates.
• K_nm measured-candidate release gate. Implemented 2026-05-20: scpn-bench knm-measured-candidate-gate now checks the committed EEG PLV, IEEE 5-bus, and IEEE 14-bus K_nm comparison artefacts. The gate fails if any candidate is promoted, changes edge count, drops the strict unit-class decision, or records physical validation as closed. The gate remains the current in-repository measured-candidate promotion boundary after the Paper 0 replay package was extracted from this repository.
• QSVT resource-estimator input hardening. Implemented 2026-05-12: the QSVT resource estimator and query-count helpers now reject non-square, dimension-mismatched, asymmetric, or non-finite K_nm/omega inputs plus invalid simulation-time and error-budget parameters before any Hamiltonian construction or resource claim is produced. The phase API documents the validation contract.
• QSP seed-angle degree validation. Implemented 2026-05-12: qsp_phase_angles() now rejects boolean, fractional, string, and negative degrees with explicit ValueError messages before reaching the non-production initial-guess path or the production synthesis gate. The phase API documents that seed angles are offline optimiser inputs, not compiled QSP phases.
• Koopman Rust-kernel routing hardening. Implemented 2026-05-12: build_koopman_generator_rust() now uses the optional native scpn_quantum_engine.koopman_generator export when available, exposes require_rust=True for benchmark/release gates that must reject fallback execution, and keeps the NumPy generator as the explicit fallback path. The Rust crate and .pyi contract now carry the Koopman dense-generator export.
• Second-order Trotter nested-commutator bound hardening. Implemented 2026-05-12: second-order trotter_error_bound() and optimal_dt() now use an exact spectral-norm nested commutator for small systems and a rigorous Pauli coefficient-norm upper bound for larger systems, replacing the previous heuristic gamma²/max(K) estimate.
• R-witness entanglement-depth claim hardening. Implemented 2026-05-12: detect_entanglement_from_R() now reports only the certified lower-bound depth from the R separability witness: 1 when the separable bound is not violated and 2 when nonseparability is certified. Stronger multipartite-depth claims are no longer inferred from heuristic R thresholds.
• PennyLane VQE observable hardening. Implemented 2026-05-12: PennyLaneRunner.run_vqe() now measures the Kuramoto order parameter from the optimized hardware-efficient ansatz using the same local X/Y-expectation phase reconstruction as the Trotter path. VQE results no longer report an unmeasured order_parameter=0.0 placeholder.
• Biological MWPM syndrome-parity hardening. Implemented 2026-05-12: BiologicalMWPMDecoder.decode_z_errors() now rejects malformed syndromes and odd syndrome parity in any connected component when no explicit rough-boundary model is present. The decoder no longer drops an unmatched defect to force even cardinality.
• Biological surface-code K-matrix hardening. Implemented 2026-05-12: BiologicalSurfaceCode now validates that K is a finite square symmetric zero-diagonal coupling matrix and that threshold is finite and non-negative before constructing the graph code, preventing malformed matrices from reaching stabilizer or MWPM logic.
• MPS long-range truncation hardening. Implemented 2026-05-12: the quimb DMRG/TEBD paths now reject non-nearest-neighbour couplings by default because SpinHam1D/LocalHam1D only represent adjacent bonds. Callers that intentionally run the truncated tensor-network diagnostic must pass allow_long_range_truncation=True; returned metadata records coupling_scope and omitted_coupling_l1.
• Paper claim-boundary audit after hardening. Implemented 2026-05-12: the Phase 1 DLA short paper now states that the IBM circuit used the nearest-neighbour truncation of the exponential coupling matrix, and the benchmark API no longer turns crossover estimates into a broad hardware-only dynamics claim.
• Legacy preprint claim-boundary audit. Implemented 2026-05-12: docs/preprint.md now frames ibm_fez material as artifact-backed legacy hardware evidence, removes "first hardware demonstration" and backend-general outperformance phrasing, and keeps 16-qubit/ansatz observations descriptive unless a later ledger review promotes a specific raw-count claim.
• Coauthor meeting claim-readiness pass. Implemented 2026-05-12 for the 2026-05-13 15:00 Europe/Zurich Teams meeting: paper/submissions/submission_001_ibm_fez_synchronisation/main.tex, docs/preprint.md, and docs/PAPER_CLAIMS.md now avoid first-hardware, hardware-DTC, backend-general coherence, generic-outperformance, and clean-readout overclaims. The FIM manuscript boundary remains regression-tested as a backend/circuit- specific negative hardware result, and an internal meeting brief was added at <private-internal-record>.
• NumPy NQS sampling-contract hardening. Implemented 2026-05-12: vmc_ground_state() now rejects explicit n_samples because the current NumPy RBM path performs exact enumeration with central finite-difference gradients, not sampled VMC. Returned metadata labels sampling_mode, n_samples_used, and gradient_method.
• Trapped-ion proxy-basis hardening. Implemented 2026-05-12: transpile_for_trapped_ion() now requires allow_proxy_basis=True when multiqubit instructions need the CX proxy for native MS/RXX-style trapped-ion gates. Returned circuit metadata records the representative basis, all-to-all connectivity model, and non-calibrated hardware-claim boundary.
• GPU batch VQE backend-contract hardening. Implemented 2026-05-12: batch_vqe_scan() no longer ignores use_gpu=True; that request requires PyTorch plus CUDA or fails clearly. The default NumPy path now reports backend, product-Ry diagnostic ansatz, random-scan optimizer, and no-hardware-claim metadata while rejecting invalid sample and parameter counts before execution.
• Integrated-information wrapper production route. Implemented 2026-05-12: IntegratedInformationPhi now routes explicit coupling_matrix and natural_frequencies inputs to the compute_quantum_phi Kuramoto-XY density-matrix engine with shape, symmetry, and finite-value validation. Counts-only entropy remains an opt-in entropy_proxy diagnostic and is never returned as phi.
• QFI production-adapter input hardening. Implemented 2026-05-12: QuantumFisherInformation now rejects non-finite Hamiltonian inputs, non-integer measurement budgets, boolean shot counts, and malformed/out-of-range/diagonal coupling_pairs before routing to the spectral QFI engine. The analysis API documents the production contract and keeps counts-derived sync/DLA estimates labelled as opt-in proxy diagnostics.
• DLA-protected witness count normalisation hardening. Implemented 2026-05-12: evaluate_dla_protected_memory() now rejects fractional, boolean, and negative count values before shot total normalisation, preventing malformed hardware/simulator payloads from being coerced into invalid probability mass. The DLA-protected subspace docs record the count contract.
• Logical-sync fidelity proxy domain hardening. Implemented 2026-05-12: LogicalSyncWitness now rejects non-finite and out-of-range scalar fidelity diagnostics even on the explicit allow_fidelity_proxy=True path. The reproducibility table records that this remains a labelled finite unit-interval proxy, not a production logical synchronisation witness.
• RL witness-discovery wrapper contract hardening. Implemented 2026-05-12: RLDiscoveryAgent now rejects unwired compatibility parameters (runner, unsupported observables, unsupported reward_function, and non-positive n_episodes) at construction time. The wrapper contract is documented in the analysis and witness-discovery API pages.
• RL pulse-optimiser interface configuration hardening. Implemented 2026-05-12: RLPulseOptimizer now rejects missing runners, non-finite/out-of-range target_sync_order, and non-positive, fractional, or boolean episodes before the intentionally gated optimisation methods can be called. The language-tier policy records that this remains a fail-fast interface pending a real objective, benchmark, and replayable training trace.
• DLA tensor-network interface configuration hardening. Implemented 2026-05-12: dla_truncated_tn() now rejects non-square, non-finite, or asymmetric K_nm inputs, invalid bond dimensions, non-positive/non-finite DLA cutoffs, and unsupported observables before reaching the intentionally gated tensor-network implementation path. The language-tier policy records the validated fail-fast boundary.
• PEC local multi-qubit coefficient decomposition. Implemented 2026-05-12: pauli_twirl_decompose() now returns tensor-product quasi-probability coefficients for n_qubits >= 1, preserving exact single-qubit Rust parity while documenting that correlated multi-qubit noise still requires a separately characterised inverse channel.
• Behavioural-test audit automation. Implemented 2026-05-06: tools/audit_test_behaviour.py inventories test modules for assertion-bearing tests, exception contracts, parametrisation, and smoke-only tests so the manual behavioural audit can be prioritised.
• Behavioural-audit tool test coverage. Implemented 2026-05-06: tests/test_audit_test_behaviour.py now covers function and class-based test detection, assertion-helper contracts, pytest.raises contracts, parametrisation counts, smoke-only reporting, deterministic JSON/text output, and CLI failure status.
• Rust .pyi contract checker coverage. Implemented 2026-05-06: tests/test_check_rust_pyi_exports.py now covers PyO3 export parsing, namespace stripping, private .pyi helper exclusion, matching-contract success output, and missing/stale export failure reporting for the CI typing-contract gate.
• Version-consistency hook coverage. Implemented 2026-05-06: tests/test_check_version_consistency.py now covers matching version carriers, mismatched carrier reporting, missing carrier reporting, and unmatched-version-pattern reporting for the pre-commit release hook.
• Secret-scanner hook coverage. Implemented 2026-05-06: tests/test_check_secrets.py now covers vault-token extraction without reading the real vault, added-line diff scanning, keyword secret detection, placeholder and variable-reference suppression, and safe redaction output.
• Commit-trailer hook coverage. Implemented 2026-05-06: tests/test_check_commit_trailers.py now covers accepted trailer messages, missing-trailer rejection, banned-subject-word rejection, default body-word allowance, stricter body scanning, and help output.
• Benchmark CLI failure-policy coverage. Implemented 2026-05-06: tests/test_bench_cli.py now covers default stop-on-first harness failure, --keep-going execution after failures, and --no-diff skipping of artefact-drift checks.
• Local preflight tool coverage. Implemented 2026-05-06: tests/test_preflight_tool.py now covers gate pass/fail output, stdout/stderr tail reporting, --no-tests, --no-coverage, and stop-on-first-failure behaviour without running the full suite.
• Coverage-gap audit automation. Implemented 2026-05-06: tools/audit_coverage_gaps.py parses coverage.xml, inventories package source files missing from coverage reports or below a per-file threshold, and provides --fail-on-gap for release-gating without running tests or treating line coverage as scientific validation. Usage and boundaries are documented in <private-internal-record>.
• Coverage XML freshness guard. Implemented 2026-05-07: tools/audit_coverage_gaps.py now emits an explicit coverage_report_warning when no selected package source files match the supplied coverage XML, uses defusedxml for hardened XML parsing, and has focused behavioural coverage for the warning path.
• Coverage-gap CI observation. Implemented 2026-05-06: the Python 3.12 coverage job now emits coverage-gap-audit.json from tools/audit_coverage_gaps.py and uploads it as a 30-day artifact. It is intentionally observational rather than a hard per-file gate until the broader coverage-to-100% release task closes.
• Behavioural assertion-helper recognition. Implemented 2026-05-06: the behavioural audit now counts assertion helper calls such as assert_* functions instead of misclassifying them as smoke tests.
• Class-based test audit coverage. Implemented 2026-05-06: tools/audit_test_behaviour.py now counts test_ methods inside Test* classes, so class-based pytest modules are no longer misreported as empty.
• Behavioural-test audit. Completed 2026-05-06: <private-internal-record> records the final audit state after targeted hardening passes. The current automated inventory covers 319 test modules and reports no smoke-only tests; broader coverage-to-100% work remains open as a separate release task.
• Behavioural-test audit: topological coupling guard. Implemented 2026-05-06: tests/test_topological_coupling_guard.py now asserts coupling symmetrisation, diagonal clearing, omega/config retention, and no state mutation on the missing optional dependency path.
• Behavioural-test audit: readout matrix guards. Implemented 2026-05-06: tests/test_readout_matrix.py now asserts confusion matrix column stochasticity, condition number and shot accounting, bitstring parsing, invalid-count rejection, mitigation normalisation, and mean-magnetisation observables.
• Public API smoke-only audit closure. Implemented 2026-05-06: tests/test_public_api.py now asserts explicit export-count contracts in addition to delegated import/type checks, closing the remaining smoke-only module found by tools/audit_test_behaviour.py.
• Backend registry no-op contract. Implemented 2026-05-06: test_unregister_missing_is_silent now asserts registry state is unchanged when unregistering a missing backend.
• Coverage 0.95 import contracts. Implemented 2026-05-06: tests/test_coverage_095_push.py import checks now assert module identity and public surface instead of only importing modules.
• Phase-artifact fuzz boundary contracts. Implemented 2026-05-06: inclusive PLV and R boundary tests now assert retained endpoint values instead of only constructing artefacts.
• VQLS denominator guard contract. Implemented 2026-05-06: test_denominator_guard_path now asserts the patched statevector path is exercised while forcing the near-zero denominator branch.
• STDP pipeline update contract. Implemented 2026-05-06: pipeline wiring now asserts Hebbian pre/post firing increases the synaptic weight while preserving configured clamp bounds.
• Rust benchmark timing contracts. Implemented 2026-05-06: pure performance rows in tests/test_rust_path_benchmarks.py now assert finite non-negative timing samples instead of only printing benchmark lines.
• E2E boundary audit automation. Implemented 2026-05-06: tools/audit_e2e_contract_boundaries.py inventories hardware/QPU, bridge, SC-NeuroCore, Phase Orchestrator, notebook, and example workflow test boundaries without fabricating coverage for missing categories.
• Example workflow contract. Implemented 2026-05-06: tests/test_example_workflows.py statically verifies every example script parses, exposes main(), has an execution guard, and is listed in examples/README.md; the README now includes examples 19--21.
• Notebook workflow contract. Implemented 2026-05-06: tests/test_notebook_workflows.py statically verifies committed notebooks are valid nbformat-4 JSON artefacts with cells, metadata, recognised cell types, and notebook-compatible source fields.
• E2E and contract audit. Completed 2026-05-06: tools/audit_e2e_contract_boundaries.py --fail-on-missing now reports all six tracked boundaries covered; static notebook/example contracts are documented in docs/e2e_contract_boundaries.md without claiming executed scientific validation.
• Mutation-test expansion: XY Kuramoto invariants. Implemented 2026-05-06: tests/test_xy_kuramoto.py now includes mutation guards for zero-time identity evolution, internal Pauli qubit ordering, and trajectory time-grid endpoints.
• Mutation-test expansion: async runner state contracts. Implemented 2026-05-06: tests/test_async_runner.py now guards round-robin index progression, submitted timestamp capture, underlying job-handle preservation, and timeout propagation into result polling.
• Mutation-test expansion: backend registry state contracts. Implemented 2026-05-06: tests/test_backend_registry.py now guards discovery-state reset on clear(), sorted known-backend diagnostics for missing names, and one-shot discovery semantics after broken plugin load attempts.
• Mutation-test expansion: Knm Hamiltonian bridge invariants. Implemented 2026-05-06: tests/test_knm_hamiltonian.py now guards Kuramoto ring self-coupling/edge count, inclusive ansatz thresholding, and delta-zero equivalence across XY, XXZ, sparse, and dense Hamiltonian paths.
• Mutation-test expansion: DLA parity invariants. Implemented 2026-05-06: tests/test_dla_parity_theorem.py now guards parity operator popcount ordering, even/odd projector orthogonality and reconstruction, and unnormalised weight accounting.
• Mutation-test expansion: ZNE mitigation invariants. Implemented 2026-05-06: tests/test_zne.py now guards terminal measurement stripping/re-append behaviour, exact odd-scale fold gate counts, and immutable result copies for extrapolation inputs.
• Mutation-test expansion: PEC mitigation invariants. Implemented 2026-05-06: tests/test_pec.py now guards exact inverse-channel coefficient formulae, sign-distribution support, and circuit-size exponentiation of PEC overhead.
• Mutation-test expansion: readout-matrix mitigation invariants. Implemented 2026-05-06: tests/test_readout_matrix.py now guards full-basis matrix stochasticity, prepared-state shot accounting, invalid observed labels, and probability-observable contracts.
• Mutation-test expansion: symmetry-decay mitigation invariants. Implemented 2026-05-06: tests/test_symmetry_decay.py now guards copied decay-model sequences and real positive GUESS correction factors under sign-flipped symmetry readings.
• Mutation-test expansion: compound-mitigation pipeline guards. Implemented 2026-05-06: tests/test_compound_mitigation.py now asserts the CPDR plus symmetry pipeline calls the backend exactly once with the generated training circuits and preserves requested training-count boundaries.
• Mutation-test expansion: CPDR mitigation guards. Implemented 2026-05-06: tests/test_cpdr.py now guards Clifford snapping without target mutation, little-endian observable-qubit extraction, zero-slope fallback, and backend training-count boundaries.
• Mutation-test expansion: DD mitigation guards. Implemented 2026-05-06: tests/test_dd.py now guards negative idle qubit rejection, exact DD sequence ordering, idle-qubit placement, and source-circuit immutability.
• Mutation-test expansion: symmetry-verification guards. Implemented 2026-05-06: tests/test_symmetry_verification.py now guards invalid Z2 parity labels and invalid computational-basis count labels before post-selection or symmetry expansion.
• Mutation-test expansion: Mitiq integration guards. Implemented 2026-05-06: tests/test_mitiq_integration.py now guards that default ZNE and DDD executor paths forward the requested shot count into the Qiskit executor instead of silently using its default.
• Mutation-test expansion. Completed 2026-05-06 across the listed initial targets: phase/xy_kuramoto.py, hardware runner and backend registry state contracts, Hamiltonian bridge invariants, DLA-parity analysis contracts, and mitigation modules.
• Mock/stub audit automation. Implemented 2026-05-06: tools/audit_mock_stub_usage.py inventories Mock, MagicMock, patch, monkeypatch, and fake/stub helper usage, flags apparent third-party boundaries, and highlights result-term contexts for manual review without deleting or rewriting tests.
• Mock/stub audit. Completed 2026-05-06: an internal local audit record captures the tools/audit_mock_stub_usage.py inventory, manual classification of the 22 result-term contexts, and the decision that current mocks model branch control, optional dependencies, configuration, or executor boundaries rather than fabricated publication data or successful hardware execution. Ensure future mocks only model third-party boundaries and never fabricate scientific results, provenance, datasets, or successful hardware execution. Initial target list reviewed: phase/xy_kuramoto.py, hardware/async_runner.py, hardware/backends.py, bridge/*, analysis/dla*, and mitigation/*.
• Full-suite ordering audit automation. Implemented 2026-05-06: tools/audit_test_ordering_state.py inventories module reloads, monkeypatch state changes, environment mutation, module injection, random seed mutation, and backend/cache/registry global mutations so ordering risks can be reviewed without relying on ad-hoc greps.
• Full-suite ordering audit. Completed 2026-05-06: a local internal audit record documents the tools/audit_test_ordering_state.py result (420 findings), accepted isolation patterns for monkeypatch-managed state, optional dependency module injections, and paired reload tests, plus the high-density files to prioritise if ordering randomisation later exposes flakes.
• Static-analysis pass. Completed 2026-05-06: Semgrep was added as a development dependency and run with explicit Python/security rules (0 findings); Bandit was run over src, scripts, and tools, easy true positives were fixed, and the remaining 57 low findings were documented locally as accepted subprocess/provenance tooling or deterministic non-cryptographic shuffling.
• QuantumKuramotoSolver validation. Completed 2026-05-06: validates oscillator count, coupling-matrix shape/squareness, symmetric XY semantics, finite coupling and omega values, matching finite omega length, and canonicalises the solver-owned coupling diagonal to zero without mutating caller input.
• Trotter/config public surface. Completed 2026-05-06: TrotterEvolutionConfig now exposes typed defaults for evolution order, evolve() Trotter steps, and run() per-step Trotterisation; legacy trotter_order, trotter_steps, and trotter_per_step arguments remain supported and take explicit precedence.
• Expectation hot path. Implemented 2026-05-06: statevector X/Y/Z expectation fallbacks now use bitwise-vectorised NumPy instead of per-qubit dense Kronecker or Qiskit Pauli object construction; Rust remains the preferred accelerated path when the extension is available.
• Rust import failure policy. Implemented 2026-05-06: optional Rust acceleration imports now share optional_rust_engine(), treating only true scpn_quantum_engine absence as optional while propagating broken installed-extension import failures.
• README and public framing sync. Completed 2026-05-06: README, publication framing guide, and hardware status ledger now reflect promoted Phase 2 DLA artefacts, SCPN/FIM negative hardware results, benchmark dashboard/CLI availability, and conservative no-advantage/no-DLA-only/no-FIM-protection boundaries.
• Architecture data-flow diagram. Completed 2026-05-06: docs/architecture.md now documents the stable artefact-first pipeline from K_nm/omega through Hamiltonian compilation, circuit/simulator kernels, CPU/GPU/Rust/QPU execution, raw counts, mitigation, DLA/sync/FIM/VQE observables, and reports/ledgers.
• Curated researcher examples. Completed 2026-05-06: docs/application_benchmarks.md, examples/README.md, and data/public_application_benchmarks/README.md now promote selected GraphML/CSV, EEG, power-grid, plasma/tokamak, and notebook workflows with provenance boundaries and deterministic no-QPU smoke commands.
• Quantum-Kuramoto API-surface contract. Implemented 2026-05-07: scpn-bench s6-api-contract validates the proposed quantum_kuramoto.* export names, current source-module importability, duplicate/export-blocker state, and SCPN-specific warning rows while keeping package-skeleton creation blocked until boundary refactors are complete.

Active paper and submission tasks

• DLA parity preprint submission package. Completed 2026-05-06: docs/publication/dla_parity_submission_checklist_2026-05-06.md freezes the conservative parity-sector/excitation-number framing, required job IDs, committed artefact index, unsupported claims, and final no-QPU pre-upload gate. Updated 2026-05-07 after the Phase 3 state/layout analysis so the checklist and paper source explicitly reject layout-independent and same-sector-state-independent parity-protection claims.
• Rust/VQE methods paper package. Completed 2026-05-06: docs/publication/rust_vqe_methods_submission_checklist_2026-05-06.md freezes the artefact-first table boundary, supported/unsupported claims, generator-script index, scpn-bench reproduce-methods gate, and no-QPU final pre-upload checklist.
• JOSS-style software paper package. Completed 2026-05-06: docs/publication/joss_software_submission_checklist_2026-05-06.md freezes the software-paper claim boundary, JOSS/pyOpenSci framing, metadata gate, reproducibility gate, and no-QPU final pre-upload checklist.
• SCPN/FIM Hamiltonian paper package. Completed 2026-05-06: docs/campaigns/scpn_fim_submission_checklist_2026-05-06.md freezes the negative hardware falsification boundary, committed artefact index, IBM job IDs, scpn-bench fim-all gate, full-basis readout-mitigation scope, and blocked hardware-protection claims.
• Combined submission checklist. Completed 2026-05-06: docs/publication/combined_submission_checklist_2026-05-06.md ties the four paper packages together with final PDF build commands, no-QPU reproduction gates, arXiv metadata draft, URL/identifier checks, and a minimal venue-conditional AI disclosure policy.
• IBM Quantum Credits follow-up. Completed 2026-05-06: <private-internal-record> records the 5--10 hour QPU allocation boundary, current evidence package, affiliation wording, draft locations, and per-run spend gates. The older allocation draft now matches the 5--10 hour scope.

Active Phase 4 follow-up tasks

• One-command reproducibility CLI. Implemented 2026-05-06: scpn-bench reproduce-methods, scpn-bench fim-all, and scpn-bench all regenerate committed benchmark artefacts and report drift without submitting IBM jobs.
• Public benchmark dashboard. Implemented 2026-05-06: docs/methods_benchmark_dashboard.md is wired into MkDocs and links artefacts, generator scripts, provenance, reproducibility commands, optional GPU/scaling/readout harnesses, and no-QPU-spend boundaries.
• Ansatz scaling plus tensor-network baseline: initial harness. Implemented 2026-05-06 with n=4--12 ansatz scaling rows and tensor-network truncation diagnostics.
• Ansatz scaling sparse strengthening. Implemented 2026-05-06 using sparse eigensolver references where feasible for larger-n rows.
• Richer ansatz/TN reference comparisons. Implemented 2026-05-06: benchmark_ansatz_scaling_tn.py now emits generated per-n reference-comparison rows pairing MPS truncation diagnostics with committed VQE aggregate references and marks missing larger-n VQE rows as skipped rather than extrapolated.
• Native or analogue FIM compiler path. Initial lambda_fim compiler payload implemented 2026-05-06 by decomposing -lambda M^2/n into all-to-all Z_i Z_j terms for backend design studies.
• Provider-specific analogue backend export layer. Implemented 2026-05-06: generic analogue programmes can now be exported into Pulser, Bloqade, and IBM pulse-level design payloads with SDK availability metadata, platform-compatibility checks, and explicit can_submit=False boundaries.
• Executable provider analogue backend: approval-gated plan. Implemented 2026-05-06: provider exports can now be wrapped in ProviderAnalogExecutionPlan via prepare_provider_execution_plan, requiring calibration metadata and explicit approval before SDK-object construction or emulator execution is marked possible. Cloud submission remains blocked until a separately approved provider runner exists.
• Adaptive lambda feedback scaffold. Implemented 2026-05-06: AdaptiveFIMConfig, FIMWitness, propose_next_lambda, and adaptive_lambda_schedule.
• Adaptive-QPU protocol. Implemented 2026-05-06: docs/campaigns/adaptive_fim_qpu_protocol_2026-05-06.md defines the non-submitting adaptive lambda_fim hardware boundary, QPU budget gate, live transpilation gate, falsification rules, artefact names, and blocked claims before any IBM submission.
• FIM repeated full-basis readout mitigation. Implemented 2026-05-06 for the repeated dataset where the required 16-state calibration basis exists.
• Readout-mitigation eligibility markers. Implemented 2026-05-06: audit_readout_mitigation_eligibility.py generates data/readout_mitigation_eligibility/readout_mitigation_eligibility_2026-05-06.json with dataset-level and per-n markers for full-basis, partial exact-state, and missing-calibration readout-mitigation eligibility before any new QPU calibration spend.

Hardware experiment candidates

These are candidates, not authorisations to spend QPU time. Each needs offline artefacts, a preregistered manifest, depth/shot gates, and an explicit QPU-time estimate before submission.

• Multi-device DLA replication preregistration. Completed 2026-05-06: docs/campaigns/dla_multidevice_replication_prereg_2026-05-06.md defines the second-Heron backend rule, reduced n=4 circuit matrix, 148-circuit scope, 3--6 minute estimate, 10-minute ceiling, live-depth gates, analysis plan, falsification rules, and output artefact paths. QPU execution remains separate and approval-gated.
• Multi-device DLA replication execution. Completed 2026-05-06 on ibm_marrakesh after live backend selection, transpilation/depth gates, QPU budget confirmation, and explicit approval. Jobs ibm-run-63e0a1af74a38c9c and ibm-run-0f96961442e05a77 produced the raw-count artefact, generated summary, row metrics, and manifest in data/phase3_multidevice_dla/ and docs/campaigns/phase3_multidevice_dla_manifest_2026-05-06.md. Result: mixed/mostly opposite-sign backend-transfer evidence, later strengthened by matching full-basis readout correction, weakening any backend-stable DLA leakage-asymmetry claim.
• Systematic state/layout randomisation preregistration. Completed 2026-05-06: docs/campaigns/dla_state_layout_randomisation_prereg_2026-05-06.md defines the n=4 state/depth/layout matrix, 495-circuit scope, 8--15 minute estimate, 20-minute ceiling, layout-selection gates, analysis plan, readout boundary, falsification rules, and output artefact paths. QPU execution remains separate and approval-gated.
• Systematic state/layout randomisation execution. Completed 2026-05-07 on ibm_marrakesh after live backend/layout selection, transpilation/depth gates, QPU budget confirmation, and explicit approval. Jobs ibm-run-aabcf620230b1438 and ibm-run-eea172711aa52b78 produced the committed raw-count artefact data/phase3_state_layout_dla/phase3_state_layout_ibm_marrakesh_2026-05-06T224531Z.json. Result interpretation is now closed by the generated analysis and manifest artefacts: the original contrast has mixed sign across layout-depth cells, within-sector state controls are significant, and layout spread exceeds the mean original contrast.
• Systematic state/layout randomisation analysis. Completed 2026-05-07: scripts/analyse_phase3_state_layout_dla.py generated the preregistered state/depth/layout leakage summary, row metrics, layout metrics, readout metrics, manifest, decision flags, and claim boundary from the committed raw-count artefact. The DLA parity paper and submission checklist were updated 2026-05-07 to include this mechanism-boundary result.
• Systematic state/layout randomisation live submitter. Implemented 2026-05-07: scripts/phase3_state_layout_dla_ibm.py builds the preregistered 495-circuit state/layout matrix, selects three connected four-qubit windows before outcome data exists, live-transpiles with fixed initial layouts, records readiness artefacts, enforces depth/gate and 20-minute QPU ceilings, and requires both --submit and --confirm-budget before any IBM job is launched.
• Full readout-mitigation calibration preregistration. Completed 2026-05-06: docs/campaigns/readout_full_basis_calibration_prereg_2026-05-06.md defines the eligibility boundary, basis-state calibration matrix for n=4,6,8, QPU-time estimates and ceilings, live layout gates, analysis plan, falsification rules, and output artefact paths. Calibration execution remains separate and approval-gated.
• Full readout-mitigation calibration execution. Completed 2026-05-06 for the Phase 3 ibm_marrakesh n=4 dataset on physical qubits [5,6,7,8] after live readiness checks, budget confirmation, and explicit approval. Job ibm-run-ddd29a2fbcaeed61 produced the full-basis assignment artefacts under data/readout_full_basis/ and manifest docs/campaigns/readout_full_basis_manifest_2026-05-06.md. Calibration quality: mean retention 0.96999, max parity flip 0.04724, condition number 1.07570.
• Phase 3 full-basis readout correction. Completed 2026-05-06: the stable ibm_marrakesh full-basis assignment matrix was applied offline to the Phase 3 DLA rows, generating data/phase3_multidevice_dla/phase3_multidevice_readout_corrected_summary_2026-05-06.json and data/phase3_multidevice_dla/phase3_multidevice_readout_corrected_rows_2026-05-06.csv. The corrected asymmetry is non-positive at all promoted depths, so readout correction does not rescue backend-transfer replication.
• GUESS / symmetry-decay calibration preregistration. Completed 2026-05-06: docs/campaigns/guess_symmetry_decay_prereg_2026-05-06.md defines the folded-noise parity-leakage witness protocol, readiness basis, n=4 circuit matrix, 196-circuit default scope, 5--12 minute estimate, 15-minute ceiling, live folding/depth gates, analysis plan, falsification rules, and output artefact paths. Execution remains separate and approval-gated.
• GUESS / symmetry-decay calibration execution. Completed 2026-05-07 on ibm_marrakesh after backend selection, committed submitter checks, live folded-circuit readiness, conservative 12-minute QPU estimate under the 15-minute ceiling, and explicit approval. Jobs ibm-run-72cd6f926fcf15fc and ibm-run-24d759b97fcbdbac produced the raw-count artefact data/phase3_guess_dla/phase3_guess_ibm_marrakesh_2026-05-06T234602Z.json.
• GUESS / symmetry-decay calibration analysis. Completed 2026-05-07: scripts/analyse_phase3_guess_dla.py generated data/phase3_guess_dla/phase3_guess_summary_2026-05-07.json, fit rows, witness/extrapolation rows, and docs/campaigns/phase3_guess_dla_manifest_2026-05-07.md. Result: 6 raw log-survival fits and 5 exact-state-readout-corrected fits pass the preregistered monotone/R2/RMSE witness criteria; universal GUESS mitigation, backend-general transfer, and full confusion-matrix mitigation remain blocked claims.
• Layer-selective qubit assignment preregistration. Completed 2026-05-06: docs/campaigns/layer_selective_qubit_assignment_prereg_2026-05-06.md defines the coupling-aware scoring rule, comparator layouts, offline readiness matrix, optional 152-circuit hardware follow-up, 4--10 minute estimate, 12-minute ceiling, live gates, analysis plan, falsification rules, and artefact paths. Execution remains separate and approval-gated.
• Layer-selective qubit assignment offline readiness audit. Completed 2026-05-07: scripts/analyse_layer_selective_readiness.py consumes the committed Phase 3 state/layout artefact and generates data/phase3_layer_layout/layer_selective_readiness_ibm_marrakesh_2026-05-07.json, data/phase3_layer_layout/layer_selective_transpile_rows_2026-05-07.csv, and docs/campaigns/phase3_layer_layout_readiness_2026-05-07.md. Decision: blocked_missing_comparators, because the saved artefact has connected low-readout layout rows but not the preregistered default, SABRE, and true layer-selective comparator matrix required before a hardware follow-up can be promoted.
• Layer-selective comparator matrix. Completed 2026-05-07: scripts/generate_layer_selective_comparator_matrix.py opened a no-submit ibm_marrakesh backend snapshot and generated the missing default, SABRE, and true layer-selective transpilation matrix in data/phase3_layer_layout/layer_selective_comparator_matrix_ibm_marrakesh_2026-05-07.json, row CSV, SHA256 sidecar, and docs/campaigns/phase3_layer_layout_comparator_matrix_2026-05-07.md. Decision: blocked_layer_selective_worse_than_default; layer-selective increased max depth by 79.5 % and max two-qubit gates by 46.4 % versus default, so the optional hardware follow-up is not promoted.
• Layer-selective qubit assignment execution. Blocked until a future backend snapshot or revised layer-selective heuristic passes the preregistered comparator matrix. Current 2026-05-07 matrix rejects QPU submission, so running this job now would not isolate a useful layout mechanism.
• Entanglement entropy or tomography check preregistration. Completed 2026-05-06: docs/campaigns/entanglement_tomography_prereg_2026-05-06.md defines the reduced-tomography/shadow-tomography decision rule, small-n claim boundary, offline readiness matrix, optional hardware scope, circuit and QPU-time ceilings, live gates, analysis plan, falsification rules, and artefact paths. Execution remains separate and approval-gated.
• Entanglement entropy or tomography offline readiness. Completed 2026-05-07: scripts/generate_entanglement_tomography_readiness.py generated exact reduced-Pauli reference values, half-chain purity proxies, basis settings, and circuit-count gates for DLA parity plus FIM n=4 families. Artefacts: data/phase3_entanglement_tomography/entanglement_tomography_readiness_2026-05-07.json, data/phase3_entanglement_tomography/entanglement_observable_rows_2026-05-07.csv, and docs/campaigns/phase3_entanglement_tomography_readiness_2026-05-07.md. Decision: ready_for_optional_hardware_preregistration; the promoted reduced-tomography block uses 9 basis settings and 166 total circuits, so full tomography is unnecessary and the optional hardware block is scientifically promotable after backend selection, live transpilation gates, budget confirmation, and explicit approval.
• Entanglement entropy or tomography hardware execution. Completed 2026-05-20 on ibm_marrakesh after live preflight, budget-confirmed approval, and the guarded submit command. Jobs d86g7h1789is738vkreg and d86ggpis46sc73f6v170 produced the completed raw-count artefact data/phase3_entanglement_tomography/entanglement_tomography_live_ibm_marrakesh_2026-05-20T004334Z.json. First-pass analysis generated data/phase3_entanglement_tomography/entanglement_tomography_summary_2026-05-20.json, data/phase3_entanglement_tomography/entanglement_tomography_rows_2026-05-20.csv, and docs/campaigns/phase3_entanglement_tomography_manifest_2026-05-20.md. Result snapshot: 54 observable rows, mean absolute deviation 0.12989296537986128, maximum absolute deviation 0.5560906424788263, bounded to reduced-Pauli small-system mechanism analysis only.
• Entanglement tomography manuscript tables and figure assets. Completed 2026-05-20: scripts/generate_phase3_entanglement_paper_assets.py generates reproducible label-level, basis-level, and largest-deviation paper tables plus the measured-minus-exact heatmap figures/phase3/phase3_entanglement_deviation_heatmap_2026-05-20.png. The canonical LaTeX manuscript now includes the result tables, heatmap reference, discussion, and conservative mechanism-boundary interpretation.
• Entanglement tomography readout, replication, and ZNE stress-test extension. Completed 2026-05-20: same-day ibm_fez replication, pinned-layout full 16-state correlated readout calibration, and the preregistered five-channel ZNE subset were executed and reduced. ZNE jobs d86hs6qs46sc73f70h90 and d86hsltg7okc73el4lg0 produced data/phase3_entanglement_tomography/entanglement_tomography_live_ibm_fez_2026-05-20T023600Z.json; scripts/analyse_phase3_entanglement_zne.py generated data/phase3_entanglement_tomography/entanglement_zne_summary_2026-05-20_ibm_fez_zne.json, scale rows, channel summary, and docs/campaigns/phase3_entanglement_zne_manifest_2026-05-20_ibm_fez_zne.md. Result: simple global-folding ZNE does not erase the four dominant DLA transverse deviations; the FIM control channel behaves differently and improves under linear extrapolation. Remaining paper work is venue-specific formatting and bibliography polish, not additional QPU spend for this paper.
• Entanglement tomography third-backend ZNE replication. Completed 2026-05-20 on ibm_kingston after ibm_sherbrooke and ibm_torino were unavailable to the account. Readiness passed for physical qubits [141,142,143,144], 45 main ZNE circuits, 16 full-readout circuits, max depth 1686, and estimated QPU time 0.5592 minutes under the 25-minute ceiling. Jobs d86i8fas46sc73f70vg0 and d86ijr9789is738vnh30 completed; counts were hydrated into the canonical artefact and reduced with the Kingston ZNE tag. Artefacts: data/phase3_entanglement_tomography/entanglement_tomography_live_ibm_kingston_2026-05-20T030151Z.json, data/phase3_entanglement_tomography/entanglement_tomography_live_ibm_kingston_2026-05-20T030211Z.json, data/phase3_entanglement_tomography/entanglement_zne_summary_2026-05-20_ibm_kingston_zne.json, data/phase3_entanglement_tomography/entanglement_zne_channel_summary_2026-05-20_ibm_kingston_zne.csv, and docs/campaigns/phase3_entanglement_zne_manifest_2026-05-20_ibm_kingston_zne.md. Result: Kingston reproduces the qualitative Fez ZNE stress-test pattern: the four DLA transverse channels move farther from exact under linear ZNE, while the FIM control improves. Remaining paper work is manuscript/PDF integration from the committed Kingston artefacts.
• Entanglement tomography Kingston ZNE statistical repeat. Completed 2026-05-20 as a same-backend, same-layout statistical repeat for the Phase 3 paper. The repeat keeps ibm_kingston, physical qubits [141,142,143,144], the preregistered five-channel ZNE subset, noise scales 1,3,5, three repetitions per channel/scale, and full 16-state readout calibration fixed. Readiness passed with 45 main circuits, 16 readout circuits, max depth 1657, and estimated QPU time 0.5592 minutes under the 25-minute ceiling. Jobs completed: main d86pul0p0eas73dla3dg, readout d86pul8p0eas73dla3eg. Canonical completed artefact: data/phase3_entanglement_tomography/entanglement_tomography_live_ibm_kingston_2026-05-20T114719Z.json. Reducer outputs: data/phase3_entanglement_tomography/entanglement_zne_summary_2026-05-20_ibm_kingston_zne_repeat.json, scale rows, channel summary, and docs/campaigns/phase3_entanglement_zne_manifest_2026-05-20_ibm_kingston_zne_repeat.md. Result: the same-layout repeat preserves the completed Kingston pattern: all four DLA transverse channels move farther from exact under linear ZNE, while the FIM control improves.
• Depth-optimal native decomposition preregistration. Completed 2026-05-06: docs/campaigns/depth_optimal_native_decomposition_prereg_2026-05-06.md defines comparator circuits, native-target candidate rules, offline readiness matrix, equivalence gates, optional 160-circuit hardware follow-up, 12-minute ceiling, live gates, analysis plan, falsification rules, and artefact paths. Execution remains separate and approval-gated.
• Depth-optimal native decomposition offline readiness. Completed 2026-05-07: scripts/generate_native_decomposition_readiness.py generated local basis-gate resource rows plus unitary/observable equivalence rows for generic Pauli evolution, current XY compiler, and native-targeted rxx+ryy decomposition across n=4,6,8 readiness cases. Artefacts: data/phase3_native_decomposition/native_decomposition_readiness_2026-05-07.json, data/phase3_native_decomposition/native_decomposition_transpile_rows_2026-05-07.csv, data/phase3_native_decomposition/native_decomposition_equivalence_rows_2026-05-07.csv, and docs/campaigns/phase3_native_decomposition_readiness_2026-05-07.md. Decision: blocked_current_xy_invalid_no_native_gain_vs_generic. The native-targeted rxx+ryy path passes equivalence but has no median depth or two-qubit-gate gain versus the generic Pauli baseline, while the current XY compiler comparator fails equivalence and cannot be used as a valid resource baseline. Optional hardware execution is therefore not promoted.
• Depth-optimal native decomposition hardware execution. Blocked until a revised equivalent candidate shows a resource gain in the offline gate, then passes backend selection, live transpilation, budget confirmation, and explicit approval for QPU submission.
• Variational quantum simulation alternative preregistration. Completed 2026-05-06: docs/campaigns/vqs_alternative_prereg_2026-05-06.md defines candidate VQS modes, offline readiness matrix, promotion gates, tolerances, optional hardware scope, QPU-time ceilings, live gates, analysis plan, falsification rules, and artefact paths. Execution remains separate and approval-gated.
• Variational quantum simulation alternative offline readiness. Completed 2026-05-07: scripts/generate_vqs_alternative_readiness.py generated exact-state VQS refit rows for n=4 DLA, popcount, and FIM cases plus local basis-gate resource rows for n=4,6,8 Trotter and VQS candidate circuits. Artefacts: data/phase3_vqs_alternative/vqs_readiness_2026-05-07.json, data/phase3_vqs_alternative/vqs_candidate_rows_2026-05-07.csv, data/phase3_vqs_alternative/vqs_resource_rows_2026-05-07.csv, and docs/campaigns/phase3_vqs_alternative_readiness_2026-05-07.md. Decision: blocked_no_vqs_candidate_passed_promotion_gate. No shallow VQS ansatz family passed both the preregistered target-observable accuracy gate and the compiled-resource gate, so hardware execution is not promoted.
• Variational quantum simulation alternative hardware execution. Blocked until a revised VQS candidate passes the offline promotion gate, backend selection, live transpilation, budget confirmation, and explicit approval for QPU submission.
• Multi-circuit QEC demonstration preregistration. Completed 2026-05-06: docs/campaigns/multicircuit_qec_prereg_2026-05-06.md defines required baselines, logical-error metrics, observable tolerances, offline readiness matrix, promotion gates, optional 180-circuit hardware scope, 15-minute ceiling, live gates, analysis plan, ablations, falsification rules, and artefact paths. Execution remains separate and approval-gated.
• Multi-circuit QEC demonstration offline readiness. Completed 2026-05-07: scripts/generate_multicircuit_qec_readiness.py generated Monte Carlo logical-failure rows for the unencoded physical baseline, standard MWPM decoder, K-matrix-weighted physics-aware decoder, and feature-disabled ablation across ideal, depolarising, and readout-biased noise models. It also generated encoded/unencoded circuit-resource rows for the DLA parity pair. Artefacts: data/phase3_multicircuit_qec/qec_readiness_2026-05-07.json, data/phase3_multicircuit_qec/qec_decoder_rows_2026-05-07.csv, data/phase3_multicircuit_qec/qec_resource_rows_2026-05-07.csv, and docs/campaigns/phase3_multicircuit_qec_readiness_2026-05-07.md. Decision: blocked_physics_aware_decoder_did_not_beat_baselines. Hardware execution is not promoted because the physics-aware decoder did not beat the standard decoder and its own feature-disabled ablation under the preregistered logical metric.
• Multi-circuit QEC offline-boundary terminology hardening. Implemented 2026-05-12: the readiness generator and committed artefacts now identify the encoded comparator as distance3_surface_code_offline and describe the supported claim as a distance-3 surface-code offline logical-failure comparison, avoiding imprecise informal language while preserving the no-hardware and no-fault-tolerance boundary.
• Multi-circuit QEC demonstration hardware execution. Blocked until a revised QEC/decoder candidate passes offline logical metrics against unencoded, standard-decoder, and ablation baselines, then passes backend selection, live transpilation, budget confirmation, and explicit approval for QPU submission.

Visibility and registration tasks

• GitHub topics. Completed 2026-04-17.
• QOSF awesome-quantum-software. Merged 2026-03-30.
• CiteAs verification. Verified 2026-04-17.
• PyPI publication. Completed for the current package line.
• OpenSSF Best Practices. Passing badge is present.
• Zenodo DOI. Existing DOI is live.
• GitHub Pages docs. Published.
• Software Heritage SWHID follow-up. Completed 2026-05-06: docs/publication/software_heritage_swhid_2026-05-06.md records the successful Software Heritage save request, full visit status, origin SWHID, and archived snapshot SWHID.
• Zenodo communities and metadata refresh preparation. Completed 2026-05-06: <private-internal-record> records the DOI, current version consistency, community targets, related identifiers, claim boundary, and manual-session update procedure.
• Zenodo metadata refresh execution. Completed 2026-05-06: <private-internal-record> records the authenticated Zenodo API edit/publish cycle. Public metadata now reports version 0.9.6, publication date 2026-03-29, license agpl-3.0-or-later, bounded description, and refreshed keywords for record 10.5281/zenodo.18821930 under concept DOI 10.5281/zenodo.18821929.
• Zenodo community submission UI follow-up. Completed 2026-05-06: <private-internal-record> records the pending Zenodo community-inclusion request for the Research Software Engineering community. Public record community membership remains pending curator acceptance.
• Qiskit Ecosystem Catalog. Submitted 2026-05-06: docs/publication/qiskit_ecosystem_submission_2026-05-06.md records the submitted project metadata and open review issue Qiskit/ecosystem#1123.
• awesome-qiskit. Blocked until Qiskit Ecosystem membership is accepted.
• Conda-forge recipe. Submitted 2026-05-06: docs/publication/conda_forge_submission_2026-05-06.md records the staged-recipes recipe metadata and open PR conda-forge/staged-recipes#33236.
• Metriq local readiness smoke. Completed 2026-05-06: docs/publication/metriq_local_smoke_2026-05-06.md records a no-QPU metriq-gym Bernstein--Vazirani local simulator dispatch and poll using the isolated /home/anulum/.venvs/scpn-metriq environment. This proves the Metriq CLI path is executable without treating SCPN paper benchmark tables as Metriq-native results.
• Metriq submission decision. Completed 2026-05-07: docs/publication/metriq_submission_decision_2026-05-07.md records a deliberate no-upload decision. The Metriq-Gym upload path was verified in dry-run mode for the local Bernstein--Vazirani smoke result, but no public upload was made because the available artefact is a generic local simulator result rather than an SCPN/Kuramoto--XY benchmark. Arbitrary Rust/VQE/DLA/FIM project tables remain out of scope for Metriq upload unless an accepted Metriq schema exists.
• Metriq SCPN benchmark schema proposal preparation. Completed 2026-05-07: docs/publication/metriq_scpn_benchmark_schema_proposal_2026-05-07.md defines a bounded Kuramoto--XY parity-leakage benchmark proposal with required inputs, circuit definition, primary score, secondary metrics, acceptance gates, non-claim boundary, and upstream draft text. It is a prepared proposal only; it has not been submitted upstream and is not an accepted Metriq-Gym schema.
• Metriq SCPN benchmark schema upstream submission. Optional external follow-up: submit the prepared schema proposal to Metriq-Gym maintainers, wait for review or acceptance, then run and upload results only under the accepted schema.
• pyOpenSci review. Submit the software package for review and possible JOSS fast-track.
• pyOpenSci review preparation. Completed 2026-05-06: docs/publication/pyopensci_submission_readiness_2026-05-06.md records the package scope, unsupported claims, metadata gate, reviewer evidence, no-QPU pre-submission gates, and suggested issue summary. External issue submission remains open until an issue URL is recorded.
• JOSS submission. Submit after paper package and metadata are aligned.
• arXiv submission. Submit the paper set once final PDFs, references, source tarballs, and artefact links are aligned.
• arXiv source packaging preparation. Completed 2026-05-06: docs/publication/arxiv_source_packaging_readiness_2026-05-06.md records the public source-bundle boundaries, build gates, claim gates, metadata draft, and private-file exclusions for the DLA parity, Rust/VQE methods, and SCPN/FIM papers. Actual arXiv upload remains open until explicitly approved and recorded.
• SciPy 2026 CFP preparation. Completed 2026-05-06: docs/publication/scipy_2026_cfp_readiness_2026-05-06.md records the proposed SciPy angle, format options, title options, draft abstract, reviewer-visible evidence, claims to avoid, and no-QPU submission boundary. Actual CFP submission remains external and should only be marked complete after the current CFP page and submitted proposal URL are recorded.
• Community announcements. Prepare Reddit, Qiskit Slack, Unitary Discord, Hacker News, LinkedIn, and X posts only after the public preprints are live.
• Registry-listing preparation pack. Completed 2026-05-06: docs/publication/registry_listing_plan_2026-05-06.md records the canonical metadata, target-by-target readiness, account/manual blockers, submission copy, and do-not-submit-yet boundaries for Quantiki, QOSF, best-of-python, Papers With Code, SciCrunch RRID, Open Hub, and Research Software Directory.
• Community-announcement preparation pack. Completed 2026-05-06: <private-internal-record> prepares bounded Reddit, Qiskit Slack, Unitary Discord, Hacker News, LinkedIn, and X copy while keeping publication deferred until public preprint links are live.
• Registry listings. Execute only the entries that pass their readiness gates. QOSF is already listed in qosf/awesome-quantum-software; Papers With Code and community announcements remain blocked until arXiv/JOSS links exist; Quantiki, SciCrunch RRID, Open Hub, and Research Software Directory require account/manual form completion.

Deferred / CEO-gated strategic tracks

These tracks remain scoped in docs/strategic_roadmap.md and must not be executed until individually activated.

• S1 Hybrid classical--quantum feedback loop.
• S1 cross-shot feedback-loop foundation. Implemented 2026-05-06: hardware/feedback_loop.py adds scheduler/observer protocols, FeedbackRunner, step records, latency and QPU-budget gates, explicit hardware approval enforcement, and a proportional metric observer. Tests and documentation are in tests/test_feedback_loop.py and docs/campaigns/hybrid_feedback_loop_s1_2026-05-06.md. This does not submit IBM jobs and does not claim intra-shot feedback.
• S1 realtime-controller simulator scheduler. Implemented 2026-05-06: RealtimeControllerScheduler wraps RealtimeSyncFeedbackController as a zero-QPU scheduler, adds bounded cross-shot coupling overrides, deterministic finite-shot seeds, auditable simulator metrics, focused tests, and S1 documentation.
• S1 no-QPU feedback latency benchmark. Implemented 2026-05-06: scripts/benchmark_s1_feedback_loop.py and scpn-bench s1-feedback regenerate JSON/CSV latency artefacts for FeedbackRunner plus RealtimeControllerScheduler; pipeline performance documentation records the command and IBM/QPU boundary.
• S1 provider-neutral submission-readiness package. Implemented 2026-05-06: hardware/feedback_submission.py builds no-submission dynamic-circuit readiness packages with circuit summaries, QPU-budget estimates, platform capability checks, ready/blocked/manual-review decisions for IBM, generic gate, analogue, CV, and simulator targets, focused tests, and S1 documentation.
• Hardware-job dossier standard. Implemented 2026-05-06: hardware/job_dossier.py defines the required dossier schema for every submission-ready hardware job, including purpose, hypothesis, falsification condition, observables, circuit summary, QPU budget, platform fit, risks, decision tree, paper impact, follow-up avenues, possibilities opened, claim boundary, and reproducibility package. The S1 readiness package embeds the dossier by default.
• S1 preregistration manifest export. Implemented 2026-05-06: scripts/export_s1_feedback_preregistration.py exports JSON and Markdown preregistration manifests from the provider-neutral S1 package and embedded hardware-job dossier. The default budget explicitly includes a monitored feedback arm and a matched open-loop control arm; no credentials are read and no hardware job is submitted.
• S1 provider dry-run payloads. Implemented 2026-05-06: hardware/feedback_dryrun.py emits no-submit payloads for IBM Runtime dynamic circuits, provider-neutral OpenQASM 3 style gate execution, and analogue-native review. The preregistration export embeds the dry-run bundle and keeps analogue targets behind a separate native-feedback dossier requirement.
• S1 approval-gated hardware scheduler boundary. Implemented 2026-05-06: hardware/feedback_hardware_scheduler.py provides a fail-closed scheduler wrapper requiring an injected provider submitter, explicit approval, matching provider, matching preregistration package hash, and estimated/reported QPU budget compliance before any hardware submission can pass.
• S1 no-submit capability probes. Implemented 2026-05-06: hardware/feedback_capability_probe.py evaluates backend metadata snapshots against the S1 dynamic-circuit package for qubit count, shot/circuit limits, mid-circuit measurement, conditional control, conditional reset, and cross-shot batch support. The preregistration export embeds template probe decisions without reading credentials or submitting jobs.
• S1 raw-count analysis harness. Implemented 2026-05-06: scripts/analyse_s1_feedback_hardware.py defines the preregistered feedback-vs-matched-open-loop-control analysis before live hardware submission, including raw-count schema checks, per-arm summaries, target-error improvement, decision boundary, and claim boundary tests.
• S1 synthetic analysis rehearsal fixture. Implemented 2026-05-06: data/s1_feedback_loop/s1_feedback_synthetic_raw_counts_2026-05-06.json provides a non-hardware fixture for exercising the preregistered S1 raw-count analysis path, and the S1 documentation records the exact raw-count JSON schema required for live packages.
• S1 no-submit provider metadata adapters. Implemented 2026-05-06: hardware/feedback_provider_metadata.py converts provider-neutral metadata records and Qiskit-style backend objects into BackendCapabilitySnapshot inputs for the S1 capability probes without reading credentials, opening provider sessions, or submitting jobs.
• S1 one-command readiness bundle. Implemented 2026-05-06: scripts/reproduce_s1_feedback_readiness.py and scpn-bench s1-feedback-ready regenerate the no-QPU S1 latency, preregistration, provider dry-run, capability-example, and synthetic analysis artefacts in one command.
• S1 live-submission preflight checklist. Implemented 2026-05-06: docs/campaigns/s1_live_submission_preflight_2026-05-06.md records the mandatory manual gates for artefacts, scientific purpose, provider capability, budget, reproducibility, approval records, stop conditions, and post-run handling before any live S1 provider submitter may be wired.
• S1 IBM metadata probe command. Implemented 2026-05-06: scripts/probe_s1_ibm_metadata.py writes no-submit capability decisions from offline provider metadata JSON or an already-authenticated Qiskit Runtime backend metadata lookup. The command does not accept credential strings, does not submit jobs, and records hardware_submission=false.
• S1 generic gate metadata probe command. Implemented 2026-05-06: scripts/probe_s1_generic_gate_metadata.py writes no-submit, no-network capability decisions for non-IBM gate-based targets from provider-neutral metadata JSON and records hardware_submission=false plus network_access=false.
• S1 readiness index. Implemented 2026-05-06: docs/campaigns/s1_feedback_readiness_index_2026-05-06.md consolidates the S1 no-QPU readiness state, artefact inventory, exact commands, preregistered job shape, claim boundary, platform interpretation, and remaining live-submission blockers.
• S1 IBM feedback-control paper scaffold. Implemented 2026-05-20: docs/campaigns/s1_feedback_ibm_paper_plan_2026-05-20.md and paper/submissions/submission_006_s1_feedback_control/s1_feedback_control_short_paper.md define the next no-QPU preparation lane for a monitored-feedback versus matched-open-loop IBM dynamic-circuit paper. The package records the 4-qubit, 2-circuit, 1024-shot, 12-repetition, 24-second preregistered job shape, claim boundary, decision tree, and remaining live metadata, transpilation, approval, raw-count, and analysis gates. It does not authorise or perform an S1 hardware submission.
• S2 Quantum advantage benchmarks at scale.
• S2 scaling protocol manifest. Implemented 2026-05-06: benchmarks/advantage_protocol.py and scripts/export_s2_scaling_protocol.py define the S2 no-claim scaling protocol, required baselines, size grid, output schema, acceptance rules, falsification rules, and claim boundary before any heavy sweep or hardware row is promoted.
• S2 scaling row validator. Implemented 2026-05-06: validate_scaling_rows and scripts/validate_s2_scaling_rows.py enforce the preregistered S2 row schema, required baselines, known baseline labels, valid statuses, and wall-time requirements for successful rows before any scaling table or figure is promoted.
• S2 validator matrix completeness hardening. Implemented 2026-05-12: validate_scaling_rows now enforces every required baseline per observed size, rejects off-protocol sizes and protocol ids, requires finite non-negative timing plus memory for ok rows, and requires explanatory notes for skipped or failed rows.
• S2 validator provenance hardening. Implemented 2026-05-12: duplicate (n_qubits, baseline) rows are rejected, and row-level provenance payloads must use structured metric, command, machine, dependency, git-commit, and notes fields before a scaling matrix can pass validation.
• S2 lite scaling harness. Implemented 2026-05-06: scripts/bench_s2_scaling_lite.py emits protocol-compliant rows for small selected sizes, measures cheap classical ODE and dense exact diagonalisation rows, records explicit skipped rows for heavier required baselines, validates the output, and records no hardware submission or advantage claim.
• S2 lite sparse eigensolver row. Implemented 2026-05-06: the lite scaling harness now measures gated sparse_eigsh rows for small sizes, including ground energy, residual norm, Hilbert dimension, wall time, and memory estimate, instead of treating sparse classical support as skipped in the rehearsal path.
• S2 lite MPS/TN spoofability row. Implemented 2026-05-06: the lite scaling harness now measures small-size tensor-network spoofability diagnostics from exact ground-state Schmidt spectra, including max bond, worst-cut discarded weight, midchain entropy, Hilbert dimension, wall time, and memory estimate.
• S2 lite Aer/statevector row. Implemented 2026-05-06: the lite scaling harness now measures gated Aer/statevector-style statevector evolution rows through QuantumKuramotoSolver, including Trotter steps, circuit depth, final synchronisation observables, Hilbert dimension, wall time, and memory estimate.
• S2 benchmark CLI wiring. Implemented 2026-05-06: scpn-bench s2-scaling-lite regenerates the S2 scaling protocol and lite scaling rows through the canonical benchmark CLI without hardware submission or advantage claims.
• S2 claim-boundary report. Implemented 2026-05-06: scripts/report_s2_scaling_claim_boundary.py reads S2 rows, validates them, and emits JSON/Markdown reports listing allowed claims, forbidden claims, remaining blockers, validation state, hardware submission status, and advantage-claim status.
• S2 IBM advantage-readiness hard gate. Implemented 2026-05-18: the S2 claim-boundary report now emits ibm_readiness, which blocks meaningful IBM advantage spend unless row validation passes, every required classical/simulator baseline is ok at every protocol size, and at least one preregistered qpu_hardware row with raw-count provenance is present. Lite rows and partial no-QPU campaign slices remain useful science but cannot justify IBM advantage claims.
• S2 benchmark-matrix audit ingestion. Implemented 2026-05-18: scripts/run_quantum_advantage_gap_audit.py now reads the committed S2 protocol, progress, and claim-boundary artifacts, reports missing size grid entries and remaining blockers, and refuses IBM advantage readiness while the no-QPU matrix, hardware rows, or claim gates remain incomplete.
• S2 scaling readiness index. Implemented 2026-05-06: docs/campaigns/s2_scaling_readiness_index_2026-05-06.md summarises the S2 lite baseline rows, canonical command, artefacts, allowed claims, forbidden claims, full-campaign blockers, hardware boundary, and next non-QPU scaling step.
• S2 lite memory instrumentation and size gates. Implemented 2026-05-06: scripts/bench_s2_scaling_lite.py now records tracemalloc peak bytes for measured rows, keeps estimated dense or statevector bytes in metric payloads, and exposes explicit dense, sparse, tensor-network, and statevector size gates.
• S2 full-campaign execution plan. Implemented 2026-05-07: scripts/plan_s2_full_scaling_campaign.py enumerates the full N=4,6,8,10,12,14,16,18,20 scaling matrix across every protocol baseline, classifies rows as lite-measured, ready for deliberate no-QPU full-campaign execution, size-gated, optional GPU, or blocked optional hardware, and generates data/s2_advantage_scaling/s2_full_campaign_plan_2026-05-07.json, data/s2_advantage_scaling/s2_full_campaign_rows_2026-05-07.csv, and docs/campaigns/s2_full_campaign_plan_2026-05-07.md. Decision: ready_for_deliberate_no_qpu_full_classical_campaign; hardware rows and broad quantum-advantage language remain blocked.
• S2 bounded full-campaign execution slice. Implemented 2026-05-07: scripts/run_s2_full_campaign_slice.py consumes the S2 campaign plan and executes only no-QPU required rows under explicit dense, sparse, tensor-network, and statevector size caps. The default n=8 slice generated data/s2_advantage_scaling/s2_full_campaign_slice_n8_2026-05-07.json, data/s2_advantage_scaling/s2_full_campaign_slice_rows_n8_2026-05-07.csv, and docs/campaigns/s2_full_campaign_slice_n8_2026-05-07.md. Decision: completed_no_qpu_campaign_slice; this is not the full campaign, not hardware evidence, and not a quantum-advantage claim.
• S2 bounded full-campaign execution slice, n=10. Implemented 2026-05-07: the same no-QPU slice runner executed the next bounded n=10 scaling slice and generated data/s2_advantage_scaling/s2_full_campaign_slice_n10_2026-05-07.json, data/s2_advantage_scaling/s2_full_campaign_slice_rows_n10_2026-05-07.csv, and docs/campaigns/s2_full_campaign_slice_n10_2026-05-07.md. The slice recorded 5 executed rows, 5 successful rows, 0 skipped rows, and decision completed_no_qpu_campaign_slice; this is still not the full campaign, not hardware evidence, and not a quantum-advantage claim.
• S2 bounded full-campaign execution slice, n=12. Implemented 2026-05-07: the no-QPU slice runner executed the next bounded n=12 scaling slice with dense, sparse, tensor-network, and statevector caps all set to 12 qubits. It generated data/s2_advantage_scaling/s2_full_campaign_slice_n12_2026-05-07.json, data/s2_advantage_scaling/s2_full_campaign_slice_rows_n12_2026-05-07.csv, and docs/campaigns/s2_full_campaign_slice_n12_2026-05-07.md. The slice recorded 5 executed rows, 5 successful rows, 0 skipped rows, and decision completed_no_qpu_campaign_slice; this is still not the full campaign, not hardware evidence, and not a quantum-advantage claim.
• S2 slice progress aggregation. Implemented 2026-05-07: scripts/report_s2_slice_progress.py aggregates the completed bounded no-QPU S2 slices for n=8,10,12 from committed JSON/CSV artefacts and writes data/s2_advantage_scaling/s2_slice_progress_report_2026-05-07.json plus docs/campaigns/s2_slice_progress_report_2026-05-07.md. The aggregate report records 15 successful rows out of 15 executed rows, total measured wall time 510907.947844 ms, maximum recorded memory 2416024559 bytes, and decision ready_for_next_bounded_no_qpu_slice; this is still not hardware evidence, not full S2 completion, and not a quantum-advantage claim.
• S2 n=14 resource gate. Implemented 2026-05-07: scripts/report_s2_n14_resource_gate.py compares the planned n=14 required rows against the committed n=8,10,12 progress artefacts and writes data/s2_advantage_scaling/s2_n14_resource_gate_2026-05-07.json plus docs/campaigns/s2_n14_resource_gate_2026-05-07.md. The report records a dense-matrix estimate of 4294967296 bytes, prior maximum recorded memory 2416024559 bytes, dense/prior-memory ratio 1.7777, and decision blocked_for_scheduled_or_offloaded_no_qpu_run; this is a resource gate, not an n=14 execution result, not hardware evidence, and not a quantum-advantage claim.
• S2 ML350 n=14 full no-QPU slice. Implemented 2026-05-07: the resource-gated n=14 slice was executed on ML350 in the non-SAS workspace /home/anulum/scpn_quantum_control_ntfs_worktree from NTFS source commit 842c529da89f88b68a75582fba7a0076a1a34c1f. The generated artefacts are data/s2_advantage_scaling/s2_full_campaign_slice_n14_2026-05-07.json, data/s2_advantage_scaling/s2_full_campaign_slice_rows_n14_2026-05-07.csv, and docs/campaigns/s2_full_campaign_slice_n14_2026-05-07.md. The slice recorded 5 executed rows, 5 successful rows, 0 skipped rows, decision completed_no_qpu_campaign_slice, all caps set to 14 qubits, and no QPU time. The ML350 timing log reported wall time 3:03:29, peak RSS 21583552 KiB, and 0 swaps. This is still not hardware evidence, not full S2 completion, and not a quantum-advantage claim.
• S3 ML-augmented pulse / ansatz design.
• S3 deterministic pulse/ansatz design-readiness gate. Implemented 2026-05-06: benchmarks/s3_design_protocol.py and scripts/export_s3_design_readiness.py define a no-QPU candidate ranking protocol for structured Kuramoto-XY ansatz candidates and hypergeometric pulse-schedule candidates. The artefact records hardware_submission=false, ml_training_performed=false, allowed claims, forbidden claims, and required follow-ups before any ML surrogate training or pulse-level hardware work.
• S3 benchmark CLI wiring. Implemented 2026-05-06: scpn-bench s3-design-ready regenerates the deterministic S3 design-readiness JSON/Markdown artefacts through the canonical benchmark CLI without hardware submission.
• S3 design readiness index. Implemented 2026-05-06: docs/campaigns/s3_design_readiness_index_2026-05-06.md documents the candidate families, canonical command, allowed claims, forbidden claims, and the next S3 steps: held-out surrogate training, VQE or observable validation, provider-specific pulse feasibility probes, and hardware-job dossiers before execution.
• S3 held-out surrogate rehearsal. Implemented 2026-05-06: scripts/train_s3_design_surrogate.py expands the deterministic pulse/ansatz candidate grid across small system sizes, trains a closed-form ridge linear surrogate on proxy scores, and reports train, held-out, and per-family metrics. The artefact remains no-QPU, proxy-labelled, and explicitly non-evidential for hardware pulse or VQE improvement.
• S3 promoted ansatz observable validation. Implemented 2026-05-06: scripts/validate_s3_ansatz_observables.py checks lowest-resource promoted ansatz candidates against exact statevector energy expectation, dense exact ground energy, energy error, and a synchronisation proxy. The artefact is no-QPU and explicitly not VQE optimisation, pulse validation, hardware evidence, or an advantage claim.
• S3 provider-specific pulse feasibility probes. Implemented 2026-05-06: hardware/pulse_feasibility.py and scripts/probe_s3_pulse_feasibility.py assess provider metadata against the S3 hypergeometric pulse schedule for qubit count, pulse count, duration, sample spacing, pulse-control support, and native-XY support. The probe opens no provider session and submits no job.
• S3 promoted-candidate hardware-job dossiers. Implemented 2026-05-06: scripts/export_s3_hardware_dossiers.py packages the promoted ansatz and pulse follow-up routes into the standard hardware dossier schema, including purpose, hypothesis, falsification boundary, observables, QPU budget state, platform fit, risks, decision tree, paper impact, follow-up avenues, prerequisites, and reproducibility artefacts. The dossiers do not authorise hardware execution.
• S4 Multi-hardware backend + pulse-level control.
• S4 multi-hardware no-submit readiness gate. Implemented 2026-05-06: scpn-bench s4-multi-hardware-ready exports provider-specific Pulser, Bloqade, and IBM pulse-level design payloads from the analogue Kuramoto compiler, wraps them in approval-gated execution plans, documents blocked claims and promotion gates, and performs no provider contact or QPU submission.
• S4 IBM pulse-level preregistration dossier. Implemented 2026-05-06: scpn-bench s4-provider-preregistration turns the IBM pulse-level design payload into a no-submit calibration-review dossier with purpose, hypothesis, falsification condition, expected capability observables, blocked QPU budget, platform fit, decision tree, paper impact, follow-up route, and prerequisites before any pulse job.
• S4 neutral-atom preregistration dossier. Implemented 2026-05-06: scpn-bench s4-neutral-atom-preregistration packages the Pulser and Bloqade neutral-atom payloads into a no-submit provider-object review dossier with geometry, SDK, emulator-only, comparator-observable, cost/credit, and QPU-budget gates before any cloud provider session.
• S5 Open-data + classical validation harness.
• S5 public Phase 1 benchmark harness facade. Implemented 2026-05-06: scpn_quantum_control.benchmark_harness exposes load_phase1_dataset, reproduce_phase1_statistics, and run_phase1_benchmark as the community-facing open-data API over the committed Phase 1 DLA-parity raw counts and noiseless classical parity-conservation reference.
• S5 benchmark-suite CLI artefact. Implemented 2026-05-06: scpn-bench s5-benchmark-suite regenerates the Phase 1 benchmark harness JSON/Markdown artefacts from committed raw counts, verifies published statistics against the tolerance bundle, records the classical baseline, and performs no QPU submission.
• S5 benchmark registry/index. Implemented 2026-05-06: scpn-bench s5-benchmark-registry exports a public benchmark-family registry distinguishing the implemented Phase 1 DLA-parity benchmark from planned CHSH, BKT, OTOC, and DLA-dimension entries, with blockers and claim boundaries so unavailable rows are not mistaken for results.
• S6 Decoupled quantum-kuramoto subpackage.
• S6 import-graph split audit. Implemented 2026-05-07: scpn-bench s6-split-audit inventories candidate phase, bridge, hardware, and accel modules, classifies each row as reusable, needs-review, or SCPN-specific, records blockers and next steps, and explicitly blocks second-package publication until the boundary is manually reviewed.
• S6 boundary-review report. Implemented 2026-05-07: scpn-bench s6-boundary-review converts the split audit into a conservative public-API proposal, defers config/provenance/analysis dependent rows, records compatibility requirements, and keeps the quantum_kuramoto package skeleton blocked until refactors close.
• S7 Fault-tolerant / logical-level extension roadmap. Implemented 2026-05-20: scpn-bench s7-logical-dla-roadmap exports the S7 logical-DLA parity resource table and public roadmap note. The gate estimates flat surface-code resources for 16 logical oscillators at distances 3, 5, and 7, compares the existing MS-QEC hierarchy, and keeps DLA parity survival under logical encoding blocked until the representation-theory and logical-observable prerequisites are closed.
• S8 Mid-circuit adaptive branching readiness. Implemented 2026-05-20: scpn-bench s8-adaptive-branching-readiness exports the no-submit S8 branch-policy table and readiness note. The gate records local-order, DLA-parity-leakage, and chimera-cluster branch policies, requires mid-circuit measurement plus conditional control/reset and cross-shot batching before live execution, and keeps adaptive-advantage claims blocked until the preregistered equal-depth open-loop comparison passes.
• S9 Quantum thermodynamics of synchronisation transitions readiness. Implemented 2026-05-20: scpn-bench s9-quantum-thermo-readiness exports the no-submit entropy-production, heat-current, irreversibility, and K-sweep readiness artefacts. The gate keeps thermodynamic peak claims blocked until the DLA-sector formalism, classical Lindblad/QuTiP reference, raw-count execution, and falsification controls pass.
• S10 Analog-native Kuramoto backends readiness. Implemented 2026-05-20: scpn-bench s10-analog-native-readiness exports the no-submit primitive-accounting and provider-readiness artefacts for neutral-atom, Bloqade, and IBM Pulse export targets. The gate keeps analog-advantage and provider-execution claims blocked until SDK construction, calibrated unit constraints, matched-tolerance digital baselines, and raw execution records pass.
• S11 DLA-driven quantum sensing via sync order parameter readiness. Implemented 2026-05-20: scpn-bench s11-quantum-sensing-readiness exports the no-submit QFI/classical-Fisher proxy scan and public sensing note. The gate keeps sensing-advantage claims blocked until the preregistered perturbation benchmark, classical Fisher baseline, hardware shot budget, and raw-count uncertainty archive pass.
• S12--S53 Scientific, foundational, and applied post-v1.0 differentiation tracks listed in the strategic roadmap.

Recently closed

• S5 Open-data + classical validation for the DLA-parity dataset — closed 2026-04-18. New subpackage scpn_quantum_control.dla_parity: schema, JSON loader + SHA-256 integrity check, statistical reproducer against the published summary, classical (numpy + qutip) leakage reference, one-call run_full_harness, CLI at scripts/run_dla_parity_suite.py, docs page, [dla-parity] extra, CI smoke step. 100 % line coverage across all modules; 128 / 128 tests green.
• B1 arXiv LaTeX preprint — closed 2026-04-18 (commit 457b734): paper/submissions/submission_002_phase1_dla_parity/phase1_dla_parity.tex compiles to a 4-page PDF; paper/README.md documents the submission packaging.
• Qiskit 1.x → 2.x migration — closed 2026-04-18 (commit b786fc2): pin bump to qiskit>=2.2,<3.0, closed the Dependabot PR #44 technical debt, misleading PauliEvolutionGate comment corrected. No source changes needed (our call sites were already 2.x-compatible). Full test suite 5045 / 0 / 95.

Completed

v0.1.0–v0.8.0 (February 2026)

• Core modules: qsnn, phase, control, bridge, qec, hardware, crypto, mitigation
• 20 hardware experiments, ZNE + DD error mitigation, Heron r2 noise model
• 5 Jupyter notebooks, 10 examples, GitHub Pages docs
• Property-based tests (hypothesis), integration + regression suites
• Identity subpackage: VQE attractor, coherence budget, entanglement witness, fingerprint
• 483 tests, 99%+ coverage

v0.9.0 (March 2026)

• 100% line coverage, security scanning (bandit + pip-audit)
• CODEOWNERS, SPDX headers on 130 files, AGPL-3.0 dual-license
• Enterprise hardening: 7 CI workflows, Dockerfile, Makefile, GOVERNANCE, SUPPORT
• Identity subpackage: 4 modules, 43 tests
• 553 tests

v0.9.0+9modules (March 2026)

9 v1.0 modules implemented:

Module	Description
mitigation/pec.py	Probabilistic Error Cancellation (Temme et al. PRL 119 180509)
hardware/trapped_ion.py	Trapped-ion representative noise model (MS gate, T1/T2) + explicit CX-proxy transpilation metadata
control/q_disruption_iter.py	ITER 11-feature disruption classifier + synthetic data
benchmarks/quantum_advantage.py	Classical vs quantum scaling + crossover extrapolation
bridge/snn_adapter.py	SNN ↔ quantum bridge + ArcaneNeuronBridge (sc-neurocore)
bridge/ssgf_adapter.py	SSGF ↔ quantum bridge + SSGFQuantumLoop
identity/binding_spec.py	6-layer 18-oscillator identity topology + orchestrator mapping
qsnn/training.py	Parameter-shift gradient training for QuantumDenseLayer
qec/fault_tolerant.py	Repetition-code logical qubits + transversal RZZ

Cross-repo integrations wired:

• sc-neurocore: ArcaneNeuron spike collection → quantum forward → current feedback
• SSGF engine: W/theta read → Trotter evolve → theta writeback (quantum-in-the-loop)
• scpn-phase-orchestrator: 18↔35 oscillator phase mapping (identity_coherence domainpack)
• scpn-fusion-core: NPZ archive shot data → ITER 11-feature vector

679 tests, 100% coverage, all 6 preflight gates passing.

v0.9.1 (March 2026)

• 15-dimension codebase audit: 5 critical, 12 high, 6 medium findings fixed (39 files)
• Removed hardcoded IBM CRN, fabricated CVE, broken Dockerfile
• SPDX headers on all files, line-6 descriptors on all __init__.py
• CI tool pins: ruff 0.15.6, mypy 1.19.1, bandit 1.9.4
• knm_to_hamiltonian dedup, Makefile/pre-commit preflight fix
• 1789 tests, 100% coverage

v0.9.2 (March 2026)

• Coverage expanded to include runner.py and experiments.py (previously omitted)
• 38 new runner tests covering all simulator-path methods
• 22 new experiment tests covering all 20 experiment functions
• Rust engine (scpn_quantum_engine) rebuilt, parity tests green
• Stale README cross-refs updated (phase-orchestrator v0.5.0, test count, hardware status)
• Failed/cancelled CI runs cleaned
• 1932+ tests

v0.9.3 (March 2026)

• Rust engine expanded 11→15 functions: lanczos_b_coefficients, otoc_from_eigendecomp, build_xy_hamiltonian_dense, all_xy_expectations
• Measured benchmarks: 264× OTOC (n=4), 27× Lanczos (n=3) vs Python; Rust XY-Hamiltonian construction 96.5× (L=4) / 33.7× (L=8) against a warmed Qiskit baseline [amended 2026-07-16: the original "5401×" Hamiltonian figure was a cold-start artefact and is retracted — see docs/native_speedup_benchmark.md; the baseline is a local regression guard, not a published performance claim]
• 8 modules migrated to Rust Hamiltonian path (knm_to_dense_matrix), zero .to_matrix() callers
• IBM hardware evidence ledgered for ibm_fez (Heron r2)
  • legacy artifact-backed Bell, QKD, VQE, ZNE, Trotter, and UPDE rows retained
  • no broad-advantage or frontier claim promoted from the baseline campaign
• 14 publication figures (simulation + hardware)
• JAX GPU backend (jax_accel.py) — vectorised coupling scans
• PyPI Rust wheel CI for 5 platforms (rust-wheels.yml)
• Kaggle registered, notebook pushed, ORCID profile filled
• New docs: rust_engine.md with benchmark tables, API updates across 4 doc pages

v0.9.4 (March 2026)

• 81 new tests (PennyLane mock, JAX mock, ripser mock, hardware runner mock, fallbacks)
• OpenSSF Best Practices badge (100% passing)
• 3 benchmark API docs: gpu_baseline, mps_baseline, appqsim_protocol
• Coverage 95%→98%, 2715 tests
• Experiment roadmap + crypto branch updated for completed March hardware campaign

v0.9.5 (March–April 2026)

• 10X Architecture: Strange Loop co-evolution engine (DynamicCouplingEngine, TopologicalCouplingOptimizer)
• BiologicalSurfaceCode: native topological QEC on SCPN 16-layer graph
• LindbladSyncEngine: MCWF trajectory path for large-N open systems
• StructuredAnsatz: topology-informed variational circuits for arbitrary coupling graphs
• EEG Classification: PLV-to-quantum pipeline for brain state analysis
• 27 FIM notebooks (NB14–47): 19 discoveries, 6 honest negative results
• IBM hardware v2: aggregate-only ibm_fez artifacts retained but unpromoted until raw counts, private retrieval map, and reproduction analysis are reviewed
• Rust engine expanded: 15→22 functions (correlation_matrix_xy, lindblad_jump_ops_coo, lindblad_anti_hermitian_diag, parity_filter_mask)
• Documentation audit: 21 discrepancies fixed, 26 analysis + 10 phase + 3 bridge exports added
• Fixed backend_dispatch jax.numpy AttributeError
• 165 Python modules, 22 Rust functions, 47 notebooks, 21 examples
• 2813+ tests, 95% coverage

v1.0.0 (Target: Q3 2026)

Remaining items:

• IBM Heron r2 hardware evidence ledgered; promoted claims narrowed to raw-count-backed or artifact-named rows
• Coverage push to 100% (tracked in the internal coverage queue; latest documented baseline is ~97.6%, 315 uncovered lines)
• arXiv preprint: "Quantum simulation of coupled-oscillator synchronization on a 156-qubit superconducting processor"
• Quantum advantage figure: exact-simulation crossover and hardware-budget boundary; no broad hardware-advantage claim until a preregistered raw-count campaign passes the validation gate
• IBM Quantum Credits campaign (applied 2026-03-29, pending review)
• Version bump to 1.0.0

Future

• Phase 4 / follow-up validation and reproducibility
  • One-command reproducibility CLI — implemented 2026-05-06. scpn-bench reproduce-methods, scpn-bench fim-all, and scpn-bench all now run committed benchmark harness groups, regenerate JSON/CSV artefacts, and report drift against committed files without submitting IBM jobs.
  • Public benchmark dashboard — implemented 2026-05-06. docs/methods_benchmark_dashboard.md is wired into MkDocs and links benchmark artefacts, generator scripts, machine provenance, reproducibility commands, optional GPU/scaling/readout harnesses, and no-QPU-spend boundaries.
  • Ansatz scaling plus tensor-network baseline — initial harness implemented 2026-05-06; next strengthening task. scripts/benchmark_ansatz_scaling_tn.py generates n=4--12 ansatz scaling rows and tensor-network truncation diagnostics. Sparse eigensolver references now strengthen larger-n rows where feasible. Remaining work: extend beyond ground-state truncation diagnostics into richer MPS/VQE reference comparisons, then update methods-paper claims only from regenerated artefacts.
  • Native or analogue FIM implementation — initial compiler path implemented 2026-05-06. The analogue Kuramoto compiler accepts lambda_fim and decomposes -lambda M^2/n into the documented all-to-all Z_i Z_j payload for backend design studies. Remaining work: wire a real provider-specific Pulser/Bloqade/pulse-level backend before making any execution claim.
  • Adaptive lambda feedback loop — scaffold implemented 2026-05-06. AdaptiveFIMConfig, FIMWitness, propose_next_lambda, and adaptive_lambda_schedule provide a deterministic controller over measured or simulated witnesses. Remaining work: design a separately approved hardware protocol before any adaptive-QPU claim.
  • Scalable readout-mitigation cross-check — FIM repeated dataset implemented 2026-05-06. Full 16-state readout-matrix inversion is implemented for the repeated SCPN/FIM follow-up, where the required calibration basis exists. Remaining work: add dataset-level eligibility markers for any n<=8 campaign that lacks complete basis calibration, before spending QPU time on new calibration circuits.
• Fault-tolerant UPDE on surface code logical qubits (post-2030, hardware-dependent)
• QSNN training loop on real hardware (parameter-shift STDP)
• Quantum disruption classifier on ITER disruption database
• Trapped-ion hardware runs (IonQ Aria / Quantinuum H2)
• SSGF quantum-in-the-loop with live SSGFEngine on GPU

Strategic Differentiation (post-v1.0)

Fifty-three post-v1.0 differentiation tracks are scoped in docs/strategic_roadmap.md. All are DEFERRED / CEO-gated; no execution until individually activated. Quarterly review cadence.

Infrastructure + ML (S1–S7):

• S1 Hybrid classical–quantum feedback loop
• S2 Quantum advantage benchmarks at scale (N = 4 → 20+)
• S3 ML-augmented pulse / ansatz design
• S4 Multi-hardware backend + pulse-level control
• S5 Open-data + classical validation harness
• S6 Decoupled quantum-kuramoto subpackage
• S7 Fault-tolerant / logical-level extension roadmap

Scientific directions — batch 1 (S8–S14):

• S8 Mid-circuit adaptive branching (Dynamic Circuits follow-up to S1)
• S9 Quantum thermodynamics of synchronisation transitions
• S10 Analog-native Kuramoto backends (Rydberg / neutral-atom / CV photonic)
• S11 DLA-driven quantum sensing via sync order parameter
• S12 Automated phase-diagram exploration via Bayesian optimisation
• S13 Bosonic / continuous-variable quantum Kuramoto
• S14 Hybrid quantum-classical forecasting engine

Scientific directions — batch 2 (S15–S21):

• S15 DLA-protected many-body scars for long-lived synchronisation
• S16 Quantum network tomography (reconstruct K_nm from observables)
• S17 Higher-order (simplicial / hypergraph) quantum Kuramoto
• S18 Synchronisation-protected quantum memories and repeaters
• S19 Entanglement phase diagram + magic + Krylov complexity
• S20 Quantum Kuramoto universal control-benchmark suite
• S21 Multi-scale quantum → classical bridging layer

Scientific directions — batch 3 (S22–S28):

• S22 Non-Hermitian / PT-symmetric Kuramoto with exceptional points
• S23 Quantum reservoir computing on Kuramoto transients
• S24 Quantum speed limits for collective synchronisation
• S25 Topological defects + vortex dynamics on 2D quantum lattices
• S26 Entanglement-mediated long-range synchronisation
• S27 Hardware-in-the-loop inverse design of oscillator networks
• S28 Synchronisation-enhanced distributed quantum metrology

Scientific directions — batch 4 (S29–S35):

• S29 Floquet Kuramoto time crystals + subharmonic sync
• S30 Quantum Kuramoto for community detection and modularity
• S31 DLA-protected many-body localisation / delocalisation
• S32 Monitored quantum Kuramoto (measurement-induced transitions)
• S33 Quantum-enhanced Lyapunov spectra for chaotic Kuramoto
• S34 Self-organising Kuramoto (autonomous drive engineering)
• S35 Quantum Kuramoto native simulator for active matter

Foundational tracks (S36–S53) — geometric, categorical, field-theoretic, foundations:

• S36 Information geometry on quantum sync manifolds
• S37 Categorical / compositional quantum Kuramoto
• S38 Quantum Kuramoto field theory continuum limit + RG flows
• S39 Autopoietic / self-referential networks
• S40 Holographic duals via quantum synchronisation
• S41 Quantum causal discovery with intervention
• S42 Symplectic structure-preserving Trotterisation
• S43 Full resource theory of quantum synchronisation
• S44 Objective collapse / macroscopic foundations testbed
• S45 Biologically faithful Kuramoto simulator + IIT
• S46 Phase-transition / attractor-landscape quantum programming
• S47 Analogue gravity (relativistic, cosmological, baryogenesis, emergent spacetime)
• S48 Self-healing qubit fabrics + continuous QEC via sync
• S49 Quantum fluctuation theorems across sync transitions
• S50 Quantum kernels from sync manifolds (ML)
• S51 Hayden–Preskill / black-hole information dynamics simulator
• S52 Distributed quantum consensus via global sync (quantum internet)
• S53 Engineered self-organised criticality

Applied verticals (cross-cutting over S1–S53, no separate physics tracks): fusion plasma stabilisation; tipping-point early-warning; IIT consciousness testbed; quantum biology engineering; quantum internet infrastructure; autonomous AI physicist (discovery engine). Each applied vertical is an activation target for one or more physics tracks listed above.

Paper 0 downstream replay extraction

• Status: intentionally extracted from this repository in 97f9e910.
• Scope boundary: the former Paper 0 source-accounting register, generated replay scripts, replay data, documentation pages, and package release blockers moved out of the public package surface.
• Current in-repository boundary: measured-system K_nm validation remains blocked until calibrated coupling magnitudes with per-edge uncertainty are available, and the active package gate is the K_nm measured-candidate release gate. Historical changelog entries for the removed replay package remain historical records only.