Skip to content

Changelog

Why this page exists

This file is the release history anchor for API, evidence, and policy changes. Use it to verify when a claim boundary, verification path, or gating rule changed before that version was promoted.

Dated list of changes. Format follows Keep a Changelog.

Full detailed changelog: CHANGELOG.md.

Unreleased

[1.1.0] - 2026-07-29

Added

  • Complete deterministic documentation catalogs for all Python modules and public module-level symbols, notebooks, and public Markdown pages, plus an evidence-bounded solutions and market-value guide.
  • BL-54 finite parity-sector and fixed-active-set topology JVP/VJP boundaries, projected synthetic optimisation, and complete API/evidence documentation.
  • BL-88 topology-aware exact-statevector fidelity-kernel product with matched graph controls, custody-checked classification, and bounded synthetic representability evidence.
  • BL-63 power-grid, Josephson, EEG-like, and ITER-inspired application honesty kits with privacy/provenance metadata and explicit domain non-claims.
  • BL-98 theory-hook promotion registry with corrected QSL, Hamiltonian- learning, Koopman, QMI, magic, SFF, and IIT-proxy boundaries.
  • BL-102 RL research governance: disabled-by-default preregistration, fixed multi-seed suites, iteration/evaluation budgets, deterministic zero-noise replay, explicit dense-reward gaming boundaries, BL-53 no-preregistration refusal, and current unimplemented/BL-58 pulse blockers.
  • BL-84 initial deep-analysis research-lane registry: 73 immutable human-reviewed analysis/ and gauge/ rows, explicit maturity/differentiable-hook/claim classifications, deterministic evidence, and an exact inventory gate that rejects unregistered new modules. Registry membership grants no product, control, hardware, or publication claim.

Fixed

  • The optional jax extra now includes FlatBuffers so registered Phase-QNode AOT export serialization and replay work after installing the documented extra.

[1.0.0] - 2026-07-17

First stable release. The version carriers move to 1.0.0 and the public API is placed under an explicit stable-contract commitment.

Security

  • Torch graph-artifact deserialisation is digest-gated (KIMI-12): the AOTAutograd and dynamic-shape export replays reach torch.load(..., weights_only=False) / torch.export.load only after the on-disk artefact re-hashes to the SHA-256 recorded at save time, and the serialisation-surface audit flags every ungated weights_only=False.

Added

  • Verifiability and evidence: March job-id commitments with IBM retrieval receipts, committed tensor-network classical baselines for the flagship workloads, and the preregistered matched-basis QBER re-run (KIMI-7/8/11).

Changed

  • Public claims are evidence-governed with dated amendments; published-record sources are frozen behind a SHA-256 guard and corrections land on live surfaces or new Zenodo versions, never in place. The first real FIM notebook-batch execution fixed the kernel-cwd and external-dataset exclusions in Notebook CI.

[0.10.0] - 2026-06-26

Added

  • Released the v0.10 public surface set: QRNG streaming and FIPS/NIST health reports, ML-DSA-65/PQC trigger signing, UltraScale+ HLS pulse-source generation, NV-centre 20 T magnetometry simulation, FRC pulsed-shot QAOA scheduling, realtime loop telemetry, Studio federation manifests, and Kuramoto variant/acceleration routes.
  • Added bounded native whole-program AD factorisation support for full-output static inverse, vector-solve, and matrix-RHS solve traces through 7x7, while keeping 8x8+, more than four RHS columns, and shape-changing linalg fail-closed.
  • Refreshed the API selector, tutorial routes, example gallery, new-module index, README, site home, reproducibility metadata, citation metadata, release-readiness scope, and generated capability manifests for the 0.10.0 package line.

Changed

  • Removed obsolete Paper 0 artifact requirements from the package release readiness gate after Paper 0 was intentionally extracted from this repository into the GOTM root register.
  • Hardened MLIR compiler record validation so malformed transform-plan rows and non-integer kernel-verification sample counts fail closed with typed ValueError diagnostics.
  • Hardened compiler alias-activity evidence records so dynamic public inputs with whitespace identifiers, mutable case containers, boolean count drift, or empty focused-test evidence fail closed before promotion-batch assembly.
  • Kept hardware, RNG, cryptographic, HLS, realtime, and benchmark wording bound to existing evidence classes rather than promoting new hardware or production cryptographic claims.

[0.9.12] - 2026-06-15

Added

  • Public release metadata and documentation alignment for the current differentiable-programming hardening queue, including README, site home, onboarding, tutorial, notebook, API, reproducibility, hardware-ledger, release-readiness, citation, Zenodo, and generated capability inventory surfaces.
  • Adoption-oriented documentation routes explaining the software purpose, application lanes, commercial value, first user paths, notebook governance, API selection, evidence classes, and release-hygiene boundary without promoting unsupported hardware, clinical, or broad advantage claims.

Security

  • Updated the pinned CI pip package from 26.1.1 to 26.1.2 across the Python 3.10-3.13 lockfiles to remediate PYSEC-2026-196 / CVE-2026-8643.

Fixed

  • Hardened the internal TODO policy guard for exported source trees, including Docker images where .git is intentionally excluded, while preserving filesystem detection of competing TODO queues.

[0.9.11] - 2026-06-05

Added

  • Public documentation and release metadata alignment for the current differentiable-programming hardening state, including onboarding, tutorials, notebooks, API routes, reproducibility metadata, release-readiness scope, and generated capability inventory.
  • Native compiler-backed whole-program AD determinant support widened to verified static dense 6x6 through 19x19, with 20x20+ traces documented as fail-closed until a stronger determinant-partial helper is verified.

[0.9.10] - 2026-06-04

Added

  • Public documentation and release metadata polish for the current differentiable-programming implementation state, including onboarding, README positioning, tutorial and notebook routing, API boundary text, gradient evidence checklists, and the generated capability inventory.
  • Explicit enterprise-readiness boundary for supported parameter-shift, composed-objective, provider-gradient, transform-nesting, and compiler/program-AD routes, with unsupported arbitrary-program, hardware, and framework-native paths kept fail-closed.

[0.9.9] - 2026-06-02

Added

  • Public differentiable-programming documentation route covering quantum gradients, current compiler/program AD boundaries, public differentiable API entry points, notebooks, tutorials, planned framework adapters, benchmark evidence, and fail-closed unsupported scenarios.

[0.9.8] - 2026-06-01

Added

  • Public onboarding route covering project purpose, user routes, application lanes, commercial route, evidence boundaries, and release-documentation scope.
  • Refreshed first-path docs across README, documentation index, quickstart, tutorials, notebook guide, API overview, examples guide, installation guide, MkDocs navigation, and release-readiness page.

[0.9.7] - 2026-05-19

Paper 0 preregistered replay + capability-manifest system.

  • GOTM-SCPN Paper 0 K_nm preregistered replay artefact set: deterministic JSON/Markdown outputs, comparator gate, release-safe scpn-bench wiring, null-model diagnostics, input digest manifests, and an explicit do-not-promote hardware boundary
  • Paper 0 methodology-paper outline, first downstream preregistered experiment design, replay contract, and measured-coupling evidence checklist
  • Repository capability-manifest system (generated JSON/Markdown inventory + README snapshot) with scpn-bench capability-manifest-check, so public capability counts are generated from one source of truth
  • Release-readiness audit helper and public gate document composing version consistency, coverage-gap readiness, behavioural-test density, and required claim-boundary artefacts into one deterministic tag blocker

[0.9.6] - 2026-04-17

Changed (2026-04-29)

  • Added documented classical baseline surfaces for SciPy ODE, optional QuTiP Lindblad, and optional MPS TEBD runs.
  • Re-routed API documentation so stable facades are the first-path entry point and low-level module references sit under advanced navigation.
  • Added a mkdocstrings stable facades API page for first-path public facades.
  • Added a physics-first Kuramoto-XY tutorial with a tested arbitrary K_nm/omega workflow.
  • Refreshed README, docs index, and results gallery from the dated hardware status ledger.
  • Documented the licence boundary for a possible future lightweight Kuramoto-XY core package; no relicensing is implied.
  • Added a kuramoto_core facade for validated K_nm/omega problems, Hamiltonian/circuit compilation, dense Rust-backed Hamiltonians, and order-parameter measurement.
  • Added tools/check_dependency_drift.py so release checks can verify that requirements.txt still mirrors pyproject.toml.
  • QuantumKuramotoSolver validates coupling shape, finite values, symmetry, Trotter settings, and run-grid parameters before building circuits.
  • Added a public hardware status ledger that classifies theory, simulator, hardware, mitigated, and noise-limited claims and links visible docs to that evidence index.
  • requirements.txt mirrors the canonical runtime bounds from pyproject.toml; [all] is portable; CUDA/JAX packages live behind [accelerated]; CI includes a fresh editable [all] install smoke job.

Security

  • Bumped rand 0.9.2 → 0.9.4 (RUSTSEC-2026-0097).
  • Removed IBM Cloud CRN prefix log in scripts/retrieve_ibm_job.py.
  • Bumped pytest 9.0.2 → 9.0.3 (CVE-2025-71176).

Added

  • SCPNConfig via pydantic-settings ([config] extra) — unified runtime config.
  • structlog bootstrap ([logging] extra) — configure_logging + get_logger.
  • AsyncHardwareRunner — concurrent IBM submission over asyncio.
  • Plugin backends registry via entry_points group scpn_quantum_control.backends.
  • accel/ package — multi-language dispatcher with Rust → Julia → Python chain. [julia] extra adds juliacall-bridged order_parameter.
  • provenance.capture_provenance() — git hash + runtime + hostname on every result JSON.
  • Cross-validation test suite vs QuTiP + JAX/Dynamiqs ([xvalidate] extra).
  • Criterion Rust benches + GitHub workflow.
  • Issue triage policy (docs/triage.md), language policy (docs/language_policy.md).
  • docs/pipeline_performance.md §"Multi-language accel chain" with measured wall-times.

Changed

  • CI dev-tool matrix (pytest 9.0.3, mypy 1.20.1, ruff 0.15.10, hypothesis 6.151.13, build 1.4.3).
  • _order_param in hardware/classical.py now dispatches through the accel chain.
  • hardware/runner.py print() calls migrated to structured events.
  • HardwareRunner.DEFAULT_INSTANCE → _default_instance() reads SCPNConfig.

Repository hygiene

  • Public tracked files stripped of agent-name mentions; internal logs keep identity.
  • CHANGELOG.md terse-rewritten (dates + changes only, no attributions).
  • docs/contributing.md moved to private local workspace/ (gitignored).

[0.9.5] - 2026-03-29 / 2026-04-11

Phase 1 IBM raw-count observation of the DLA parity asymmetry + five strategic tweaks (GUESS, DynQ, ICI, hypergeometric, FFI hardening) + repository hygiene (gitleaks + custom secret scanner).

Phase 1 IBM ibm_kingston campaign (2026-04-10)

  • First publishable raw-count hardware observation of the DLA parity asymmetry on IBM ibm_kingston (Heron r2, 156 qubits). 342 circuits with up to 21 reps per (depth, sector) point at \(n = 4\). Mean asymmetry \(+10.8\,\%\) for Trotter depths \(\ge 4\), peak \(+17.48\,\%\) at depth 6. Welch's t-test 7/8 depths at \(p < 0.05\), Fisher's combined \(p \ll 10^{-16}\). Consistent with the \(4.5\text{–}9.6\,\%\) apriori simulator prediction.
  • Full statistical analysis script with error bars, Welch t-test, and matplotlib figures (scripts/analyse_phase1_dla_parity.py, figures/phase1/).
  • 267-line short-paper draft for Quantum Science and Technology / Physical Review Research in paper/.

Strategic tweaks from deep-research audit (2026-04-08)

  • GUESS symmetry-decay ZNE (mitigation/symmetry_decay.py, scpn_quantum_engine/src/symmetry_decay.rs) — physics-informed zero-noise extrapolation using \(\sum Z_i\) as the guide observable. Oliva del Moral et al., arXiv:2603.13060. 20 multi-angle tests.
  • DynQ topology-agnostic qubit mapper (hardware/qubit_mapper.py, scpn_quantum_engine/src/community.rs) — Louvain community detection on calibration-weighted QPU graphs. Liu et al., arXiv:2601.19635. 17 multi-angle tests.
  • PMP / ICI pulse sequences + (α,β)-hypergeometric pulse shaping (phase/pulse_shaping.py, scpn_quantum_engine/src/pulse_shaping.rs) — Liu et al. (2023) and Ventura Meinersen et al., arXiv:2504.08031. Rust paths give 44× hypergeometric envelope speedup and 1,665× ICI three-level evolution speedup, both verified to machine precision.
  • FFI boundary hardening — every #[pyfunction] returns PyResult<T> and validates inputs via validation.rs. 16 unit tests.
  • Long-form docs ≥ 567 lines: docs/symmetry_decay_guess.md (891) and docs/dynq_qubit_mapping.md (878).

Repository hygiene (2026-04-10)

  • tools/check_secrets.py — custom vault-pattern secret scanner with Shannon-entropy filter and keyword-based password detection.
  • gitleaks v8.21.2 pre-commit hook for generic secret detection.
  • Incident report (<private-local-record>) for the prevented FTP-credentials leak that motivated the new scanners.
  • Tests collected: 2,813 → 4,828 (97%+ coverage).
  • Python modules 165 → 201, subpackages 17 → 19, Rust functions 22 → 36 across 20 source files.

Earlier v0.9.5 work (2026-03-29 — 2026-04-07)

  • Multi-Scale QEC, Free Energy Principle, Ψ-field lattice gauge theory.
  • 10X Strange Loop co-evolution engine, DynamicCouplingEngine, TopologicalCouplingOptimizer, BiologicalSurfaceCode, LindbladSyncEngine (MCWF), StructuredAnsatz.
  • 27 FIM experiment notebooks (NB14–47), 81 FIM tests, IBM hardware v2 aggregate artifacts retained as unpromoted evidence.
  • 19 scientific discoveries + 6 honest negative results.

[0.9.4] - 2026-03-29

Coverage 98%, OpenSSF badge, 2715 tests.

  • 81 new tests, PennyLane/JAX/ripser mock guards
  • OpenSSF Best Practices badge (100%)
  • 3 benchmark API docs

[0.9.3] - 2026-03-28

Rust engine 15 functions, IBM baseline artifacts, 16 figures.

  • 4 new Rust functions (11→15): lanczos, OTOC, dense Hamiltonian, batch Pauli
  • IBM baseline artifacts on ibm_fez, including CHSH and 16q UPDE rows
  • JAX GPU backend, PyPI Rust wheel CI, BKT universality

[0.9.2] - 2026-03-26

Runner + experiment test coverage.

  • 38 runner + 22 experiment coverage tests
  • requirements.txt pinned versions

[0.9.1] - 2026-03-25

33 research gems, 56 new modules, 9,772 lines.

  • 33 research modules (Rounds 1–8): witnesses, PH, OTOC, ADAPT-VQE, VarQITE, AVQDS, Floquet DTC, BKT, DLA parity
  • 14 analysis modules, 9 phase modules, 6 hardware/bridge modules
  • IBM hardware campaign: 9 jobs on ibm_fez

[0.9.0] - 2026-03-02

SCPN-native quantum control.

  • Analysis: shadow tomography, Koopman, DLA, entanglement spectrum, QFI
  • Hardware: GPU offload, circuit cutting, trapped-ion, PennyLane/Cirq
  • Identity: VQE attractor, coherence budget, entanglement witness
  • Gauge: U(1) Wilson loops, vortex detection, CFT, universality
  • 1789 tests, 100% coverage

[0.8.0] - 2026-03-02

Cross-repo bridges, crypto, QEC expansion.

  • Bridge: SSGF, SPN-to-circuit, SNN adapter, orchestrator
  • Crypto: BB84, Bell tests, topology QKD
  • QEC: fault-tolerant UPDE, surface code, error budget

[0.7.1] - 2026-03-02

Test/typing hygiene + packaging guards.

  • Meaningful assertions replace trivial assert True; 3 duplicate tests removed; 10 error-path tests added
  • mypy warn_unreachable + check_untyped_defs; per-module ignore_missing_imports; __all__ gaps closed (knm_to_ansatz, classical_exact_evolution, JobResult, SurfaceCode, MWPMDecoder)
  • wheel-check CI job; sdist excludes; build in dev extras; matplotlib capped <4.0

[0.7.0] - 2026-03-02

Packaging, RNG hygiene, CI, exports hardening.

  • Fix crypto __all__ (callable symbols, not module names)
  • PEP 561 py.typed marker
  • Seeded RNG in QuantumDenseLayer, inf → nan in PhaseVQE
  • Named constants in percolation, pip cache in CI, dependency upper bounds
  • 456 tests, 99%+ coverage

[0.6.4] - 2026-03-01

Docs/metadata hardening.

  • Fix stale test counts (411/424 → 442) in docs pages
  • Fix header generator version string (v0.5.1 → v0.6.3)
  • Complete README architecture tree (3 missing files)
  • Add scpn-phase-orchestrator to Related Repositories
  • Update SECURITY.md supported versions (0.6.x)
  • Add pip ecosystem to dependabot
  • Version bump 0.6.2 → 0.6.4

[0.6.3] - 2026-03-01

Coverage gate, mitigation API docs, notebook table.

  • Coverage gate in CI, mitigation API docs, notebook summary table, ruff fix
  • Test count: 424 → 442

[0.6.2] - 2026-03-01

Notebook fixes + Knm heatmap figure.

  • Notebooks 01/03/04: classical_kuramoto_ode → classical_kuramoto_reference
  • Notebook 03: ZNE scales [1,2,3,4,5] → [1,3,5,7,9] (odd required by gate_fold_circuit)
  • Notebook 04: rewrite to 8-qubit Trotter + 16-layer classical (16-qubit statevector intractable on laptop)
  • figures/generate_knm_heatmap.py + figures/knm_heatmap.png (16×16 K_nm coupling matrix)
  • Knm heatmap figure in README with annotated calibration anchors
  • All 4 notebooks executed with embedded outputs
  • Remove misplaced docs/SESSION_LOG and docs/HANDOVER (duplicates of private local workspace/)

[0.6.1] - 2026-03-01

mypy + Zenodo metadata fixes.

  • mypy errors in bridge module: remove FloatArray type alias (incompatible with Python 3.9), fix Path(None) in control_plasma_knm.py
  • Zenodo metadata enriched (.zenodo.json, CITATION.cff)

[0.6.0] - 2026-03-01

Hardening + high-level API.

  • Input validation guards on all public API constructors — prevents div-by-zero in qlif, qsynapse, qstdp, qaoa_mpc, classical_kuramoto_reference; bounds-checks on bell_inequality_test, best_entanglement_path
  • PhaseVQE.solve() now returns exact_energy, energy_gap, relative_error_pct, n_params
  • Notebook 02 fixed to use enriched solve() dict
  • 13 validation tests (424 total)

[0.5.1] - 2026-03-01

Version alignment. Fixes __version__ mismatch from v0.5.0 tag timing.

[0.5.0] - 2026-03-01

Quantum cryptography hardware experiments.

  • 3 new crypto experiments: Bell test (CHSH), ZZ correlator, QKD QBER
  • devetak_winter_rate() key rate computation
  • 20 experiments in registry, 411 tests

[0.4.0] - 2026-02-28

Docs, notebooks, and test depth.

  • GitHub Pages MkDocs Material site (7 pages, auto-deploy)
  • 4 Jupyter notebooks: Kuramoto XY, VQE ground state, ZNE mitigation, UPDE-16
  • 10 new hardware experiments (noise baseline, 8-osc ZNE, Trotter order-2, etc.)
  • 14 property-based tests (hypothesis), 8 edge-case tests, 13 coverage-gap tests
  • 4 integration tests, 7 regression tests
  • Test count: 208 → 254

[0.3.0] - 2026-02-28

README rewrite and validation hardening.

  • Motivating abstract: SCPN→XY isomorphism, Kuramoto-to-Hamiltonian derivation, 4 figures
  • mypy expanded to 30 source files (8 module paths), zero errors
  • VALIDATION.md test count 88 → 199
  • Paper 27 citation, examples/README.md walkthrough

[0.2.0–0.2.7] - 2026-02-28

Error mitigation, QEC fixes, classical references.

  • ZNE (unitary folding + Richardson), DD (XY4, X2) for idle qubits
  • Heron r2 noise model factory
  • MWPM decoder: 3 bug fixes (dual edges, seam-crossing, d=5 > d=3)
  • Classical endianness fix verified against Statevector to 1e-6
  • QAOA Ising encoding fix, Petri net multi-input AND gating
  • Second-order Trotter, energy tracking
  • 20 classical reference tests, parametrized quantum-vs-classical validation
  • Test count: 88 → 208

[0.1.0] - 2026-02-28

Initial release.

  • qsnn/: Quantum LIF neuron, CRy synapse, parameter-shift STDP, dense layer
  • phase/: Kuramoto XY solver, 16-layer Trotter UPDE, VQE ground state
  • control/: QAOA-MPC, VQLS Grad-Shafranov, quantum Petri net, disruption classifier
  • bridge/: Knm→Hamiltonian compiler, SPN→circuit, bitstream→rotation
  • qec/: Toric surface code + MWPM decoder (Knm-weighted)
  • hardware/: IBM Quantum runner for ibm_fez Heron r2
  • 88 tests, 4 examples, 19 hardware result files
  • Hardware: VQE 0.05% error, 12-point decoherence curve, 16-layer UPDE snapshot