Public Evidence Status¶
This page is the truth source for what WolfXL currently claims publicly.
Status¶
- WolfXL package version:
2.0.0 - Repository under audit:
SynthGL/wolfxl - Last local verification pass: 2026-07-30 current-source benchmark and claim-audit refresh
- Dated release-artifact benchmark snapshot:
docs/trust/release-artifact-benchmark-rerun.md - Current supported-scope SOTA gate: green
- Overall all-future-surface SOTA claim: not ready
- Benchmark posture: the dated 2026-07-30 clean-source OpenPyXL and selected-surface Rust/Rust-backed benchmarks passed their configured gates. Benchmark source relevance: current. No weak speed or memory rows under the configured gates. Memory pass/fail applies the configured absolute RSS noise tolerance.
- PyPI Rust-backed discovery fetched: true; 0 unclassified PyPI Rust-backed discovery hits.
- Public-claim wording audit:
docs/trust/public-claim-wording-audit.json - Public-claim dimensions audit:
docs/trust/public-claim-dimensions-audit.md - Release-artifact smoke:
docs/trust/release-artifact-smoke.md - Release-artifact benchmark smoke:
docs/trust/release-artifact-benchmark-smoke.md - Full release-artifact benchmark rerun:
docs/trust/release-artifact-benchmark-rerun.md - Final blocker audit:
docs/trust/final-sota-blockers.md - SOTA snapshot wording audit:
docs/trust/sota-snapshot-wording-audit.md - Cheap trust-report freshness audit:
docs/trust/trust-report-freshness-audit.md - Launch-safe claim brief:
docs/trust/launch-claim-brief.md
Broad Claim Boundaries¶
These are the five reasons the universal "no reason to keep using openpyxl" claim remains gated. The supported-scope gate is green for the current measured clean-source lanes; the broader claim remains bounded by open-ended evidence:
| Boundary | Still-unproven families | Plain-English meaning |
|---|---|---|
| Registered release-artifact lanes are proven, but not every future package route or installer. | unregistered_future_wheel_lanes, installer_and_resolver_contexts, alternate_distribution_channels, workflow_topology_and_trigger_paths |
The current wheel and sdist lanes are covered; future wheel lanes, installers, dependency resolvers, distribution channels, and CI or release trigger paths still need their own proof before public copy relies on them. |
| Openpyxl still has ecosystem maturity and long-tail workflow history. | unsupported_openpyxl_api, unvalidated_business_template, olap_external_cache_or_pivot_visuals, adjacent_spreadsheet_surfaces, organizational_dependency_familiarity |
Some teams may reasonably keep openpyxl for an unsupported API, an unvalidated template, OLAP or external pivot caches, adjacent spreadsheet formats, staff familiarity, old integrations, or dependency approval history. |
| High-risk render variant space remains open-ended. | unseen_feature_edit_combinations, template_specific_visual_acceptance, excel_renderer_version_variants |
The current render evidence is strong for the tested space, but it cannot cover every feature-edit combination, every template-specific visual expectation, or every Excel renderer version. |
| Click-level Excel interaction variant space remains open-ended. | slicer_timeline_control_variants, prompt_and_dialog_variants, destructive_axis_external_tool_boundaries |
The current interaction probes are strong for the tested space, but they cannot cover every slicer, timeline, control, prompt, dialog, or destructive external-tool path. |
| Future or unseen real-world Excel surfaces cannot be fully exhausted. | unseen_ooxml_part_families, unseen_relationship_or_content_types, future_excel_feature_extensions |
No finite workbook corpus can prove that every future OOXML part, relationship, content type, or future Excel extension is already handled. |
Verified Now¶
These points may be repeated publicly with the status and evidence boundaries shown.
| Claim | Status | Evidence |
|---|---|---|
| WolfXL ships read, write, and modify modes behind an openpyxl-style API. | Publishable | README.md, docs/getting-started/quickstart.md |
WolfXL 2.0 adds pivot cache construction, pivot table construction, pivot-chart linkage, and pivot-bearing copy_worksheet. |
Publishable | docs/release-notes-2.0.md |
| WolfXL preserves untouched workbook parts during modify-mode saves, including charts, images, macros, and pivot-bearing files. | Publishable with limitations | README.md, docs/trust/limitations.md, local test suite |
| The dated 2026-07-30 clean-source OpenPyXL benchmark passed its configured bounded speed and memory gates. Benchmark source relevance: current. | Current measured-scope evidence | docs/internal/claims/baselines/2026-07-30-head-cf0ee75b-current-sota-claim-audit.md |
| The dated 2026-07-30 required Rust/Rust-backed competitor snapshot is complete for its recorded benchmark set and passed the selected-surface speed and memory gates. Raw cross-surface rows remain disclosed and outside the selected gate. Benchmark source relevance: current. | Current measured-scope evidence | rust_xlsxwriter 0.96.0, crates.io package xlsxwriter 0.6.1 with repo alias xlsxwriter-rs (not the separate crates.io xlsxwriter-rs package at 0.1.0), calamine 0.36.1, umya-spreadsheet 3.0.1, and office_oxide 0.1.8 are benchmarked through direct Rust APIs; fastexcel 0.20.2 and python-calamine 0.8.2 are benchmarked through their Python binding APIs; requested-name resolution and current registry checks are recorded in docs/performance/baselines/2026-07-28-rust-competitor-set-live-recheck.json |
| The older local release-artifact install smoke passed for its recorded macOS arm64 lane, but it is not current enough for headline proof after later source changes. | Promising but not fully proven | docs/trust/release-artifact-smoke.md; currentness boundary in docs/trust/public-claim-dimensions-audit.md |
| The older bounded release-artifact benchmark smoke passed for its measured smoke lanes, but it is stale after later source changes. | Promising but not fully proven | docs/trust/release-artifact-benchmark-smoke.md; currentness boundary in docs/trust/public-claim-dimensions-audit.md |
| The dated 2026-07-30 macOS arm64 CPython 3.13 full clean-wheel benchmark rerun passed its configured bounded OpenPyXL and selected-surface Rust speed and memory gates. Raw Rust public-API speed rows remain visible as non-gating cross-surface boundary evidence. Benchmark source relevance: current. | Current measured-scope evidence | docs/trust/release-artifact-benchmark-rerun.md |
| No Excel-only proof is deferred for this release-artifact blocker. | Publishable as a boundary, not as a universal SOTA claim | docs/trust/final-sota-blockers.md |
| Older generated SOTA snapshots are explicitly marked historical so stale blocker wording is not mistaken for the current claim boundary. | Publishable as a trust/audit hygiene note | docs/trust/sota-snapshot-wording-audit.md |
| Cheap generated trust reports regenerate to the checked-in JSON and Markdown files. | Publishable as trust/audit hygiene, not as a benchmark rerun | docs/trust/trust-report-freshness-audit.md |
| Launch copy has a supported-scope default claim and explicit avoid-list. | Publishable as wording guidance | docs/trust/launch-claim-brief.md |
| ExcelBench contains reproducible fidelity and performance harnesses, generated artifacts, and methodology docs. | Publishable | https://github.com/SynthGL/ExcelBench, ExcelBench/METHODOLOGY.md |
The dated 2026-04-28 WolfXL 2.0 wheel-backed ExcelBench rerun reaches 18/18 green features and 100% pass rate. |
Publishable with timestamp | ExcelBench/results-release-2026-04-28/README.md |
Gated Claims¶
These points need measured threshold improvements or clearer caveats before they should appear as headline launch copy.
| Claim Type | Current State | Action Before Headline Use |
|---|---|---|
Universal WolfXL 2.0 speedup headline vs openpyxl |
Supported only for measured lanes | Cite the current SOTA audit and exact workload rows instead of one universal number |
| "Only library" or "first library" ecosystem claims | Supported only inside the listed comparison scope | Keep qualified wording and link the compatibility matrix plus Rust competitor-set audit |
| "No reason to keep using openpyxl" | Not ready | Keep the stronger all-future-surface claim gated until the strict SOTA audit reports sota_claim_ready: true |
| "Better in every dimension" | Not ready | Keep the dimensions report green for all rows before using this kind of headline |
Public benchmark tables in docs/performance/benchmark-results.md |
Historical v1.7 baseline | Treat as archived context, not current release proof |
| Older ExcelBench dashboard snapshots | Accurate for their timestamp, not current release evidence | Label as archived snapshots in README links and launch copy |
| Next release quality claims | Thirteen of 21 registered release-artifact lanes are current: Linux aarch64 and Linux x86_64 CPython 3.10 through 3.13, macOS arm64 CPython 3.10 through 3.13, and the source distribution. The remaining eight macOS x86_64 and Windows x86_64 lanes are not current, and raw cross-surface weak rows plus broader future package and install routes remain outside universal proof. | Refresh the macOS x86_64 and Windows x86_64 lanes from the current source before making a full cross-platform release-quality claim |
Release Evidence Checklist¶
Use this set before any post, PyPI release thread, HN launch, or comparison blog post.
cargo test --workspaceuv run --no-sync pytest -quv run pytest tests/parity -q -x- Fresh wheel/sdist build and clean install smoke, or cite the current dated release-artifact smoke reports when no new artifact is being published
- Dated ExcelBench fidelity/perf evidence tied to the exact snapshot being cited, with older dashboard links marked as archived context
- Current release-artifact benchmark rerun, final blocker audit, public-claim
dimensions audit, and trust-report freshness audit under
docs/trust/ - Update public timestamps in both READMEs and any dashboard summary pages
Public Wording Policy¶
- Safe: "Rust-backed, openpyxl-compatible Excel automation with surgical modify mode."
- Safe: "Pivot tables are constructible from Python in WolfXL 2.0."
- Safe with caveat: "The supported-scope SOTA gate is green for the current measured clean-source lanes; the broader all-future-surface claim remains gated."
- Safe with caveat: "The broader all-future-surface claim remains gated because future package/install routes and open-ended proof boundaries are not universal proof."
- Safe with timestamp: "On the 2026-04-29 wheel-backed ExcelBench release snapshot..."
- Default launch wording: use
docs/trust/launch-claim-brief.mdbefore drafting README, post, or comparison copy. - Avoid: broad
10x-100xrelease-headline speed claims that ignore workload shape. - Avoid until the strict audit is fully green: "there is no reason to keep using openpyxl."
- Avoid: presenting historical v1.7 benchmark tables as WolfXL 2.0 release proof.
Cross-Repo Alignment Notes¶
- WolfXL is on
2.0.0inCargo.tomland release docs. - ExcelBench contains a dated 2026-04-28 WolfXL 2.0 wheel-backed release snapshot in
results-release-2026-04-28/alongside older historical snapshots. - When in doubt, prefer explicit dates, exact commands, and raw artifact links over compressed marketing summaries.