Quidra v0.5.0

04 / detailed analysis

Ecosystem

These results describe the evidence snapshot of 23 September 2026, not the ecosystem today. Workers' preserved findings were recertified into five fixed components per requirement, and the runner converted levels into points. The rubric file linked below matches the input hash recorded for each selected worker. Low absence scores refer to evidence obtained under the benchmark's fixed retrieval policy; they are not a new survey of present-day packages, tools or adoption.

Quidra’s evaluation rank
10th of 10
Evaluation score
32.50 / 100

Recorded run · 27 September 2026. These scores evaluate Quidra v0.3.0 at 38b7137. The site’s current compiler release is v0.5.0.

Run 2026-09-27-38b7137-gh31 · Methodology reference · Published score breakdown

Run’s declared inference model: claude-sonnet-5. Any separate review stages in the archived assessments are identified in the evidence notes.

Score records belong to this run; methodology links point to a later source snapshot. Its full template hash differs from the run’s recorded template hash. The explanations distinguish verified rubric inputs and score calculations from missing original observations.

15 requirements / 0–100 scores

Per-requirement comparison

Higher is better in every row. “Best other” is the highest-scoring non-Quidra language for that requirement, with every tie included. Δ is Quidra’s score minus the best other score. Published scores.

Scroll horizontally to see every column.

Requirement Quidra Best other Other score Δ
Third-party library availability / domain coverage 0.00 Python, C++, Java, TypeScript 100.00 −100.00
Package ecosystem activity / maintenance 30.00 Python, Rust, Go, Java, TypeScript, Kotlin 100.00 −70.00
Third-party tool availability 0.00 Python, C++, Rust, Go, Java, TypeScript, Kotlin 100.00 −100.00
Production adoption / deployment evidence 0.00 Python, C++, Rust, Go, Java, TypeScript, Kotlin 100.00 −100.00
Community / public knowledge availability 15.00 Python, C++, Rust, Go, Java, TypeScript 100.00 −85.00
Core tooling availability / quality 75.00 Python, Rust, Go, Java, TypeScript, Kotlin, Swift 100.00 −25.00
Package / dependency management quality 60.00 Rust, Go, Java, TypeScript, Kotlin 100.00 −40.00
IDE / editor support quality 70.00 Python, C++, Go, Java, TypeScript, Kotlin 100.00 −30.00
Debugger / profiler support 45.00 Python, C++, Go, Java, TypeScript 100.00 −55.00
Build / test integration 60.00 Python, Go, Java, TypeScript, Kotlin, Swift 100.00 −40.00
Documentation quality 65.00 Python, Rust, Go, Java, TypeScript, Kotlin 100.00 −35.00
Installation / distribution experience 60.00 Python, Rust, Go, Java, TypeScript 100.00 −40.00
Toolchain stability / release maturity 50.00 Python, C++, Go, Java, Kotlin, Swift 100.00 −50.00
Implemented platform coverage 40.00 C++ 100.00 −60.00
External integration coverage 35.00 Python, C++, Java 100.00 −65.00

01 / metric.third_party_library_availability_domain_coverage

Effective evaluation weight 10%

Third-party library availability / domain coverage

Evidence of usable independently maintained libraries across five prescribed domain groups.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Inspect the same five fixed domain groups for every language; prefer the official package index or official ecosystem directory, then independent package/project evidence. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

0.00 / 100

The recorded levels are 0 + 0 + 0 + 0 + 0 = 0 out of 20 level units. The runner's conversion gives 0 × 5 = 0.00/100.

Recorded judgment for Web network, Data numerical, ML / AI, Storage interchange, Application testing: No library (first-party or third-party) was found via targeted search for: web/HTTP framework, database/ORM, testing framework beyond built-in, serialization beyond built-in JSON, CLI framework beyond built-in, logging, date/time extension, or cryptography/security. 8 of the 10 declared domains returned no evidence of any library.

Recorded evidence
Web network Level 0/4 → 0/20 points
Data numerical Level 0/4 → 0/20 points
ML / AI Level 0/4 → 0/20 points
Storage interchange Level 0/4 → 0/20 points
Application testing Level 0/4 → 0/20 points
Evidence snapshot 2026-09-23

Evidence limits

Both discovered packages are maintained inside the same GitHub organization (quidra-lang) that owns the compiler and are described in the language's own docs as 'first-party' rather than independently authored third-party contributions, so even the 2/10 domain coverage found is not evidence of an independent third-party ecosystem. | GitHub star/fork counts were observed (0 stars, 0 forks on all three quidra-lang repositories checked) but are not used as the scoring basis by themselves; they are recorded only as corroborating context per the methodology's restriction on raw popularity proxies.

The archive also preserves an earlier 10/100 judgment. The published 0.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 4 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Web network, Data numerical, ML / AI, Storage interchange, Application testing: All 11 predeclared functional domains have a canonical, officially-recognized, maintained third-party (or first-party-adjacent, e.g. Django) library available on PyPI.

Recorded evidence
Web network Level 4/4 → 20/20 points
Data numerical Level 4/4 → 20/20 points
ML / AI Level 4/4 → 20/20 points
Storage interchange Level 4/4 → 20/20 points
Application testing Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

Download counts and GitHub star counts cited above (e.g. FastAPI GitHub stars, requests daily download counts) are used only as declared corroborating context per package, never as the sole basis for either metric's score, consistent with the predeclared-sampling rule against single-proxy scoring. | This shard evaluates only the Python language; cross-language comparability of the same domain/selection-rule structure is the responsibility of the sibling per-language shards and the runner's aggregation step. | Evidence was gathered via a single provider-brokered web-search retrieval route per package (project page + latest metadata + changelog), applied uniformly across all sampled packages to keep examination depth constant.

The archive also preserves an earlier 98/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −100.00 points

All requirements

02 / metric.package_ecosystem_activity_maintenance

Effective evaluation weight 10%

Package ecosystem activity / maintenance

Evidence of package discovery, maintenance, metadata, distribution and supply-chain support.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Use the primary package registry/ecosystem for the language when one exists and evaluate then-current observable maintenance within the fixed activity window; package count alone is never sufficient. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

30.00 / 100

The recorded levels are 0 + 1 + 2 + 2 + 1 = 6 out of 20 level units. The runner's conversion gives 6 × 5 = 30.00/100.

Recorded judgment for Registry discovery, Dependency metadata: All observed pull requests across quidra-lang/quidra, quidra-lang/dnn, and quidra-lang/vision (core release PR, first-party package compatibility CI PR, and both packages' initial release PRs) are authored and merged by the same single account, koba-jon, who is also recorded as making dnn's and vision's first contributions.

Recorded judgment for Recent maintenance, Distribution artifacts: Positive signals exist (recent releases, an explicit cross-repository compatibility CI process, signed/verified release tags, structured versioning against core compatibility ranges), which keep the score above a floor of zero. However the entire observable ecosystem is two packages maintained by a single contributor inside the language's own organization, with zero external stars/forks/independent contributors and only two releases each in a ~48-hour window, so there is essentially no independent community maintenance signal yet.

Recorded judgment for Security provenance: The core repository added CI specifically to continuously test the released vision and dnn packages against Quidra changes before they land in develop/main, so core changes cannot silently break first-party packages, which is evidence of deliberate maintenance process rather than ad hoc releases.

Recorded evidence
Registry discovery Level 0/4 → 0/20 points
Recent maintenance Level 1/4 → 5/20 points
Dependency metadata Level 2/4 → 10/20 points
Distribution artifacts Level 2/4 → 10/20 points
Security provenance Level 1/4 → 5/20 points
Evidence snapshot 2026-09-23

Evidence limits

Star/fork counts are recorded as declared evidence per the methodology but are explicitly not used as a standalone proxy; they are near-zero for a project this new and are not comparable in scale to established ecosystems for other benchmarked languages.

The archive also preserves an earlier 15/100 judgment. The published 30.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 6 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Registry discovery: Evidence was gathered via a single provider-brokered web-search retrieval route per package (project page + latest metadata + changelog), applied uniformly across all sampled packages to keep examination depth constant.

Recorded judgment for Recent maintenance, Distribution artifacts: Every sampled package across the frozen domain list shows releases or changelog entries within the 2026 calendar year, several within the last one to two months of the snapshot date, and most track new CPython release readiness within the same release cycle (3.14 classifiers already present for numpy, pandas, requests, cryptography, boto3, pytest, fastapi, aiohttp, pydantic per the pyreadiness.org cross-check). This indicates a highly active, well-maintained package ecosystem across sampled domains, supporting a high but not maximal normalized score to leave room for packages/domains outside the sampled set that may show slower cadence.

Recorded judgment for Dependency metadata: Latest version 2.34.2, last published May 14, 2026; 160 releases; release frequency about 1 month 4 days

Recorded judgment for Security provenance: All 11 predeclared functional domains have a canonical, officially-recognized, maintained third-party (or first-party-adjacent, e.g. Django) library available on PyPI.

Recorded evidence
Registry discovery Level 4/4 → 20/20 points
Recent maintenance Level 4/4 → 20/20 points
Dependency metadata Level 4/4 → 20/20 points
Distribution artifacts Level 4/4 → 20/20 points
Security provenance Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

Download counts and GitHub star counts cited above (e.g. FastAPI GitHub stars, requests daily download counts) are used only as declared corroborating context per package, never as the sole basis for either metric's score, consistent with the predeclared-sampling rule against single-proxy scoring. | This shard evaluates only the Python language; cross-language comparability of the same domain/selection-rule structure is the responsibility of the sibling per-language shards and the runner's aggregation step. | Evidence was gathered via a single provider-brokered web-search retrieval route per package (project page + latest metadata + changelog), applied uniformly across all sampled packages to keep examination depth constant.

The archive also preserves an earlier 97/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −70.00 points

All requirements

03 / metric.third_party_tool_availability

Effective evaluation weight 10%

Third-party tool availability

Evidence of independently maintained editor, analysis, build, debugging and developer-operations tools.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Inspect the same fixed tool categories for every language and require independently maintained then-current tools; official first-party tools may be cited only as context, not as third-party availability. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

0.00 / 100

The recorded levels are 0 + 0 + 0 + 0 + 0 = 0 out of 20 level units. The runner's conversion gives 0 × 5 = 0.00/100.

Recorded judgment for Editor IDE, Lint format static, Build test coverage, Debug profile, CI devops: No third-party (non quidra-lang-authored) editor plugins, linters, formatters, build-system integrations, CI marketplace actions, or debuggers were found. All tooling identified is official/first-party.

Recorded evidence
Editor IDE Level 0/4 → 0/20 points
Lint format static Level 0/4 → 0/20 points
Build test coverage Level 0/4 → 0/20 points
Debug profile Level 0/4 → 0/20 points
CI devops Level 0/4 → 0/20 points
Evidence snapshot 2026-09-23

Evidence limits

Star/fork counts and search-result absence are weak proxies (a tool could exist without being indexed yet); however, the consistent first-party-only pattern across every discovered artifact (compiler, LSP mentioned only in official docs, playground, formatter, first-party packages) plus zero independent hits after multiple targeted queries is convergent evidence of essentially no third-party tool ecosystem.

The archive also preserves an earlier 2/100 judgment. The published 0.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 7 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Editor IDE, Build test coverage, CI devops: Python exhibits essentially complete third-party tool coverage across every standard category (IDE, editor extension, linter/formatter, type-checker, package manager, debugger, profiler, test framework, CI/CD integration), with multiple competing, actively maintained options in most categories and a documented pattern of continued new-tool investment (uv/Ruff rapid adoption). This represents near-maximal ecosystem tool maturity, scored 97/100; the deduction reflects that a handful of niche/embedded or highly specialized tooling gaps still exist relative to a hypothetical perfect score.

Recorded judgment for Lint format static, Debug profile: A new generation of Rust-implemented Python tooling (uv, Ruff, Polars) has been adopted rapidly, with uv reaching double-digit usage within its first year, showing active third-party investment in the ecosystem beyond legacy tools.

Recorded evidence
Editor IDE Level 4/4 → 20/20 points
Lint format static Level 4/4 → 20/20 points
Build test coverage Level 4/4 → 20/20 points
Debug profile Level 4/4 → 20/20 points
CI devops Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 97/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −100.00 points

All requirements

04 / metric.production_adoption_deployment_evidence

Effective evaluation weight 10%

Production adoption / deployment evidence

Publicly checkable deployments, their scale, diversity, continuity and depth.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Use public, independently checkable production evidence; named deployments must be documented by the deploying organization or corroborated by credible independent sources. Popularity surveys alone do not prove deployment. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

0.00 / 100

The recorded levels are 0 + 0 + 0 + 0 + 0 = 0 out of 20 level units. The runner's conversion gives 0 × 5 = 0.00/100.

Recorded judgment for Named organizations, Scale criticality, Domain diversity, Deployment depth: No evidence of any production deployment, external company adoption, or real-world usage of Quidra was found. Given the language's release history begins only days before this snapshot, this outcome is expected and consistent with the language having no discoverable production footprint yet.

Recorded judgment for Current continuity: Quidra's own README/release history shows the language reaching only v0.1.0-v0.3.1, first tagged 2026-09-16 and most recently 2026-09-20, i.e. released within the week preceding this evaluation snapshot (2026-09-23).

Recorded evidence
Named organizations Level 0/4 → 0/20 points
Scale criticality Level 0/4 → 0/20 points
Domain diversity Level 0/4 → 0/20 points
Current continuity Level 0/4 → 0/20 points
Deployment depth Level 0/4 → 0/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

Best other · 7 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Named organizations, Scale criticality, Domain diversity, Current continuity: Python has abundant, independently verifiable production deployment evidence at extreme scale across multiple industry-leading companies and domains (social media, streaming, cloud storage, ride-hailing, aerospace), plus a large, work-majority professional survey population and growing production-web-framework adoption. The primary limitation noted in the source data itself is that Python's server-side web-hosting *market share* (per W3Techs) is a small single-digit percentage and includes a documented long tail of unpatched Python 2 deployments, which is why the score is 96/100 rather than a perfect score: overwhelming production evidence in application/backend/data domains, with a caveat o

Recorded judgment for Deployment depth: Independently documented, large-scale production use of Python exists at Instagram/Meta (backend serving over a billion daily users via Django), Netflix (data engineering, A/B testing, content delivery, and a failover system that shifts AWS-region traffic in minutes), Spotify (backend services and data pipelines, plus PyCon/EuroPython sponsorship), Dropbox (desktop client and backend architecture), Uber/Lyft/Pinterest (real-time location/routing and application-layer services), NASA (CAE/CAD/PDM infrastructure and the OpenMDAO framework), and Google/Amazon (recommendation engines and backend systems).

Recorded evidence
Named organizations Level 4/4 → 20/20 points
Scale criticality Level 4/4 → 20/20 points
Domain diversity Level 4/4 → 20/20 points
Current continuity Level 4/4 → 20/20 points
Deployment depth Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 96/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −100.00 points

All requirements

05 / metric.community_public_knowledge_availability

Effective evaluation weight 10%

Community / public knowledge availability

Searchable discussion, independent learning material, active groups and the breadth of public knowledge.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Inspect the same knowledge channels for every language and use then-current, publicly accessible evidence rather than raw search-result counts alone. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

15.00 / 100

The recorded levels are 1 + 1 + 0 + 1 + 0 = 3 out of 20 level units. The runner's conversion gives 3 × 5 = 15.00/100.

Recorded judgment for Q&A discussion, Learning materials, Current knowledge, International breadth: Normalized 0-100 (100 = best/richest independent community knowledge). Score of 5 reflects: (a) credit for genuine, substantial first-party documentation and an official interactive playground, which lower-bounds the score above zero and above a purely non-existent/undocumented language; (b) an essentially complete absence, within the retrievable public web, of independent third-party community knowledge channels (Q&A sites, forums, chat communities, tutorials, blogs, books) that the metric is specifically measuring, which is the dominant factor and keeps the score near the low end of the scale rather than at a mid-range value reserved for languages with some but limited community presence.

Recorded judgment for Events groups: Very small pull-request count (single digits), authored by one maintainer account (koba-jon), several with Claude Code co-authorship notes, indicating a small, largely non-community-contributed codebase rather than an active outside contributor base.

Recorded evidence
Q&A discussion Level 1/4 → 5/20 points
Learning materials Level 1/4 → 5/20 points
Events groups Level 0/4 → 0/20 points
Current knowledge Level 1/4 → 5/20 points
International breadth Level 0/4 → 0/20 points
Evidence snapshot 2026-09-23

Evidence limits

Web search snapshots are not exhaustive and could miss small/private community venues (e.g. a closed Discord invite-only server) that do not surface in general search indexes. | Raw popularity indicators such as GitHub star/fork counts were not retrievable via the search tool in this session (no aggregator page reported a numeric count), so this evidence relies on presence/absence signals from search results rather than a numeric popularity proxy; per methodology this is used only as declared, non-standalone evidence. | This is a single-language shard; no cross-language comparison table was computed here, only the Quidra-specific raw evidence, consistent with the language-sharded task assignment.

The archive also preserves an earlier 5/100 judgment. The published 15.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 6 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Q&A discussion, Current knowledge, International breadth: Python exhibits comprehensive, high-depth publicly available community knowledge across every examined channel category: (a) largest or near-largest Q&A archive on the dominant general-purpose Q&A platform with sustained multi-year growth driven by then-current (AI/ML) demand rather than legacy decline, (b) an officially curated, multi-channel community index maintained by the language's own foundation (python.org), unusual among languages for its breadth (chat, forum, newsletter, social, IRC all officially linked from one page), (c) a very large (430K+), actively staffed real-time help community, and (d) a mature, officially-adopted governance discussion forum that has fully replaced legacy mail

Recorded judgment for Learning materials: pythondiscord.com: over 100 dedicated expert Helpers available around the clock in every timezone

Recorded judgment for Events groups: python.org/community/ lists 6+ distinct official channel types: Slack (PySlackers), Discord (Python Discord), LinkedIn group, IRC (Libera.Chat), weekly newsletter (Python Weekly), plus links to local user groups/PSF survey

Recorded evidence
Q&A discussion Level 4/4 → 20/20 points
Learning materials Level 4/4 → 20/20 points
Events groups Level 4/4 → 20/20 points
Current knowledge Level 4/4 → 20/20 points
International breadth Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

Stack Overflow tag counts and Discord member counts are raw popularity indicators; per methodology they are used only as one of four declared evidence legs, not as the sole basis for the score. Discourse topic-volume counts could not be retrieved as a discrete number in the time/route budget available (only qualitative descriptions were retrievable), so that leg is supported qualitatively rather than quantitatively. All four retrieval routes and depth were held identical to the process that would be applied to every other assigned language in this shard set.

The archive also preserves an earlier 97/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −85.00 points

All requirements

06 / metric.core_tooling_availability_quality

Effective evaluation weight 5%

Core tooling availability / quality

The build, formatting, testing, documentation and diagnostic tools available in the core workflow.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Evaluate the maintained first-party/de-facto core developer toolchain actually available at the recorded snapshot; score implemented capability, not planned design. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

75.00 / 100

The recorded levels are 4 + 3 + 2 + 3 + 3 = 15 out of 20 level units. The runner's conversion gives 15 × 5 = 75.00/100.

Recorded judgment for Build execute, Format lint, Diagnostic quality: Quidra ships a coherent, well-documented, first-party core tool surface (compiler+diagnostics, formatter, REPL, inspect/patch, describe) that is unusually strong for LLM-oriented tooling, which supports a mid score rather than a low one. However availability/quality here is bounded by there being a single implementation, no external/third-party tool ecosystem, and pre-1.0 status, which keeps the score in the middle of the scale rather than high.

Recorded judgment for Test: Release checks (docs/development.md) enumerate GCC/Clang builds, Linux/macOS/Windows CI, sanitizer and warnings-as-errors builds, ctest, parser/checker negative tests, native runtime smoke tests, standard-library/module tests, package/version/lockfile tests, REPL/CLI tests, and packaging tests as release gates, indicating a real (if internally-defined) build/test integration discipline.

Recorded judgment for Documentation / package tooling: Frozen proxy: score the availability and functional quality of the core first-party developer tool surface (compiler CLI, diagnostics, formatter, REPL, source-introspection/patch tooling, build/link integration) as observable from the frozen repository documentation snapshot. Sub-scores: (a) compiler CLI completeness/diagnostics quality, (b) formatter presence/determinism, (c) REPL/interactive tooling, (d) machine-tooling (inspect/patch/describe) for programmatic editing, (e) build/link integration and cross-platform CI coverage, (f) project maturity signal (single vs multiple implementations, versioning discipline).

Recorded evidence
Build execute Level 4/4 → 20/20 points
Format lint Level 3/4 → 15/20 points
Test Level 2/4 → 10/20 points
Documentation / package tooling Level 3/4 → 15/20 points
Diagnostic quality Level 3/4 → 15/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 55/100 judgment. The published 75.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 7 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Build execute, Documentation / package tooling: Measures maturity, availability, and quality of the core first-party/de-facto-standard developer tooling stack that is NOT already covered by the separate Debugger/Profiler Support metric or the separate Package/Dependency Management metric: (1) official interpreter/runtime implementation, (2) standardized build/compile tooling, (3) linter tooling, (4) formatter tooling, (5) test framework, (6) static type-checking tooling. Excludes debugger/profiler products (scored under metric.debugger_profiler_support) and excludes the package manager / registry (scored under metric.package_dependency_management_quality) to avoid double counting per the Ecosystem spec's mixed-concept split.

Recorded judgment for Format lint: Ruff has become the recommended unified linter+formatter for new Python projects, replacing the Flake8/Black/isort/pyupgrade patchwork.

Recorded judgment for Test: Test framework: existence of a mature, actively maintained, de-facto-standard test framework plus a built-in/standard-library fallback.

Recorded judgment for Diagnostic quality: By 2026 the toolchain has consolidated around fast Rust-based tools (uv, Ruff) for performance-critical parts alongside mature Python libraries (pytest, mypy) for the rest.

Recorded evidence
Build execute Level 4/4 → 20/20 points
Format lint Level 4/4 → 20/20 points
Test Level 4/4 → 20/20 points
Documentation / package tooling Level 4/4 → 20/20 points
Diagnostic quality Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

Scores reflect a single analyst's reading of the frozen sub-criteria against 2026 secondary-source writeups plus one primary source (python.org/Wikipedia CPython) rather than a full independent hands-on tooling audit; multiple competing tools (mypy vs pyright vs ty) were credited as ecosystem strength but also reflect some fragmentation, reflected in the lower static-type-checking sub-score.

The archive also preserves an earlier 96/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −25.00 points

All requirements

07 / metric.package_dependency_management_quality

Effective evaluation weight 5%

Package / dependency management quality

The quality of dependency resolution, manifests, workspaces, security and publishing workflows.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Evaluate the standard/de-facto package workflow using official documentation and then-current registry/tool behavior; ecosystem size is scored elsewhere. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

60.00 / 100

The recorded levels are 4 + 3 + 1 + 2 + 2 = 12 out of 20 level units. The runner's conversion gives 12 × 5 = 60.00/100.

Recorded judgment for Resolver lock: The lockfile format (quidra.lock v3) records distribution name, import name, version, and a SHA-256 content hash per package, and compilation checks both version and hash, giving strong reproducibility guarantees per docs/packages.md.

Recorded judgment for Manifest constraints, Workspace multi, Registry security, Publish version: The package-manager mechanism itself (manifest schema, semver ranges, ABI-vs-version separation, hashed lockfiles, immutable-tag-only installs) is comparatively well engineered and would score well in isolation. But package/dependency management quality per the Ecosystem methodology also requires package-manager maturity and package ecosystem activity together: Quidra has no registry, no automatic transitive dependency resolution (a core capability of mature package managers), and an ecosystem consisting of only two known first-party packages.

Recorded evidence
Resolver lock Level 4/4 → 20/20 points
Manifest constraints Level 3/4 → 15/20 points
Workspace multi Level 1/4 → 5/20 points
Registry security Level 2/4 → 10/20 points
Publish version Level 2/4 → 10/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 27/100 judgment. The published 60.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 5 tied languages

Rust 1.95.0

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Resolver lock, Registry security: Meets top-band criteria: fully integrated first-party package manager, Cargo.lock reproducibility, semver resolution with feature unification, native workspace support, and a large-scale, high-traffic central registry (crates.io) with community security-auditing tooling; not a perfect 100 because some auditing/supply-chain tooling (cargo-audit, cargo-deny) remains third-party rather than bundled by default.

Recorded judgment for Manifest constraints, Workspace multi, Publish version: Cargo is a first-party, deeply integrated package manager backed by the large, actively used crates.io registry, with lockfiles, semver-based resolution, and workspace support.

Recorded evidence
Resolver lock Level 4/4 → 20/20 points
Manifest constraints Level 4/4 → 20/20 points
Workspace multi Level 4/4 → 20/20 points
Registry security Level 4/4 → 20/20 points
Publish version Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 96/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −40.00 points

All requirements

08 / metric.ide_editor_support_quality

Effective evaluation weight 5%

IDE / editor support quality

Implemented navigation, completion, refactoring, debugging and editor support.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Inspect the same editor capabilities across at least the dominant general-purpose editor and any major dedicated IDE when available. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

70.00 / 100

The recorded levels are 4 + 4 + 3 + 2 + 1 = 14 out of 20 level units. The runner's conversion gives 14 × 5 = 70.00/100.

Recorded judgment for Syntax navigation: No dedicated TextMate/semantic-highlighting grammar package for Quidra was found distributed independently of the playground's own embedded editor.

Recorded judgment for Completion types: 7/20 (2 of 6 tracked features -- diagnostics and formatting -- documented; completion/hover/go-to-def/rename not documented)

Recorded judgment for Refactor actions: A search for an official Quidra VS Code/JetBrains/Open VSX marketplace extension returned no matching listing; only unrelated languages (e.g. Quil, Quint, Q#, q) were found under similar-sounding names.

Recorded judgment for Debug integration, Breadth maintenance: A first-party browser Playground (quidra-lang/playground) exists, running the real compiler frontend compiled to WebAssembly, offering in-editor diagnostics with codes/ranges, formatting, typed-IR view, structural inspect (node ids/kinds/spans/hashes/types), and hash-checked patch application -- but it explicitly excludes program execution ('No Run, REPL, JIT or any other execution'), multi-file editing, and package management.

Recorded evidence
Syntax navigation Level 4/4 → 20/20 points
Completion types Level 4/4 → 20/20 points
Refactor actions Level 3/4 → 15/20 points
Debug integration Level 2/4 → 10/20 points
Breadth maintenance Level 1/4 → 5/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 40/100 judgment. The published 70.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 6 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Syntax navigation, Refactor actions: Refactoring/navigation: Pylance's Code Actions include 'Move symbol to...' and 'Move symbol to new file', accessible via a light bulb Code Action or right-click Refactor menu, and Workspace Symbol search now surfaces symbols from packages installed in the active virtual environment as of the March 2026 release.

Recorded judgment for Completion types: Language server maturity: Pylance is described as the default language server for Python in VS Code, based on Microsoft's Pyright static type checking tool, leveraging type stubs and lazy type inferencing for a highly-performant development experience.

Recorded judgment for Debug integration: JetBrains: PyCharm is a comprehensive dedicated Python IDE combining intelligent code assistance with advanced debugging, testing, and profiling tools, with integrated version control and database tools per ComponentSource's PyCharm product listing.

Recorded judgment for Breadth maintenance: Emacs: lsp-pyright is a maintained lsp-mode client leveraging the pyright and basedpyright language servers for Emacs, giving Emacs the same LSP-based IntelliSense as VS Code and Neovim.

Recorded evidence
Syntax navigation Level 4/4 → 20/20 points
Completion types Level 4/4 → 20/20 points
Refactor actions Level 4/4 → 20/20 points
Debug integration Level 4/4 → 20/20 points
Breadth maintenance Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 95/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −30.00 points

All requirements

09 / metric.debugger_profiler_support

Effective evaluation weight 5%

Debugger / profiler support

Usable debugging and profiling facilities around the implementation.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Evaluate implemented, documented debugging/profiling workflows on then-current supported toolchains; generic OS tools count only when language-aware use is documented. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

45.00 / 100

The recorded levels are 3 + 2 + 1 + 1 + 2 = 9 out of 20 level units. The runner's conversion gives 9 × 5 = 45.00/100.

Recorded judgment for Basic debug: The same README passage explicitly disclaims full source mapping: 'then-current integration guarantees native function/symbol visibility; fine-grained Quidra source-line DWARF mapping is not yet claimed,' meaning breakpoints/stepping are only guaranteed at the native symbol level, not reliable Quidra-source-line granularity.

Recorded judgment for Advanced debug, CPU profile, Memory profile: 8/20 (ASan/UBSan instrumentation exists for generated programs, but documentation frames it around compiler CI rather than a general user-facing debugging/profiling workflow)

Recorded judgment for Integration platforms: Same source: 'quidra debug FILE.qui launches LLDB or GDB (or QUIDRA_DEBUGGER) against that executable,' confirming real integration with mainstream native debuggers rather than a bespoke one.

Recorded evidence
Basic debug Level 3/4 → 15/20 points
Advanced debug Level 2/4 → 10/20 points
CPU profile Level 1/4 → 5/20 points
Memory profile Level 1/4 → 5/20 points
Integration platforms Level 2/4 → 10/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

Best other · 5 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Basic debug: Built-in debugger: Python ships pdb in the standard library, documented as enabling users to 'step through code, analyze stack frames and set breakpoints' as part of the stdlib debugging and profiling docs.

Recorded judgment for Advanced debug: Advanced debugger capabilities: debugpy (the VS Code/DAP-based debugger) supports remote attach debugging — 'Debugpy supports remote debugging, allowing you to attach to running processes on remote servers' — plus conditional breakpoints, hit counts, logpoints, and multi-threaded application debugging (with noted limitations for some features).

Recorded judgment for CPU profile, Memory profile: Profiler breadth: Python's standard library includes cProfile and profile for deterministic profiling, described as providing 'deterministic profiling of Python programs' with pstats as a statistics browser.

Recorded judgment for Integration platforms: IDE integration: The VS Code Python Debugger extension 'is automatically installed along with the Python extension for VS Code' and 'offers debugging features with debugpy for several types of Python applications, including scripts, web apps, remote processes and more.'

Recorded evidence
Basic debug Level 4/4 → 20/20 points
Advanced debug Level 4/4 → 20/20 points
CPU profile Level 4/4 → 20/20 points
Memory profile Level 4/4 → 20/20 points
Integration platforms Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 95/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −55.00 points

All requirements

10 / metric.build_test_integration

Effective evaluation weight 5%

Build / test integration

Evidence of repeatable builds, test execution, automation and integration.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Evaluate the standard/de-facto build-to-test-to-CI path with official tool documentation plus then-current integration evidence. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

60.00 / 100

The recorded levels are 3 + 2 + 3 + 1 + 3 = 12 out of 20 level units. The runner's conversion gives 12 × 5 = 60.00/100.

Recorded judgment for Build system: Is there one canonical, documented build system with a single source of truth for project metadata, usable identically across supported platforms?

Recorded judgment for Test framework, CI matrix: A small deduction (-6) is applied because evidence is observational (repository docs, PR/release descriptions and CI-matrix summaries) rather than a fully reproduced live CI log inspection, and because the project is a young pre-1.0 codebase with only two tagged releases observed.

Recorded judgment for Coverage, Config integration: No direct access to live CI run logs or coverage numbers was available in this packet; scoring relies on the project's own frozen release/build documentation cross-checked against externally observable release and PR history for the same repository.

Recorded evidence
Build system Level 3/4 → 15/20 points
Test framework Level 2/4 → 10/20 points
CI matrix Level 3/4 → 15/20 points
Coverage Level 1/4 → 5/20 points
Config integration Level 3/4 → 15/20 points
Evidence snapshot 2026-09-23

Evidence limits

No direct access to live CI run logs or coverage numbers was available in this packet; scoring relies on the project's own frozen release/build documentation cross-checked against externally observable release and PR history for the same repository.

The archive also preserves an earlier 91/100 judgment. The published 60.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 6 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Build system, Config integration: Frozen 5-criterion rubric (20 points each, 0-100 total), applied identically to all 10 languages in this evaluation. Measures maturity of the build-to-test-to-CI pipeline: (1) standard/native build tooling and its integration with the packaging system; (2) maturity/adoption of standard or de-facto test framework(s); (3) ease of CI integration incl. multi-version matrix testing; (4) coverage tooling integration and CI-gate support; (5) ability to unify build/test/coverage/lint config in one place.

Recorded judgment for Test framework: Standard or de-facto test framework maturity and adoption (stdlib unittest + de-facto pytest)

Recorded judgment for CI matrix: tox and nox provide first-class multi-Python-version matrix testing, with documented GitHub Actions recipes (setup-python + tox) and native uv acceleration paths.

Recorded judgment for Coverage: coverage.py integrates with pytest via pytest-cov with minimal setup, supports branch coverage, and multiple report formats (term, term-missing, html, xml) usable for CI gating via --cov-fail-under.

Recorded evidence
Build system Level 4/4 → 20/20 points
Test framework Level 4/4 → 20/20 points
CI matrix Level 4/4 → 20/20 points
Coverage Level 4/4 → 20/20 points
Config integration Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 94/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −40.00 points

All requirements

11 / metric.documentation_quality

Effective evaluation weight 5%

Documentation quality

The coverage, usability and maintenance of language and toolchain documentation.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Prioritize official then-current-version documentation, then independently maintained supplementary material; community size alone is not documentation quality. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

65.00 / 100

The recorded levels are 4 + 3 + 3 + 2 + 1 = 13 out of 20 level units. The runner's conversion gives 13 × 5 = 65.00/100.

Recorded judgment for Language reference, Examples guides, Version navigation, Generation localization: A deduction (-6) reflects that documentation is concentrated in in-repo markdown/spec files rather than a generated, versioned public docs site, and that some cross-file restatement risk is only partially mitigated (README examples had to be corrected to match contracts, per observed PRs).

Recorded judgment for Library API: Exhaustive language semantics reference covering types, arrays, bin, classes, references, generics, enums, tensors, neural values, and standard namespaces in one document.

Recorded evidence
Language reference Level 4/4 → 20/20 points
Library API Level 3/4 → 15/20 points
Examples guides Level 3/4 → 15/20 points
Version navigation Level 2/4 → 10/20 points
Generation localization Level 1/4 → 5/20 points
Evidence snapshot 2026-09-23

Evidence limits

Evidence is drawn from the frozen docs/ directory embedded in this task plus externally observable repository artifacts (README, PRs, releases); no independent readability/clarity NLP scoring was performed, consistent with the proxy-based rubric used here.

The archive also preserves an earlier 90/100 judgment. The published 65.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 6 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Language reference, Library API: The official docs cover the language reference, tutorial, and full standard library API reference as a single maintained corpus licensed under the PSF License, with examples additionally under a permissive license.

Recorded judgment for Examples guides: The official docs have a dedicated, PEP-545-governed translation program with a public dashboard and multiple in-production language translations reachable via a language switcher, plus a large surrounding community documentation ecosystem (Real Python, countless guides).

Recorded judgment for Version navigation: Sphinx, originally built for Python's own docs, provides mature autodoc extraction from docstrings into multi-format output (HTML, PDF, EPUB) and is the tool used for Python's official documentation itself as well as major libraries.

Recorded judgment for Generation localization: Frozen 5-criterion rubric (20 points each, 0-100 total), applied identically to all 10 languages. Measures the maturity and completeness of the documentation ecosystem: (1) official reference/tutorial completeness and currency; (2) standard-library API reference completeness; (3) maturity of auto-doc-generation tooling (docstring-to-published-docs pipeline); (4) consistency/officiality of a docstring/comment convention; (5) breadth of supplementary community documentation and internationalization.

Recorded evidence
Language reference Level 4/4 → 20/20 points
Library API Level 4/4 → 20/20 points
Examples guides Level 4/4 → 20/20 points
Version navigation Level 4/4 → 20/20 points
Generation localization Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 93/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −35.00 points

All requirements

12 / metric.installation_distribution_experience

Effective evaluation weight 5%

Installation / distribution experience

How well users can obtain, install, update and verify supported distributions.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Evaluate then-current official/de-facto installation paths on the same major desktop/server platforms and use only documented distribution channels. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

60.00 / 100

The recorded levels are 3 + 1 + 3 + 2 + 3 = 12 out of 20 level units. The runner's conversion gives 12 × 5 = 60.00/100.

Recorded judgment for Installers: quidra install resolves only exact vMAJOR.MINOR.PATCH release tags via git, optionally downloading HTTPS platform assets (asset.<platform>) that are inspected before extraction; asset-backed installs additionally require curl and tar, and official Linux packages declare those runtime dependencies. This is the package manager for Quidra *libraries* (dnn, vision), not for installing the Quidra compiler/toolchain itself.

Recorded judgment for Package version manager, Channel breadth, Install docs: Positive signals: automated, CI-gated multi-asset GitHub Releases for every tagged version; a deliberately safety-hardened library package installer (exact-tag resolution, manifest/hash verification, sandboxed asset extraction) documented in packages.md; clear, reproducible source-build instructions. Negative signals: no first-party OS/language package-manager listing found (apt/brew/choco/winget/scoop), no curl-pipe-bash installer, no Docker image, no version-manager plugin, and building from source requires several native library dev headers (libcurl, libpng, libjpeg, libtiff, libwebp) plus a pre-existing Clang 15+ install because the project intentionally does not bundle a backend toolcha

Recorded judgment for Artifact integrity: Tagged releases v0.1.0, v0.2.0, v0.3.0, v0.3.1 each carry attached release assets (6, 6, 6, and 4 assets respectively), consistent with automated per-platform artifact publication, but no evidence was found of listing in any OS or language package manager (apt, Homebrew, Chocolatey, winget, scoop), no install-script (curl|bash) entry point, and no Docker image or version-manager (asdf/mise-style) integration.

Recorded evidence
Installers Level 3/4 → 15/20 points
Package version manager Level 1/4 → 5/20 points
Channel breadth Level 3/4 → 15/20 points
Artifact integrity Level 2/4 → 10/20 points
Install docs Level 3/4 → 15/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 42/100 judgment. The published 60.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 5 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Installers: Python.org provides dedicated Windows and macOS installer pages; per search evidence, 'On Windows the most stable build is available from the official download page' and users 'should download and run the installer from that page' (wiki.python.org/BeginnersGuide). macOS installer packages are versioned and dated on python.org/downloads/macos/ (e.g. Python 3.14.7 - Aug. 5, 2026).

Recorded judgment for Package version manager: pip ships bundled with the interpreter (Wikipedia: 'Most distributions of Python come with pip preinstalled... Python 2.7.9 and later... and Python 3.4 and later include pip by default').

Recorded judgment for Channel breadth: Multiple independently maintained install/version-manager channels exist beyond the base installer: the new Python Install Manager and Microsoft Store on Windows, Homebrew on macOS, apt/yum/pacman on Linux, and version managers pyenv, conda/Anaconda/Miniconda, and uv (which 'can install and manage Python versions directly, replacing the need for tools like pyenv' and pulls 'prebuilt CPython binaries from python-build-standalone' with installs that 'don’t need admin privileges'). uv itself ships via a standalone shell/PowerShell installer, pip/pipx, Homebrew, Chocolatey, WinGet and Scoop. Score: 20/20.

Recorded judgment for Artifact integrity: C4 Artifact integrity: code signing / checksums / provenance attestation on official release artifacts - 20 pts

Recorded judgment for Install docs: Official documentation (python.org/downloads, docs.python.org/3/using/windows.html, wiki.python.org BeginnersGuide) gives per-OS install guidance, and third-party tutorials (Real Python) closely mirror and extend it with platform-specific nuance (Store vs. traditional installer vs. WSL on Windows; Homebrew vs. official installer on macOS; package manager vs. source build on Linux).

Recorded evidence
Installers Level 4/4 → 20/20 points
Package version manager Level 4/4 → 20/20 points
Channel breadth Level 4/4 → 20/20 points
Artifact integrity Level 4/4 → 20/20 points
Install docs Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

Component weighting and thresholds are proxy judgments made once and intended to be reused unchanged for the other 9 languages by sibling shards; raw popularity/download counts were not used as a standalone score input.

The archive also preserves an earlier 97/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −40.00 points

All requirements

13 / metric.toolchain_stability_release_maturity

Effective evaluation weight 5%

Toolchain stability / release maturity

Observable release, compatibility and maintenance commitments; language age itself earns no points.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Score then-current observable release/support practices. Language age and first-release date are context only and never directly score. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

50.00 / 100

The recorded levels are 2 + 1 + 2 + 3 + 2 = 10 out of 20 level units. The runner's conversion gives 10 × 5 = 50.00/100.

Recorded judgment for Release cadence, Support policy, Compat deprecation, Release integrity, Governance change: Positive signals: a genuinely mechanized, reproducible, single-source-of-truth release pipeline (project.toml -> sync_metadata.py -> CI-gated, immutable, bot-created tags) with independently versioned ABI/IR/schema compatibility numbers, a documented and enforced SemVer policy, an unusually rigorous pre-release CI/test checklist, verified (GPG-signed) release commits, and real observed history including a patch-level maintenance release (v0.3.1). Negative signals: the language is still pre-1.0, so by its own documented policy 0.X.0 releases may include breaking language/stdlib changes and there is no published LTS/backport support-window commitment beyond 'fix release blockers on develop'; t

Recorded evidence
Release cadence Level 2/4 → 10/20 points
Support policy Level 1/4 → 5/20 points
Compat deprecation Level 2/4 → 10/20 points
Release integrity Level 3/4 → 15/20 points
Governance change Level 2/4 → 10/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 61/100 judgment. The published 50.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 6 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Release cadence: D1 Documented, predictable release cadence actually followed in practice - 20 pts

Recorded judgment for Support policy: PEP 602 documents an explicit five-year lifecycle per release: 'After the release of Python 3.X.0, the 3.X series is maintained for five years,' with roughly two years of full bugfix support followed by three years of security-only, source-only fixes ('For the next thirty six months (3 years) it receives security updates and source-only releases are made on an as-needed basis'). Concrete EOL dates are published and tracked (e.g. 'Python 3.10 reaches end of life on October 31, 2026').

Recorded judgment for Compat deprecation: Python maintains a documented deprecation/removal policy (PEP 387) with warnings issued at least one release ahead of removal, and stdlib/runtime changes are gated through the PEP process; however, real behavioral changes occur release-over-release (e.g. removals of previously deprecated stdlib modules, and multi-year phased rollout of free-threading changes to the C API surface), which is a modest compatibility-stability cost, though it is a governed rather than ad hoc process, as shown by the phased PEP 703/PEP 779 rollout ('Phase II would make the free-threaded build officially supported but still optional'). Score: 18/20.

Recorded judgment for Release integrity: Release artifacts carry verifiable provenance: Sigstore signing (from 3.9.14/3.10.7/3.11.0 onward), Apple Developer ID-signed macOS installers, and Authenticode/Azure Trusted Signing-signed Windows binaries, all documented on the official downloads page. Score: 19/20.

Recorded judgment for Governance change: A concrete recent test of governance stability: PEP 2026 proposed switching CPython to calendar versioning ('Python 3.26 will be the first version after this change, released in 2026'), went through the full PEP process, and was ultimately not adopted — external tracking confirms 'the Steering Council rejected it on 2025-02-06' and 'Python continues with the existing 3.x numbering.' This shows disruptive scheme changes are filtered through formal deliberation rather than applied unilaterally, preserving existing versioning continuity. Similarly, the multi-year, criteria-gated PEP 703/PEP 779 process for free-threading ('this PEP establishes clear expectations and requirements for moving to P

Recorded evidence
Release cadence Level 4/4 → 20/20 points
Support policy Level 4/4 → 20/20 points
Compat deprecation Level 4/4 → 20/20 points
Release integrity Level 4/4 → 20/20 points
Governance change Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

Compatibility-policy and governance-track-record components involve qualitative judgment about the severity/frequency of breaking changes and are intended to be applied with the same rubric and threshold language to the other 9 languages by sibling shards; language age/first-release year were excluded from scoring per the frozen instruction.

The archive also preserves an earlier 95/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −50.00 points

All requirements

14 / metric.implemented_platform_coverage

Effective evaluation weight 5%

Implemented platform coverage

The operating systems, architectures and deployment targets actually supported at the evidence snapshot.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Count only platforms with at the recorded snapshot implemented and documented build/run support; design intent without working support does not count. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

40.00 / 100

The recorded levels are 4 + 4 + 0 + 0 + 0 = 8 out of 20 level units. The runner's conversion gives 8 × 5 = 40.00/100.

Recorded judgment for Desktop server OS, Mobile: Platform confirmation relies on documentation/README/CI descriptions rather than direct execution of the release pipeline; packaging detail (Linux DEB/TGZ, macOS TGZ, Windows ZIP via CPack, per README search snippet) corroborates 3 desktop OS artifacts but ARM64/mobile targets are not evidenced and are scored as absent rather than assumed present.

Recorded judgment for Architectures: No architecture-specific (ARM64) Linux target is named in the docs or README beyond WSL/NVIDIA driver interop notes; only generic 'Linux' CI is described.

Recorded judgment for Embedded: Same CI line covers macOS; native runtime boundary documents Mach-O executable-path API self-location and Metal compute path is macOS/Apple Silicon specific.

Recorded judgment for Web WASM other: Architecture doc describes a dedicated 'WebAssembly frontend boundary' with QUIDRA_BUILD_WASM_FRONTEND building quidra_core plus wasm_api.cpp and a versioned quidra_wasm_invoke entry point; the quidra-lang/playground repo independently confirms a shipped product described as 'the real Quidra compiler frontend, compiled to WebAssembly and running in your browser tab.'

Recorded evidence
Desktop server OS Level 4/4 → 20/20 points
Architectures Level 4/4 → 20/20 points
Mobile Level 0/4 → 0/20 points
Embedded Level 0/4 → 0/20 points
Web WASM other Level 0/4 → 0/20 points
Evidence snapshot 2026-09-23

Evidence limits

Platform confirmation relies on documentation/README/CI descriptions rather than direct execution of the release pipeline; packaging detail (Linux DEB/TGZ, macOS TGZ, Windows ZIP via CPack, per README search snippet) corroborates 3 desktop OS artifacts but ARM64/mobile targets are not evidenced and are scored as absent rather than assumed present.

The architecture component was recertified at 4/4, contributing 20 points, but its preserved finding describes only generic Linux CI and says an ARM64 Linux target was not evidenced. The row's limitation also says ARM64 targets were not evidenced. This inherited text does not independently justify full coverage across the rubric's major CPU architectures; the displayed points reproduce the recorded level.

The WebAssembly finding describes running the compiler frontend in a browser. By itself, that does not demonstrate WebAssembly as a deployment target for generated Quidra programs. The recertified web/WASM/other component is 0/4; the inherited frontend finding does not directly establish or refute support under that deployment-target criterion. The embedded component's attached finding describes macOS and Apple Silicon runtime details, which do not independently evaluate embedded or bare-metal deployment support for its recorded 0/4 level.

The archive also preserves an earlier 80/100 judgment. The published 40.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other

C++ 18.1.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for Desktop server OS, Mobile: Frozen proxy: a fixed universe of 10 platform-target categories, each scored 10 points (binary) if a mainstream, actively-maintained toolchain/SDK produces production-grade artifacts for that target. Categories: (1) Linux x86_64, (2) Linux ARM64, (3) Windows x86_64, (4) macOS (x86_64 and ARM64), (5) Android (mobile, via NDK), (6) iOS (mobile, via Xcode/Clang), (7) WebAssembly (browser/edge/server), (8) Embedded/bare-metal microcontrollers (no OS), (9) Major cloud serverless functions (custom runtime), (10) BSD/Unix variants.

Recorded judgment for Architectures: GCC and Clang are the reference C++ toolchains for Linux on both x86_64 and ARM64; treated as common knowledge baseline, not separately re-verified by search in this pass.

Recorded judgment for Embedded: GCC-based bare-metal cross toolchains (e.g. arm-none-eabi-gcc, AVR-GCC) and Clang/LLVM embedded targets are long-standing, widely deployed C++ toolchains for microcontroller firmware; common-knowledge baseline, not separately re-verified by search in this pass.

Recorded judgment for Web WASM other: Verified: Emscripten compiles C and C++ to WebAssembly using LLVM and Binaryen, with output able to run on the Web, in Node.js, and in wasm runtimes.

Recorded evidence
Desktop server OS Level 4/4 → 20/20 points
Architectures Level 4/4 → 20/20 points
Mobile Level 4/4 → 20/20 points
Embedded Level 4/4 → 20/20 points
Web WASM other Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

Binary per-category scoring does not capture depth/quality differences within a supported platform (e.g. tier-1 CI-tested vs best-effort community support); this is a coarse breadth proxy, not a quality proxy. Some categories were confirmed via search this pass, others rely on stable, well-established toolchain facts not independently re-searched to conserve the shared query budget.

Quidra − best other: −60.00 points

All requirements

15 / metric.external_integration_coverage

Effective evaluation weight 5%

External integration coverage

Working bindings and integrations with external languages, native systems, data systems and services.

How it is measured

Five equally weighted components are each assigned a judgment level from 0 to 4. The trusted runner converts each level to 0, 5, 10, 15 or 20 points and sums the five contributions. Evaluate working documented integrations and binding ecosystems, not only theoretical FFI design; FFI language design itself belongs to Language Quality. Score breakdown · Methodology reference.

The difference is the sum of the recorded component-level differences, multiplied by five. A tie means equal published rubric totals; the underlying findings can differ.

Evidence snapshot: 2026-09-23. These are archived evidence judgments under a fixed retrieval budget; an absence finding is scoped to that search and date.

Quidra

Quidra 0.3.0

35.00 / 100

The recorded levels are 4 + 0 + 0 + 1 + 2 = 7 out of 20 level units. The runner's conversion gives 7 × 5 = 35.00/100.

Recorded judgment for C / native: C FFI / native library interop

Recorded judgment for Language bindings, Database data: No database driver, gRPC/WebSocket, or other network-protocol standard module beyond HTTP(S) is documented anywhere in the reviewed specification or README.

Recorded judgment for Network cloud: score = 100 * confirmed_count / category_total, then adjusted -5 for narrower reliability/breadth signal (single-protocol network stack, no DB/queue integrations) = 90 - 5

Recorded judgment for Extension embedding: Language spec defines a narrow explicit C ABI boundary: 'External C declarations remain explicit in the checked program and typed IR' with scalar/void results and explicit string/bin borrow expansion to (pointer, uint64 length) pairs.

Recorded evidence
C / native Level 4/4 → 20/20 points
Language bindings Level 0/4 → 0/20 points
Database data Level 0/4 → 0/20 points
Network cloud Level 1/4 → 5/20 points
Extension embedding Level 2/4 → 10/20 points
Evidence snapshot 2026-09-23

Evidence limits

Confirmation is documentation/README-based rather than execution-based; integration 'reliability' (e.g., libcurl timeout/redirect limits, FFmpeg lazy decoder reconstruction, OpenAI structured-output re-validation) is asserted in source docs but not independently load-tested by this worker. The one unconfirmed category (database/other network protocols) is treated as absent rather than assumed present, per the rubric's fail-closed design.

The archive also preserves an earlier 85/100 judgment. The published 35.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Best other · 3 tied languages

Python 3.12.3

100.00 / 100

The recorded levels are 4 + 4 + 4 + 4 + 4 = 20 out of 20 level units. The runner's conversion gives 20 × 5 = 100.00/100.

Recorded judgment for C / native: psycopg/psycopg2 (PostgreSQL), mysqlclient/PyMySQL (MySQL) and the standard-library sqlite3 module are the long-standing de-facto default drivers used across the ecosystem.

Recorded judgment for Language bindings: Django, Flask and FastAPI are mainstream, actively maintained web frameworks each with large plugin ecosystems.

Recorded judgment for Database data, Network cloud, Extension embedding: Breadth and reliability of actively-maintained, production-ready working integrations with external systems and platforms (not the language's FFI/interop design, which is scored under Language Quality).

Recorded evidence
C / native Level 4/4 → 20/20 points
Language bindings Level 4/4 → 20/20 points
Database data Level 4/4 → 20/20 points
Network cloud Level 4/4 → 20/20 points
Extension embedding Level 4/4 → 20/20 points
Evidence snapshot 2026-09-23

Evidence limits

No new provider call was made for this recertification row. Historical normalized scores are retained only for audit comparison and are not used by the v2 runner arithmetic.

The archive also preserves an earlier 94/100 judgment. The published 100.00 is the later runner-owned rubric total; the earlier number is not used in the calculation.

Quidra − best other: −65.00 points

All requirements

Read the result in context

These are descriptive results for the frozen benchmark sample. They do not establish statistical superiority beyond that sample. Each evaluation stands on its own; the project publishes no cross-evaluation total.

The analysis above reads the recorded evidence alongside the scores. The machine-readable rankings and score breakdown are the published result. The methodology reference describes the linked source snapshot; its full template is not identical to the run’s recorded template.

Return to all ten languages in Ecosystem