aswritten.ai · public perspective

Perspective › Rubric Accuracy

Rubric Accuracy

10 nodes in this area of the perspective

This hub tracks accuracy assessments across the corpus, evaluating factual precision, named-entity integrity, and citation discipline. Technical documentation, including [collective-memory] (Assess_Accuracy_collective_memory) and [code-memory] (Assess_Accuracy_cm) assessments, emphasizes precise type signatures, SHACL schemas, and explicit correction logging. Strategic and GTM materials, such as the [GTM rubric] (RubricAssess_Accuracy_GTM) and [investor-facing assessments] (RubricAssess_Accuracy_1), are scrutinized for quantified metrics and alignment with product architecture. Decision records ([ADR-001] (RubricAssess_Accuracy_ADR001), [ADR-002] (RubricAssess_Accuracy_ADR002)) focus on cross-reference consistency and status clarity. The [IP-crisis entry] (Assessment_Accuracy_IPCrisis) and [customer architect call] (Assessment_Accuracy_CustomerArchitect) highlight the preservation of verbatim quotes and the explicit labeling of exploratory or uncommitted claims. Broad [foundational rubrics] (RubricAssess_Accuracy) validate internal consistency and entity naming, though they occasionally note the absence of external citations for high-level paradigm claims. Recurring markers of quality include verifiable incidents, consistent terminology, and the avoidance of false precision.

Witness phrases

verbatim
Precise type signatures; accurate code examples
verbatim
explicit 'uncommitted' status for exploratory claims
verbatim
rework rate 15% → <5%, board prep days → 15 minutes
verbatim
Factually accurate named entities (the enterprise customer, USPS, Österreichische Post, Poland Post)
verbatim
decision status clearly stated; no citations needed for internal decision record
verbatim
Some assessments identify a lack of external citations for strategic claims, while others deem them unnecessary for internal narratives.
verbatim
Tension exists between the demand for verifiable external data and the sufficiency of internal consistency for strategic documents.

In this area

RubricAssessment Assess_Accuracy_cm
Precise type signatures; accurate code examples; consistent terminology; references to prior memory (collective memory Data Model & API Architecture Discussion); SHACL validation schema included
RubricAssessment Assess_Accuracy_collective_memory
Precise technical specifications; explicit corrections documented ('Technical Corrections Made' section); consistent type definitions; accurate sequence diagrams; verifiable claims about system behavior
RubricAssessment Assessment_Accuracy_CustomerArchitect
Factually accurate named entities (the enterprise customer, USPS, Österreichische Post, Poland Post); precise pricing signal ([amount]-150K); concrete timeline (late Feb beta); no unsupported claims.
RubricAssessment Assessment_Accuracy_IPCrisis
Accurate references to ADR-002; explicit 'uncommitted' status for exploratory claims; preserved verbatim quotes; acknowledges open questions; no false precision
RubricAssessment RubricAssess_Accuracy
Factual claims (company start date, MVP status, six years of model experience) plausible and internally consistent. Brand name 'aswritten.ai' correctly styled. No citations provided for external claims, but none required for personal narrat
RubricAssessment RubricAssess_accuracy
Consistent brand name; clear claims; no citations needed for strategic document; factual accuracy maintained
RubricAssessment RubricAssess_Accuracy_1
Precise technical references (n8n, Apache Jena Riot, Outseta, Helicone, Open Router). Accurate product state ('missing piece', 'to do'). No falsifiable claims detected.
RubricAssessment RubricAssess_Accuracy_ADR001
Specific dates, dollar figures, and entity names (Capitol.ai, VCSA); decision status clearly stated; no citations needed for internal decision record; factual claims verifiable
RubricAssessment RubricAssess_Accuracy_ADR002
Named entities (Claude Code, Cursor, Windsurf) accurate. Cross-references to ADR-001, ADR-003, ADR-005, ADR-006 present. No unsupported claims; decision rationale explicit.
RubricAssessment RubricAssess_Accuracy_GTM
Specific metrics cited (days→15 min, weeks→days); references prior memories (ADR-002, role-based value props); no unsupported claims; could add external citations for 5.0.

Each node links from the entity index.