A read-only scoring readiness layer that defines future evidence scoring dimensions, source type baselines, claim-level scoring rules, and capsule-readiness boundaries before live scoring is enabled.
Every TheoB pathway can move through Past, Present, and Future without losing context.
Read current signals, conditions, and live context.
Voice ready
Discovery Source Scoring Readiness defines how TheoB will eventually score sources: authority, primary-source status, citations, freshness, relevance, conflict, contamination, multimodal interpretability, and capsule readiness.
Source scoring readiness is active as a non-destructive scoring policy layer. TheoB can define how future sources will be scored, but it cannot score live provider results, reject sources automatically, ingest Vault records, or compress capsules yet.
Score whether a source is institutionally, scientifically, legally, or historically authoritative for the claim being evaluated.
Authority is contextual. A source can be authoritative in one domain and weak in another.weight: 14Prefer original data, official records, direct research, first-party documentation, and primary evidence over reposts or summaries.
Primary-source status must be labeled, not assumed.weight: 14Score whether the source provides clear references, links, citations, datasets, DOI trails, or supporting evidence.
Claims without trails should not receive high confidence.weight: 10Track when the source was published and when TheoB retrieved it, especially for current or changing topics.
Old sources can still be valuable, but freshness must match the query.weight: 8Score whether the author, publisher, institution, or source owner is identifiable and relevant.
Anonymous or unclear authorship should reduce trust unless the source is otherwise verifiable.weight: 8Score how directly the source supports the exact claim, question, entity, region, or timeframe.
Do not cite sources that only vaguely relate to the claim.weight: 10Detect whether many results are copies of the same original source or independently corroborating sources.
Repetition is not verification.weight: 6Score whether the source agrees, partially conflicts, or strongly conflicts with other credible sources.
Conflict should be surfaced, not hidden.weight: 8Detect affiliate content, content farms, thin SEO pages, promotional claims, and low-evidence commercial pages.
Useful commercial sources can exist, but incentives must be visible.weight: 6Score whether the source explains how data, measurements, surveys, experiments, or conclusions were produced.
Opaque methods reduce confidence.weight: 6Score whether images, diagrams, CAD, maps, schematics, or datasets can be interpreted with enough clarity for structured use.
Visual meaning must not be flattened into weak summaries.weight: 5Score whether the source can later become a TheoB Intelligence Capsule without losing source trail, uncertainty, or context.
Compression must reduce size, not truth.weight: 5statistics, public records, climate, agriculture, economic, civic, and infrastructure datasets
Check methodology, reporting year, regional definitions, and revisions.scientific claims, citations, research trails, experimental findings, medical or agriculture research
Separate peer-reviewed, preprint, abstract-only, and commentary sources.entity context, definitions, relationships, historical orientation, and citation discovery
Use as context, then verify important claims with primary or specialist sources.broad discovery, comparison, source finding, and contradiction detection
Aggressively filter SEO, duplication, weak authorship, and commercial motives.current developments, event timelines, public statements, and cross-outlet comparison
Speed is not accuracy. Preserve publisher, timestamp, and update history.cacao, climate, agriculture, supply chain, ceremony, sustainability, and specialized source streams
Separate cultural, scientific, commercial, and ceremonial claims respectfully.images, PDFs, CAD, maps, diagrams, datasets, architecture, engineering and schematic interpretation
Require file validation, rights awareness, redaction, and uncertainty zones.A score helps rank evidence quality, but it does not magically make a source true.
Scores must remain explainable and revisable.A source can be strong for one claim and weak for another.
Avoid giving one blanket score to an entire domain.Primary evidence should outrank summaries when the question requires factual grounding.
Summaries can guide discovery but should not replace proof.When credible sources disagree, TheoB must preserve the disagreement and show why.
Do not hide uncertainty to sound more confident.This layer defines scoring readiness only and does not score live provider results.
No live provider calls or Vault writes.Bad-looking sources should be flagged before they are automatically rejected.
Avoid throwing away minority or emerging evidence too early.Images, CAD, schematics, maps, and datasets need specialized scoring criteria.
Visual evidence should not be judged with text-only rules.A source cannot be capsule-ready unless its claim trail, uncertainty, and retrieval context survive compression.
A capsule must preserve enough truth to be reawakened faithfully.