Genesis PreviewE1 admittedSemantic Snapshot implemented500 Atlas identitiesQuery Lab: liveRead-onlyNo arbitrary URLsHosted CI: startup blocked

Live read-only Genesis baseline plus protected E4.7 agent, ontology and model candidates. Public MCP, frontier-model execution and TNOE training are not deployed; no arbitrary URLs.

TWIRX Reference implementation and public service of the Typed Web Commons

Ontology Workbench

Make semantic judgment explicit, reviewable, and reversible.

The Workbench collapses repeated record-level claims into unique Semantic Mapping Units, separates relation from admission, records signed human decisions, and exports eligible training evidence. Today all fifteen units remain pending.

Volume is not semantic diversity

747,958 repeated record-level mapping claims collapse into only 15 unique Semantic Mapping Units. TWIRX reports both numbers because they answer different questions.

Corpus impact

How many packets and frames would a mapping decision affect?

Semantic diversity

How many distinct native-field-to-concept judgments actually require review?

Current semantic review state

Source families
2
Unique mapping units
15
Signed decisions
0
Reviewed hard negatives
0
Diversity audit
Native schemas2
Native fields15
Candidate semantic concepts12
Candidate frame roles14
Mapping relationunresolved
Admission statuspending

“Unresolved” describes the semantic relation. “Pending” describes human admission. They are deliberately separate dimensions.

The Workbench produces review evidence, not automatic canon

  1. Explore native structureSchema, field, path, type, source family, sample values and neighboring fields.
  2. Inspect candidate meaningConcept, frame role, examples, counterexamples and affected packet/frame counts.
  3. Choose an actionAccept, reject, defer or contextual—with a separate exact, close, broad, narrow, contextual or unresolved relation.
  4. Record reasons and contextReason codes and explicit conditions prevent a local equivalence from becoming universal.
  5. Sign and versionAn external review key binds reviewer identity, time, prior decision and semantic diff.
  6. Export eligible training evidenceOnly verified signed decisions enter the training ledger.

Competency includes knowing what cannot be concluded

Opportunity

6 positive and negative cases pass. Personalized eligibility, current-open state and ranking stay unresolved when evidence is insufficient.

World State

6 cases pass. Causality, cross-series comparability and missing methodology are not inferred from a numeric observation.

Total competency failures: 0. Passing authored constraints does not approve any pending mapping unit.

What models may and may not do

May propose

  • Concept candidates
  • Frame roles
  • Potential relations
  • Ambiguity and abstention
  • High-impact review ordering

May not establish

  • Mapping admission
  • Source or publisher authority
  • Identity equivalence
  • Policy permission
  • Canonical ontology state

See the TNOE roadmap Workbench report