The strongest public story is not that we have a large corpus. It is that we are building a disciplined knowledge-engineering system: evidence has provenance, uncertainty is named, synthetic material is controlled, and mistakes can be corrected.
CKES (CHIKARA Knowledge Engineering Standard) is the public methodology blueprint for corpus engineering, dataset engineering, evidence engineering and knowledge governance. It is proposed architecture — not a claim of completed certification.
Corpus lifecycle
Scope → acquisition → rights → validation → release → correction → retirement.
Evidence engineering
Provenance, conflict, freshness and uncertainty — not silent selection of a preferred claim.
Synthetic data controls
Grounded generation, adversarial testing, human review. Model consensus is not truth.
Transparency
Corpus Cards, Dataset Cards, quality gates and an explicit list of open decisions.