AI and context
Subject What a model writes under kapi's governance, and whether the governance reached it. Evidence Scores estimate performance on the stated corpus. Model-judged scores also depend on the judge's validated agreement with people. Each card reports corpus size to make the sample limitations visible. Gates CI No. Results are tracked over time; regressions require investigation.
3 measured4 partial2 unvalidated
What governs content at the point it sits: the coordinates, the voice profile, the terms, and the content memory holding what was already approved.Rests on the engine putting the right content in front of the governance.
The call itself: what goes into a prompt, how many blocks travel together, and how the text was split before any of that.Rests on the context layer having resolved what applies before a prompt is built.
The checks that read finished content and report what breaks the rules, and whether they are right about it.Rests on every layer below, since a check reports on what they produced.