decision: close the first-value evidence loop

[tt-a1i]

Parent

#255

What to build

Close the first-value evidence loop with one explicit product decision grounded in the release, packaged first-use journey, and complete pilot evidence: retain the current design, isolate one first-use defect, or open a separate specification for a repeated preference gap.

Acceptance criteria

• The decision cites every prior evidence frontier and its limitations.
• The outcome is classified as exactly one of: retain current design, isolated first-use defect, or repeated preference gap.
• Recommendation disagreement alone does not trigger routing, ranking, embedding, or model changes.
• An isolated defect produces only a bounded follow-up specification or ticket; a repeated preference gap produces a separate policy specification; retaining the design closes the work without speculative features.
• The conclusion states that token savings, labor savings, model quality, and universal Core-versus-On-demand superiority remain unproven.
• The completion receipt distinguishes every ticket, validation gate, external effect, and remaining authority boundary.

Blocked by

• research: run the independent Roster recommendation pilot #261