# Curator reproduction run 1

Run on 2026-09-09 with the exact Curator 2.0.0 Hermes profile archive.

## Lineup review

Newsletter Scout A and Newsletter Researcher B show likely, but not proven, overlap. They share a weekly cadence, source scope, subject, and three supplied story labels. The evidence is not conclusive because the sample covers one week, the labels may hide different stories, and Researcher B includes commentary that Scout A does not.

- **Newsletter Scout A — improve, or combine after testing (medium confidence).** Its stated job is a narrower version of Researcher B's. Preserve its instructions and history. Do not merge until a comparison test shows that it adds no distinct value.
- **Newsletter Researcher B — keep (high confidence on keeping; medium on distinctness).** It has the broader brief and adds a commentary dimension.
- **Invoice Helper — keep; effectiveness unknown.** It is an occasional, unrelated role with no supplied logs. Missing evidence is not poor performance.
- **Retire or add — neither is supported by this evidence.**

## Proposed comparison test

Run Scout A and Researcher B unchanged for the same calendar week against an identical candidate pool. Have a person tag every result by source, news versus commentary, and usefulness. Compare duplicate rate and unique useful items across three weeks. If duplication is consistently high and Scout A adds no unique useful result, propose a merger for approval. Otherwise clarify the division of labor and keep both.

No instructions were edited, no profile was merged or archived, and no schedule was created.