Quest: Open Questions as Ranked Artifacts (b307ad54-a95).
Acceptance:
- Tournament loop: select pairs of open_question artifacts in same field_tag, ask domain expert persona (e.g., Hongkui for connectomics, Ed Lein for SEA-AD/AD, Claire Gustavson for immunology) which is more important to answer.
- Update importance_elo with K=24 (standard), separate Elo per field_tag.
- Schedule: nightly batch picking ~50 pairs per active field, prioritizing pairs where Elo gap is small (highest information gain).
- Audit: every pairwise judgment logged with rationale → artifact_comments on both questions for percolation.
- Pause logic: stop within a field if all questions converged (low rating volatility).
- Spec: docs/planning/specs/open_question_elo_tournament.md.
Why now: blank Elo isn't useful — the tournament is what makes per-field rankings meaningful.