Rank by User Fit

Workflow

Per candidate pool:
- [ ] Step 1: Score each on 5 dimensions (intuition_density_fit, goal_alignment, dedup_penalty, source_reliability, freshness)
- [ ] Step 2: Weighted sum → rank
- [ ] Step 3: Top items with score > threshold (default 30), max 10 → keep list
- [ ] Step 4: Remaining → drop list; per-drop one-line reason using worst-scoring dimension
- [ ] Step 5: Enforce minimum: ≥2-3 drops visible even in slow weeks (ranker transparency)

Scoring rubric