Magic Lists already performs candidate preparation before LLM ranking. One additional low-cost signal that may improve playlist quality is album track position.
Rationale:
- In many studio albums, earlier tracks are more likely to be singles, opening statements, or otherwise prominent tracks.
- This would not need to be a hard rule, only a light weighting signal.
- It seems especially appropriate for borderline ranking decisions, not as a dominant factor.
Potential implementation:
- Capture
track_number (and optionally disc_number) in the artist track fetch path.
- Pass it through to the candidate scoring layer.
- Apply a small deterministic bias favoring earlier album positions.
- Optionally also expose it to the LLM payload for secondary use.
Important constraint:
- The weighting should remain weak. Many strong tracks occur later in albums, especially in metal/hard rock catalogs, so this should be a tie-breaker or mild prior, not an aggressive penalty against later tracks.
Why it may fit Magic Lists well:
- It complements the existing pre-filter/scoring architecture.
- It should be backward compatible.
- It does not require any workflow change from users.
- It may improve perceived “This Is:” playlist correctness without adding much complexity.
Magic Lists already performs candidate preparation before LLM ranking. One additional low-cost signal that may improve playlist quality is album track position.
Rationale:
Potential implementation:
track_number(and optionallydisc_number) in the artist track fetch path.Important constraint:
Why it may fit Magic Lists well: