Personalized Autoplay 2.0: Feedback-Learning, Smart Skip, and Better Continuity

Personalized Autoplay 2.0 Is Live

We just shipped a major upgrade to Personalized Autoplay. This release focuses on one core goal: make recommendations adapt faster to each listener while keeping playback stable and uninterrupted.

What changed

1) Skip now affects recommendation weight

When a user skips a Personalized track, Beatra records that as a negative signal for recommendation scoring. Similar artists and adjacent candidates are weighted down over time.

This improves two things immediately:

• Fewer repeated misses from the same recommendation cluster
• Faster convergence toward the listener's actual taste

2) New Like and Dislike feedback buttons for cold-start users

If Beatra detects that it does not know the listener well enough yet, it now shows two feedback buttons directly on Personalized tracks:

• Like
• Dislike

Behavior:

• Like: increases the recommendation weight for similar artists and candidates
• Dislike: instantly skips the track and applies a stronger negative weight

This solves cold-start much faster than passive listening alone.

3) Personalized autoplay now uses a rolling queue

Autoplay no longer relies on single-track fetch behavior. It now maintains a rolling queue model:

• Target queue size: 5 tracks
• Queue is refilled proactively as it drops
• Recommendation logic is re-triggered continuously during playback

Result: skip remains responsive, and the next candidate is always ready.

4) Better continuity for autoplay and 24/7 mode

We improved continuity behavior for reconnect and restart scenarios:

• Playback state is persisted
• Current track and queue are restored when possible
• Autoplay and 24/7 mode settings are preserved

This gives a much more normal-play style recovery after transient failures.

5) Recommendation engine now uses explicit feedback scoring

Recommendation ranking is no longer based only on listening duration and affinity. It now also uses explicit user feedback signals:

• Like boosts score
• Dislike and skip reduce score
• Artist-level weighting is updated dynamically

In short, the engine now learns from both what you play and what you reject.

---

Personalized Autoplay will continue improving as more feedback arrives. If you use the new buttons, recommendation quality should improve noticeably faster in your first sessions.