Ponio

Ponio is a Spotify-only, voice-first private AI DJ. It is actively being built around playback reliability, recommendation quality, conversational transparency, and durable taste memory.

When someone says “play something for a rainy afternoon,” the system must understand the scene, find fitting music, play on the right device, maintain continuity, and explain failures clearly.

An AI DJ earns trust not by sounding conversational, but by causing the right musical action after the user speaks.

The listener describes a scene, similarity, or energy direction. Ponio interprets it, plans a short set, executes Spotify playback, shows what happened, and adjusts from follow-up feedback.

The system separates what benefits from judgment from what must remain reliable.

• The LLM extracts musical direction; the selection engine owns song choice.
• Play, pause, skip, and volume call Spotify directly.
• The DJ asks only when the listener's taste contains a genuine fork.

Real use exposed text-without-playback failures, repeated artists, opaque device state, and unsafe memory capture. The product evolved toward one routing gate, one song-selection owner, conservative session flow, and stricter durable-memory rules.

Current evidence comes from a working Spotify product path, issues observed in real use, continued regression testing, and architecture changes tied to specific failures. A demo video will be added later.

An AI product cannot hide uncertainty behind a fluent reply. Judgment, execution, and feedback need clear ownership so the user can understand what the system actually did.