Seeker

Note: This project is planned: design and architecture are underway but development has not started yet.

Project Overview: Seeker is a podcast player that rethinks how listeners manage what they’re hearing. Traditional apps force every show into a single now-playing queue—switch from a work podcast to something casual and you lose your spot, your up-next list gets shuffled, and you’re left digging through history to pick up where you left off. Seeker solves this with multi-queue playback: each queue is an independent lane with its own now-playing, progress, and up-next list, so jumping between unrelated shows is instant and non-destructive.

The Problem

Every major podcast player—Apple Podcasts, Spotify, Overcast, Pocket Casts—uses a single-queue model. That works fine when you listen to one show at a time, but real listening habits are messier:

• A commuter queues up industry news for the drive, then switches to a comedy show at the gym.
• A student lines up lecture recaps for study sessions but wants a true-crime binge on weekends.
• Anyone who taps play on a new episode mid-queue watches their carefully ordered list get replaced or reordered.

In each case the listener has to manually re-find, re-order, or re-start episodes because the app assumes one queue is enough.

Core Concepts

Multi-Queue Playback

• Create named queues (e.g. “Work”, “Wind Down”, “Road Trip”).
• Each queue maintains its own now-playing episode, playback position, and up-next list.
• Switching queues instantly resumes from where you left off—no searching, no re-ordering.
• Queues can be pinned, archived, or set to auto-populate from subscriptions.

Agentic Capabilities

Seeker layers AI on top of the listening experience:

• Natural-Language Discovery — describe what you’re in the mood for (“something about the history of bridges” or “a funny interview with a chef”) and Seeker surfaces matching episodes across the catalog.
• Queue Playground — an interactive sandbox where an agent can suggest, re-order, or remix your queue based on mood, topic, or time budget (“build me a 45-minute commute queue on AI news”).
• Smart Summaries — get episode digests before you commit to listening, powered by transcript analysis.
• Cross-Queue Insights — surface connections between episodes across different queues (“the guest on your Work queue also appeared in this Wind Down episode”).

Planned Features

1. Queue Management — create, rename, reorder, pin, and archive independent playback queues.
2. Seamless Queue Switching — one tap to jump between queues with full state preservation.
3. Subscription Routing — assign shows to default queues so new episodes land in the right lane automatically.
4. Natural-Language Search — find shows and episodes by describing what you want, not just by title or keyword.
5. Queue Playground — let the AI agent build, remix, or optimize a queue for a given context.
6. Episode Summaries — AI-generated digests from transcripts so you can decide what’s worth your time.
7. Playback Sync — seamless handoff across devices with per-queue state.
8. Import & Migration — bring your subscriptions and history from Apple Podcasts, Spotify, and OPML feeds.

Technology Stack (Planned)

• Frontend: React Native (iOS & Android), Expo
• Backend: Node.js, Hono
• Database: Postgres (Supabase), Vector Store (Supabase)
• AI Integration: OpenAI, Vercel AI SDK
• Audio: Expo AV / custom playback engine

Outcome

In the planning phase—architecture and design work is in progress.