diff --git a/ROADMAP.md b/ROADMAP.md
new file mode 100644
index 0000000..3c57bd5
--- /dev/null
+++ b/ROADMAP.md
@@ -0,0 +1,92 @@
+# Vynl Feature Roadmap
+
+Ordered by impact + ease of implementation.
+
+## Tier 1 — Big Wins, Easy to Build
+
+### 1. "More Like This" Button
+- Button on every recommendation card
+- Uses that single song as seed for a new discovery
+- Just pre-fills the Discover query with "find songs similar to [artist] - [title]"
+- **Effort: 30 min** — frontend-only, no backend changes
+
+### 2. "Why Do I Like This?" Reverse Analysis
+- User pastes a song, AI explains what draws them to it
+- Then finds more songs with those same qualities
+- New mode on Discover page or standalone input
+- **Effort: 1 hour** — new prompt template, reuse existing UI
+
+### 3. Mood Scanner
+- Simple mood sliders: happy/sad, energetic/chill
+- Injects mood context into the recommendation prompt
+- "I'm feeling chill and melancholy — find me music for that"
+- **Effort: 1-2 hours** — UI sliders + prompt modification
+
+### 4. Share Discoveries
+- "Share" button generates a public link like deepcutsai.com/share/abc123
+- Shows recommendations without requiring login
+- Great for viral growth on social media
+- **Effort: 2-3 hours** — new endpoint + public page
+
+### 5. "Surprise Me" Mode
+- One button, no input needed
+- AI picks a random obscure angle from user's taste profile
+- "You seem to like songs in D minor with fingerpicked guitar — here's a rabbit hole"
+- **Effort: 1-2 hours** — new discovery mode, creative prompt
+
+## Tier 2 — Medium Effort, High Value
+
+### 6. Playlist Generator
+- Generate a full 20-30 song playlist around a theme
+- "Road trip playlist", "Sunday morning cooking", "90s nostalgia"
+- Save as a playlist in the app, export as text
+- **Effort: 3-4 hours** — new endpoint, playlist creation, UI
+
+### 7. Daily Digest Email
+- Automated daily/weekly email with 5 personalized recommendations
+- Brings users back without them having to open the app
+- Requires Resend integration (already planned)
+- **Effort: 4-5 hours** — Resend setup, email template, Celery scheduled task
+
+### 8. Artist Deep Dive
+- Click any artist name to get an AI card
+- Shows: why they matter, influences, best album to start with, similar artists
+- Modal or dedicated page
+- **Effort: 3-4 hours** — new endpoint + UI component
+
+### 9. Song Preview Embeds
+- Inline YouTube Music player in recommendation cards
+- Users can listen without leaving the app
+- YouTube iframe embed with the video ID
+- **Effort: 2-3 hours** — need to fetch YouTube video IDs (ytmusicapi), embed iframe
+
+### 10. Music Timeline
+- Visual timeline of when recommended/liked artists released music
+- Shows patterns in user's taste ("you love 2008-2012 indie")
+- Uses MusicBrainz release dates
+- **Effort: 4-5 hours** — data fetching + timeline visualization
+
+## Tier 3 — Bigger Builds, Major Differentiators
+
+### 11. Collaborative Discovery / Taste Compatibility
+- Two users compare taste profiles
+- Compatibility percentage score
+- Shared "you'd both love this" recommendations
+- **Effort: 6-8 hours** — friend system, profile comparison logic, shared rec generation
+
+### 12. Vinyl Crate Simulator
+- Gamified discovery: swipe through 50 random albums
+- Tinder-style UI: swipe right to save, left to pass
+- Rapid feedback builds taste profile faster
+- **Effort: 8-10 hours** — swipe UI, album data source, feedback loop
+
+### 13. Concert Finder
+- After recommending an artist, check if they're touring nearby
+- Link to tickets (Songkick API or Bandsintown API)
+- Potential affiliate revenue
+- **Effort: 6-8 hours** — external API integration, location input, UI
+
+### 14. "Surprise Me" Advanced — Rabbit Holes
+- Multi-step guided journey: "Start here → then try this → go deeper"
+- Each step builds on the last, taking the user on a curated path
+- **Effort: 8-10 hours** — multi-step UI, state management, chained prompts