ScrollSage: Decode Your Feed

🌱 Inspiration

We’re constantly scrolling—but rarely stopping to ask: why does this content speak to me? Who benefits from this narrative? What identity is my feed shaping?

ScrollSage was inspired by this curiosity. We wanted to create a tool that reveals the invisible architecture of our social media bubbles: the values, ideologies, influencers, and brands that drive them. Think of it as cultural x-ray vision for your media taste.

🚀 What it does

ScrollSage lets users upload a few pieces of media they love—like influencers, brands, or shows—and turns that into a personalized cultural analysis.

• It maps your taste profile using Qloo’s Taste Graph.
• It uses an LLM (Gemini) to explain the underlying narratives, ideologies, and marketing patterns in your media bubble.
• It suggests opposite-bubble content to expand your perspective.
• It highlights how media consumption shapes identity—for better or worse.

Each insight is delivered as a fluorescent, animated, glassy card, making abstract cultural ideas feel fun, visual, and intuitive.

🛠️ How we built it

• Frontend: HTML + TailwindCSS for a calm, Gen-Z, glassmorphism look with animated insight cards.
• Backend: Flask handles the API calls and flow logic, no database involved.
• Taste Engine: We use Qloo's /search and /v2/insights endpoints to map user inputs (like “Zendaya” or “Glossier”) into structured cultural entities.

• LLM (Gemini): We feed Qloo results into a structured prompt that Gemini uses to:

  • Decode shared ideologies and marketing patterns
  • Reveal cultural identities
  • Suggest counter-content

• Deployment: Lightweight, Docker-ready stack built for fast deployment with just an API key and Flask server.

🧗 Challenges we ran into

• Structuring Qloo API calls correctly: getting valid entity_ids, filter.type, and valid signal combinations took some trial and error.
• Getting Gemini to return nuanced, critical insights rather than general summaries required careful prompt engineering.
• Designing the UI to be both aesthetic and informative—we didn’t want it to look like a dashboard, but like a mirror.
• Working without a database meant we had to design clean, ephemeral flows powered purely by user input and real-time APIs.

🏆 Accomplishments that we're proud of

• Built a fully functioning, database-free cultural analysis engine in a short sprint.
• Created a visually engaging UX that makes deep cultural insight feel accessible and scrollable.
• Successfully integrated Qloo + Gemini + Tailwind + Flask into a cohesive pipeline.
• Designed a novel use case: decoding your media bubble like a sociologist, marketer, and culture critic—all in one app.

📚 What we learned

• Media taste can be analyzed as structured cultural data, not just vibes.
• Every feed reflects an ideology, and some are shaped more by corporate interests than personal growth.
• LLMs + taste graphs = powerful tools for identity analysis and media literacy.
• Design matters—deep insight becomes usable when it feels intuitive, soft, and scrollable.

🔮 What's next for ScrollSage: Decode Your Feed

• 🔄 Let users contrast their taste graph with friends for “bubble comparison”.
• 🎯 Add interactive story interpretation: “How would you vs. someone from a different bubble read this news?”
• 📱 Build a browser extension that analyzes your actual Instagram/TikTok history in real-time.
• 🤖 Use feedback loops to improve Gemini's contrast and bias detection capabilities.
• 🌐 Deploy a live version with login-free access, powered only by API keys and local sessions.

Built With

• and
• bias-analysis
• contrast
• gen-z-inspired-ui-with-glassmorphism-and-animation-html-+-js-?-structure-and-interactions-qloo-api-?-for-building-the-taste-graph-and-cultural-insights-gemini-(llm)-?-for-narrative-decoding
• perspective
• python-?-for-backend-logic-and-api-integration-flask-?-lightweight-web-server-and-routing-tailwind-css-?-beautiful