No posts yet.
Reel-to-Second-Brain 1. Core Idea Build a personal AI-powered “Reel-to-Second-Brain” system that turns Instagram Reels into useful, structured knowledge instead of letting saved Reels disappear into an endless collection. The user simply pastes an Instagram Reel URL into a mobile-friendly interface. The system automatically watches/analyses the Reel, extracts the useful information, verifies important details, and saves it into a connected knowledge system. The goal is: «Don't just save content. Turn it into something you can use later.» --- 2. Example Use Case — NYC Trip Imagine I am planning a trip to New York City. I see an Instagram Reel showing a café in NYC. Normally: «Save Reel → Forget about it → Trip happens → Never visit café.» With this system: «Paste Reel → AI understands it → Identifies café + NYC → Saves information → Connects it to NYC → Eventually becomes part of my NYC itinerary.» For example: Instagram Reel ↓ "Café XYZ in SoHo, NYC" ↓ [[Cafe XYZ]] ↓ [[SoHo]] ↓ [[New York City]] ↓ [[NYC Trip]] The café becomes part of my accumulated NYC knowledge. --- 3. User Experience The mobile UI should be extremely simple. Home Screen ┌─────────────────────────┐ │ My Second Brain │ │ │ │ Paste Instagram Reel │ │ ┌─────────────────────┐ │ │ │ instagram.com/reel/ │ │ │ └─────────────────────┘ │ │ │ │ [ SAVE REEL ] │ │ │ │ Recent │ │ • NYC Café │ │ • Tokyo Restaurant │ │ • Photography Tip │ └─────────────────────────┘ The user should not need to manually categorize anything. The AI handles classification. --- 4. Processing Pipeline User ↓ Mobile UI ↓ n8n Webhook ↓ Instagram Reel Extraction ↓ Video + Audio + Caption + Metadata ↓ AI Video Analysis ↓ Structured Information ↓ Web / Location Verification ↓ Database ↓ Obsidian ↓ Knowledge Graph ↓ Trips / Projects / Itineraries --- 5. AI Analysis The AI should analyse both what is being said and what is visually shown. It should extract information such as: General - Topic - Summary - Important points - Recommendations - Products - People - Places - Events - Activities - Concepts - Useful facts Travel-specific - Country - City - Neighborhood - Place/business name - Address - Category - Why it is recommended - Things to do - Food/drinks mentioned - Estimated cost, if stated - Best time to visit, if stated - Original Reel URL The AI should return structured JSON rather than directly generating Markdown. Example: { "content_type": "place", "name": "Cafe XYZ", "city": "New York City", "neighborhood": "SoHo", "category": "cafe", "summary": "...", "why_saved": "...", "source": "Instagram", "source_url": "...", "confidence": 0.91 } --- 6. Verification Layer AI-generated information should not automatically be treated as fact. Important information should be verified using external sources. For a business/place: - Name - Address - City - Current existence - Website - Opening hours - Location - Category This prevents incorrect information from entering the knowledge base. --- 7. Knowledge Structure Obsidian should act as the human-readable knowledge layer. Possible structure: Second Brain/ │ ├── Cities/ │ ├── New York City.md │ ├── Tokyo.md │ └── London.md │ ├── Places/ │ ├── Cafe XYZ.md │ ├── Restaurant ABC.md │ └── The Met.md │ ├── Trips/ │ └── NYC Trip.md │ ├── Reels/ │ ├── Reel - Cafe XYZ.md │ └── Reel - Restaurant ABC.md │ ├── Topics/ │ ├── Photography.md │ └── Fitness.md │ └── Projects/ --- 8. Obsidian Linking Every entity should be connected using Obsidian Wikilinks. Example: # Cafe XYZ Located in [[New York City]]. Neighborhood: [[SoHo]] Category: [[Cafes]] Saved from: [[Reel - Cafe XYZ]] Related trip: [[NYC Trip]] This allows Obsidian's graph to naturally reveal relationships. Example: NYC Trip │ New York City / | \ SoHo Brooklyn Midtown │ Cafe XYZ │ Instagram Reel --- 9. Don't Make Every Reel an Isolated Note A major design principle: «The system should continuously enrich existing knowledge rather than creating disconnected notes.» If I save 20 NYC Reels, the system should recognize that they belong to the same destination. Instead of: Reel 1 Reel 2 Reel 3 Reel 4 ... it should build: New York City │ ├── Cafés ├── Restaurants ├── Attractions ├── Shopping ├── Activities ├── Neighborhoods └── Saved Reels --- 10. Itinerary Generation The itinerary should be generated from the accumulated knowledge. Example request: «"I'm going to NYC for 5 days."» The AI can use the existing NYC knowledge base to generate: # NYC — 5 Day Itinerary ## Day 1 — SoHo + Downtown Morning - Cafe XYZ - Walk around SoHo Afternoon - ... Evening - ... ## Day 2 — Midtown ... ## Day 3 — Brooklyn ... The system should consider: - Geographic proximity - Opening hours - User preferences - Number of saved places - Trip duration - Previously visited places - Priority - Time required - Travel time - User's available days --- 11. Important: Separate “Saved” From “Planned” A Reel being saved does not automatically mean the user definitely wants to visit/do the thing. Each item can have a state: Saved ↓ Interesting ↓ Want to Visit ↓ Added to Itinerary ↓ Visited ↓ Reviewed This keeps the knowledge base clean. --- 12. Feedback Loop After visiting a place, the user can update it: Cafe XYZ ⭐ Rating: 4/5 Status: Visited Notes: Great coffee. Would visit again. Tags: #coffee #soho #nyc This information can later influence future itinerary generation. For example: «Prioritize places similar to places I previously rated highly.» --- 13. Recommended Architecture Frontend A lightweight mobile-first web app/PWA. Automation n8n Responsible for: - Receiving Reel URLs - Calling APIs - Processing data - Running AI workflows - Searching/verifying information - Updating database - Generating Obsidian files Reel Extraction Instagram Reel scraper/API. AI Gemini Used for: - Video understanding - Audio understanding - Transcription - Visual analysis - Information extraction - Summarization - Classification - Itinerary generation Database Supabase/PostgreSQL Used as the structured source of truth. Potential entities: Users Reels Places Cities Countries Trips Categories Topics Tags Visits Recommendations Knowledge Interface Obsidian Used for: - Markdown notes - Human editing - Wikilinks - Knowledge graph - Personal notes - Long-term knowledge management --- 14. Long-Term Vision The ultimate goal is to create a personal contextual memory system. Instead of saving random pieces of content, everything I consume can become connected to something I care about. For example: My Second Brain │ ┌─────────────────┼─────────────────┐ │ │ │ Travel Learning Projects │ │ │ NYC Design YouTube │ │ │ Cafés Books Ideas │ │ │ Reels Notes Research A Reel is no longer just a Reel. It becomes a piece of knowledge connected to a larger context. --- 15. Core Philosophy Traditional See → Save → Forget This system See ↓ Capture ↓ Understand ↓ Structure ↓ Connect ↓ Remember ↓ Use The product isn't really an Instagram Reel saver. It is an: «AI-powered system that converts fleeting internet content into persistent, connected personal knowledge.» --- 16. Possible Future Features Universal Content Capture Eventually support: - Instagram Reels - YouTube Shorts - YouTube videos - TikTok - Websites - Articles - Tweets/posts - PDFs - Screenshots - Voice notes All enter the same knowledge system. Natural Language Search Ask: «"What cafés have I saved in NYC?"» «"Show me everything I've saved about Japan."» «"What places did I save that are near Central Park?"» «"What did I save about cameras?"» AI Planning Ask: «"Plan my NYC trip using everything I've saved."» «"Give me a 3-day itinerary with the restaurants I've saved."» «"Remove places that are too far apart."» «"Make this itinerary cheaper."» Knowledge Discovery The AI could proactively identify connections: «"You've saved 7 Reels about SoHo. Would you like me to create a SoHo guide?"» This transforms the system from a storage tool into an active personal knowledge assistant.
No posts yet.