BookmarX

Every bookmark represents a moment of curiosity. A thought worth returning to. I built BookmarX because we were frustrated watching our digital collections become cluttered, disorganized wastelands.

Comment

BookmarX

Inspiration

"Your bookmarks are digital graveyards." We've all been there - hundreds of saved links collecting dust, knowing there's gold somewhere but finding it feels impossible. BookmarX was born from the frustration of watching valuable content disappear into the void of our bookmark folders, never to be seen again.

What it does

BookmarX transforms your chaotic X/Twitter bookmarks into an intelligent, searchable knowledge graph. It automatically analyzes, categorizes, and connects your saved content - including videos, threads, and media - so you can finally make sense of what you've been saving. Understanding the universe starts with understanding yours.

How we built it

  • Backend: Node.js/Express with Prisma ORM, BullMQ for job queues handling bookmark syncing, media downloads, and AI analysis
  • Frontend: React with Framer Motion for fluid animations, Zustand for state management
  • AI Pipeline: xAI (Grok) for content analysis, entity extraction, and semantic scoring; OpenAI (SRRY xAI didnt have transcribe :/) for video transcription and analysis
  • Infrastructure: PostgreSQL for data persistence, worker-based architecture for processing bookmarks asynchronously
  • OAuth Integration: Secure X/Twitter authentication for bookmark access

Challenges we ran into

  • NSFW Content: Filtering and appropriately handling sensitive content during analysis without breaking the pipeline
  • Context Drift: Ensuring AI analysis stays relevant to the actual content rather than hallucinating connections
  • Organization at Scale: Building a categorization system that works for 10 bookmarks or 10,000
  • Order Preservation: Maintaining chronological and relevance-based ordering while enabling smart grouping

Accomplishments that we're proud of

  • Built a complete bookmark-to-insight pipeline with video analysis capabilities
  • Created a visual graph system showing connections between saved content
  • Implemented "magnet search" - finding related content across your entire bookmark history
  • Real-time notification system keeping users informed of processing status

What we learned

The real value isn't in saving content - it's in surfacing it at the right moment. We learned to balance AI automation with user control, and that good UX means making complex operations feel effortless.

What's next for BookmarX

  • Cross-platform bookmark aggregation (beyond X)
  • Collaborative knowledge sharing between users
  • AI-powered "rediscovery" - proactively surfacing relevant old bookmarks based on current interests
  • Export to popular note-taking tools (Notion, Obsidian)

Built With

Share this project:

Updates