X - Protocol

A consent-first proximity social mesh that maps nearby Bluetooth devices in real-time and uses AI to send handwritten invitation letters to strangers around you.

Comment

What Inspired X

I was sitting in a packed lecture hall, 200 people around me, every one of them on a device, and zero of us connected to each other. We were surrounded by presence but completely invisible to one another.

That moment made me ask: the internet knows every click I've ever made, but it has no idea who I'm physically near right now. GPS gives coordinates. Social media maps followers. But nobody has mapped physical proximity as a human experience. That gap is where X was born.

What X Is

X - General Intelligence is a consent-first proximity social mesh.

It passively senses nearby Bluetooth devices like, phones, laptops, earbuds, displays and visualizes them on a live 2D star map. X sits at the center. Every detected device is a glowing node orbiting by RSSI signal distance.

When X notices someone nearby, it uses GPT-4o-mini to draft a short, personalized AI invitation letter tailored to the device type and how close they are. The visitor receives it by scanning a QR code, no app, no login, just a chat portal. X mirrors the live conversation in real time.

How We Built It

  • BLE scanning via bleak (Python) - passive RSSI-based proximity detection
  • FastAPI + asyncio backend - WebSocket-driven real-time device state
  • React + Canvas frontend - live constellation star map with signal-strength orbits
  • GPT-4o-mini - generates contextual, device-aware invitation letters
  • Google Gemini API - powers the constellation intelligence layer, analyzing proximity patterns and generating contextual social insights for the Footprint panel. Gemini's multimodal understanding helps X interpret who is around, not just what devices are nearby.
  • qrcode[pil] - live QR portal generation per session
  • Cloudflare Tunnel - zero-config public HTTPS deployment at xprotocol.us
  • Mono/3D canvas rendering with RSSI-to-distance visual mapping

Challenges We Faced

Getting Bluetooth RSSI values to feel spatially accurate was the hardest part, signal strength varies wildly by environment and device shielding. I built a smoothing and classification layer to map raw dBm into human-readable proximity zones (next to you, in room, far away).

WebSocket frame batching at 40ms flush intervals was critical to keep the star map smooth without flooding React re-renders. The proxy tunneling through Cloudflare required careful Vite config to forward WS and API routes correctly.

What I Learned

Physical presence is data that the internet completely ignores. BLE gives us a new primitive, not location, not identity, but nearness. X proves that nearness alone is enough to start a human conversation.

Built With

Share this project:

Updates