Demos you can inspect.
Nothing on this page is a mockup — every surface responds. Start with the EyeBrowse operator admin: switch tabs and roles, then ask the Ask AI tab anything and watch it stream. The broadcast and architecture demos below explain how the system stays coherent.
The dashboard gets the first look.
This is the operator admin Ben built for EyeBrowse — designed to run a multi-location chain from one surface. Switch tabs to land on different operating pages. Switch roles to see access filtering. Ask anything in the Ask AI tab and watch the response stream.
Editorial weekly briefing for the super admin. Hero MRR, 12-week trajectory annotated with launches, three operational signals, practice constellation, leaderboard. The page is a recurring artifact — the same shape ships every Monday.
Built for desktop: 26 operating pages, role-based access, and a ⌘K command palette. Open it full-screen to explore the real thing.
Open full-screen adminAsk the admin anything in the Ask AI tab — real Claude Sonnet, streams in real time.
Move your cursor — every visitor on this page sees it instantly. This is the same Supabase Realtime infrastructure EyeBrowse uses to keep a chain of stores on the same inventory, prices, and promotions. No stale data, no “sorry, that frame is actually sold out.”
What This Architecture Powers
When a frame sells in one store, every kiosk and website across the chain reflects it within 200ms — built to keep multi-location inventory coherent. No overselling, no apologetic phone calls.
Store managers never hit refresh. Try-on analytics, inventory levels, and conversion rates stream to their screen the moment data changes.
A franchise owner updates pricing from one dashboard. Every kiosk, every storefront display, every admin panel reflects the change simultaneously.
Multiple staff editing the same order queue see who is working on what in real time — the same presence pattern behind Google Docs cursors and Figma avatars.
Six components, one pipeline. A customer walks into a store, looks at a kiosk, and the system handles everything from face detection to payment processing in under a second.
Hover any component to see what it does for the business.
Browser-native webcam capture — no app or extension
Runs natively in the browser — zero app downloads, zero friction. Customers try on eyewear from any product page without leaving the retailer's site.
468 facial landmarks detected in <16ms
Google's ML framework maps 468 facial landmarks per frame at 60fps. This precision is what makes virtual try-on feel real — glasses sit on your face, not float near it.
Serverless compute for AI orchestration
Runs auth, validation, and AI orchestration close to users so the browser stays fast without a dedicated app server.
Claude API + Gemini for product recommendations
Routes recommendation requests across providers with fallback so style guidance keeps working even when one model is slow.
WebSocket pub/sub for live updates
Built to sync inventory in real time across every store location. A product marked 'sold' in-store disappears from the try-on experience within 200ms.
Real-time inventory and analytics
Store managers see live try-on analytics — which frames are trending, conversion rates, and inventory levels updating in real time without page refreshes.
Browser-native webcam capture — no app or extension
Runs natively in the browser — zero app downloads, zero friction. Customers try on eyewear from any product page without leaving the retailer's site.
468 facial landmarks detected in <16ms
Google's ML framework maps 468 facial landmarks per frame at 60fps. This precision is what makes virtual try-on feel real — glasses sit on your face, not float near it.
Serverless compute for AI orchestration
Runs auth, validation, and AI orchestration close to users so the browser stays fast without a dedicated app server.
Claude API + Gemini for product recommendations
Routes recommendation requests across providers with fallback so style guidance keeps working even when one model is slow.
WebSocket pub/sub for live updates
Built to sync inventory in real time across every store location. A product marked 'sold' in-store disappears from the try-on experience within 200ms.
Real-time inventory and analytics
Store managers see live try-on analytics — which frames are trending, conversion rates, and inventory levels updating in real time without page refreshes.
Browser-native webcam capture — no app or extension
Runs natively in the browser — zero app downloads, zero friction. Customers try on eyewear from any product page without leaving the retailer's site.
468 facial landmarks detected in <16ms
Google's ML framework maps 468 facial landmarks per frame at 60fps. This precision is what makes virtual try-on feel real — glasses sit on your face, not float near it.
Serverless compute for AI orchestration
Runs auth, validation, and AI orchestration close to users so the browser stays fast without a dedicated app server.
Claude API + Gemini for product recommendations
Routes recommendation requests across providers with fallback so style guidance keeps working even when one model is slow.
WebSocket pub/sub for live updates
Built to sync inventory in real time across every store location. A product marked 'sold' in-store disappears from the try-on experience within 200ms.
Real-time inventory and analytics
Store managers see live try-on analytics — which frames are trending, conversion rates, and inventory levels updating in real time without page refreshes.