Solo-built live golf wagering, in production.

The challenge

Most senior marketers talk about being AI-native. Few have shipped production software solo. I wanted to close that gap — and chose a problem hard enough that the proof would be unambiguous.

BetLoopr is a full-stack golf tournament management and live wagering platform: real-time scoring across 9+ game formats, live bet resolution over WebSockets, GPS course awareness, photo-based scorecard verification, and mobile-native iOS and Android builds. A real product solving a real problem in a category I care about.

The work

I built it solo — start to finish — using Claude Code and Lovable as primary development tools, deploying on Supabase and Capacitor. The technical scope:

• 9+ game formats — stroke, match play, Stableford, Skins, Wolf, Nassau, team formats — each with its own scoring engine, handicap allocation, and settlement logic
• Real-time leaderboards and live bet resolution over WebSockets, with full lifecycle handling (initiate → accept / decline → auto-resolve → ties / pushes)
• Side games running in parallel with the main format — Snake, Skins, Birdies, Greenies — each with independent state
• Press betting system with location-based and photo-based verification (Google Vision OCR for scorecard scanning)
• Offline-first mobile with sync, GPS course awareness, camera integration, and haptic feedback
• 67+ database migrations in a production-tracked codebase — not a prototype, an operating system
• 7 serverless Edge Functions running with Row-Level Security on Supabase

Stack: React 18 + TypeScript, Supabase (Postgres + RLS + Auth + Edge Functions), Capacitor 7 for iOS / Android, Mapbox for course visualization, Google Vision OCR for scorecard scanning, Claude Code + Lovable for sustained AI-assisted development across an 8-month-plus build window.

The outcome

BetLoopr is live at betloopr.com — not a demo, not a side project, a production app with real users.

The point of including this in a marketing portfolio: most senior candidates' AI-native claims live in slideware. Mine lives in a deployed codebase. The same AI-native operating mode applied to client engagements is how Craft Batch ships — workflows, agents, analytics tooling, content pipelines built directly inside client businesses rather than handed off as recommendations.

When a client asks whether I can actually build with AI: this is the answer.