I played 1k hands of online poker and built a web app with Cursor AI (blog.rchase.com)

A hobbyist logged 1,000 online poker hands and used AI agents to build a full-featured web app (poker.rchase.com) that mirrors desktop tools like PokerTracker 4 — without writing a single line of code. Over a few days of iterative conversation with Cursor (using Claude 4.5 Sonnet), plus earlier help from Grok, he produced a Laravel MVP deployed to a DigitalOcean Debian 13 server with a private GitHub repo. The app includes a 700+ line PokerStars hand-history parser, poker-stat calculations (VPIP, PFR, 3‑Bet), multi-file/paste hand uploads, Gmail IMAP auto-import, periodic PokerStars exports and balance checks, an admin dashboard, journaling and bankroll CRUD, hand viewers, and profit/loss charts. Locally he used Herd for macOS. Technically significant because it’s a real-world example of an AI agent handling full-stack tasks: parsing messy domain-specific text, applying complex business rules (bet types, unmatched bets, split pots), debugging runtime errors from logs, and iterating UI changes in minutes. The story highlights strengths—dramatically faster feedback loops and scaffolding of robust automation—and limits: the agent needed domain teaching for poker edge cases and human product judgment for priorities, UX, and deployment decisions. For AI/ML practitioners this underscores how agent-driven development can accelerate MVP delivery while preserving the need for human oversight, testing, and product sense.