I built derby.app — a tool to help research and handicap the Kentucky Derby using AI.

In the spring of 2023, I had some time between jobs and got interested in trying to pick the winner of the Kentucky Derby. I went all in: watched replays, read bloodline charts, followed forums, and built spreadsheets. I landed on Mage, and when he won, it was one of the most fun moments I’ve had at a sporting event. I did the same in 2024 and landed on Mystik Dan.

This year, I wanted to offload some of that manual work. So I built a research agent that collects and summarizes everything I’d normally spend hours doing.

Derby.app pulls together:

- Race footage and prep summaries - Pedigree and breeding data - Expert picks and forum chatter - Real-time odds and track conditions - Contextual weather info - A GPT-4o-powered research assistant you can chat with

How it’s built:

- Next.js with AI SDK on Vercel, Postgres with Prisma, hosted on Neon - LLM: GPT-4o and Perplexity Sonar Deep Research - Data: A mix of race history, expert analysis, and live-scraped info - UI: Designed for mobile and desktop, chat assistant is context-aware

Would love feedback on what’s unclear, broken, or could be more helpful.

Thanks for checking it out, and if you're betting tomorrow, good luck!