Building WhoseHouseBurned.com: A Friday Night Hackathon During LA Fires

When the Palisades and Altadena fires shut down Los Angeles, I was stuck inside and inspired to build. With the help of Cursor's AI agents, I built WhoseHouseBurned.com - a tracker of celebrity homes affected by the January 2025 fires.

The Build Process

Armed with Cursor's AI agents and a product spec, I set out to build each feature one by one. Creating a branch and Pull Request for each feature allowed me to isolate features and continue to build while Vercel was deploying the previous changes.

One challenge with Cursor Agents is that you're limited to a single agent at a time. While having multiple agents would be great, the bottleneck is around ideas. While the agent is working, I'm unable to provide additional context on this task or other tasks. Devin.ai solves this but is lacking in other areas.

The Vercel Deploy Limits

The flow of Cursor Agent to Vercel preview is OP, but can be intoxicating. About 2 hours into development, I hit Vercel's free deployment limits. I was able to work around this by using Cursor's AI agents to deploy to a local server, but the speed slowed for 24 hours.

Keeping It Current

The site maintains its relevance through a multi-pronged data collection approach:

1. Manual updates from Instagram and Twitter monitoring
2. Automated daily ChatGPT queries that:
   • Check existing site data
   • Scan for new information across the internet
   • Identify any differences

Technical Stack

• Next.js for the frontend
• Vercel for hosting
• ChatGPT for automated data verification
• Cursor AI agents for development assistance