Sustainability · Community
The exact words I gave Claude to kick this off.
Research companies that take household recycling or waste. Check what types of products they take, if there is a cost to recycle, if they offer discounts or any rewards, and what they do with the waste. As a reference, Kitchen Warehouse does this with old pans and they offer a 10% discount for a Tefal pan. Put this in a table in Google Sheets so that it is easy to reference.
How it works, why it was built this way, and the thinking behind each decision.
You type in any item — an old frying pan, dead batteries, worn-out sneakers — and the app finds which Australian companies will take it back, what it costs, whether you get a reward, and exactly where to drop it off. It started as a spreadsheet, became a web app, and grew into a community-maintained directory that anyone can add to.
It's a React app (the interactive webpage) that talks to Claude's brain (the AI API) to understand your search and match it to the right programs. Program data lives in a shared Upstash Redis database so every visitor sees the same up-to-date list. Three small serverless functions handle the AI search, URL scraping, and email notifications — all running on Vercel without any traditional server.
Natural language search means users don't need to know the right category or keyword — "old mascara" just works. A shared database (rather than browser storage) means submissions from any user are immediately visible to everyone. The URL scraper lets the database grow quickly without manual data entry. And keeping the API key server-side via Vercel functions means it's never exposed to the public.
The code lives in a GitHub repository. Every time a file is updated on GitHub, Vercel automatically rebuilds and redeploys the site within about 60 seconds — no manual steps needed. The Anthropic API key, Google OAuth credentials, Resend email key, and database connection are all stored as encrypted environment variables in Vercel, never in the code itself.
Where this project could go with more time and feedback.
Feedback, ideas, and questions are welcome — I read everything.