The events teens actually care about don't show up on Eventbrite or Instagram. A scraper pulls local Bay Area events, the Claude API sorts them and removes the duplicates, and a weekly email digest goes out to subscribers through Resend. GitHub Actions runs the whole thing on a schedule. I built it because the gap was pretty obvious once I noticed it.
Build something that automatically finds the local Bay Area events worth going to, the ones that never make it onto Eventbrite or Instagram, and send them to subscribers in a weekly digest.
- —Figured out the main problem: the events teens want (workshops, meetups, pop-ups, volunteer days) are spread across dozens of places and never collect anywhere useful
- —Built a scraper that pulls events from a bunch of Bay Area sources into one Supabase database
- —Connected the Claude API to sort each event by how relevant it is and to catch the same event when it shows up from different sources
- —Designed and built the weekly email digest that goes out to subscribers through Resend
- —Set up GitHub Actions cron jobs so the whole scrape, sort, dedupe, and send pipeline runs every week on its own
- —Built the landing and signup site in Next.js
- —How to build a pipeline that runs on a schedule without me. The hard part isn't any one step, it's making the whole chain reliable enough that I don't have to babysit it
- —How to use an LLM as a sorting and dedupe engine instead of a chatbot, which means writing prompts that give back the same clean, parseable format every single run
- —How email deliverability actually works: why services like Resend exist, and how your sender reputation decides whether the digest lands in the inbox or in spam
- —Why it's better to build for one specific group of people, Bay Area teens, than to try to make a general event app for everyone
The events that matter most to teens, like free workshops, hackathons, library programs, volunteer days, and local pop-ups, are exactly the ones that never show up on the big platforms. They're buried in random Instagram posts, city calendars, and word of mouth. I kept finding out about stuff I would have gone to only after it already happened, just because there was no single place that pulled it all together. Drift is my fix for that.
A scraper pulls events from a bunch of Bay Area sources into a Supabase database. Then the Claude API does the part that makes it actually usable: it rates each event for how relevant it is to teens and merges the same event when it came in from different places. A weekly digest of the best ones goes out through Resend, and the whole pipeline runs on a GitHub Actions schedule so I don't have to touch it.
The scrape, sort, and digest pipeline is built and running every week. Right now I'm adding more sources and tuning the classifier so the digest stays short and worth reading as the number of events grows.