Wristband

Bay Area hackathons, demo days, VC and campus events — and what it takes to get in.

Public site: https://jasonzhang2047-art.github.io/wristband/

Private copy on claude.ai: https://claude.ai/artifact/K1UfkKARRu7rdTjzAeNYdo

How it works

crawler/crawl.mjs        every 6h  → site/events.json
  ├─ sources/luma.mjs        Luma map discovery + ~190 followed organizer calendars + ~160 followed hosts
  ├─ sources/yc.mjs          ycombinator.com/events + each event's application deadline
  ├─ sources/cerebral.mjs    Cerebral Valley events & hackathons (schema.org data)
  ├─ sources/mlh.mjs         MLH season list (Bay Area, in person)
  ├─ sources/devpost.mjs     Devpost hackathons, address-checked on each hackathon page
  ├─ sources/stanford.mjs    events.stanford.edu API (entrepreneurship, startups, hackathons…)
  ├─ sources/meetup.mjs      Meetup keyword searches
  ├─ sources/eventbrite.mjs  Eventbrite search pages (pay-to-pitch networking mills filtered out)
  └─ sources/crosscheck.mjs  compares against a public hackathon listing; anything missing is
                             re-fetched from its Luma page
data/curated.json        web & social sweep (people's posts, deadlines, recaps) — see crawler/SWEEP.md
data/learned.json        calendars and hosts the crawler discovered and will follow next run
crawler/watchlist.json   VCs, schools, labs, community hubs and people to flag (edit freely)
site/index.html          the page (Artifact format); site/events.json is its data

Each run classifies events (hackathon, pitch/demo day, VC-hosted, campus, founder, AI builder meetup, conference), drops off-topic and spam listings, merges the same event listed on several sites, keeps firstSeen history so new listings get a NEW tag, and learns new calendars/hosts from relevant events so coverage grows over time.

Run it yourself

node crawler/crawl.mjs

node crawler/build-static.mjs && python3 -m http.server 8787 --directory dist

Node 18+ with no dependencies.

Schedules

• GitHub Actions (.github/workflows/crawl.yml) runs the crawler every 6 hours and on every push to main, then deploys dist/ to GitHub Pages. It never commits: the previous crawl's events.json and learned.json are read back from the live site, so first-seen history and followed calendars carry over between runs.
• Claude desktop scheduled task wristband-crawl (every 6 hours, while the Claude app is open) runs the web & social sweep at most every 12 hours, commits data/curated.json and pushes, which starts a crawl on GitHub. It then copies the fresh events.json from the live site into the claude.ai page.