# Joinaroo

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Angelo Castigliola](https://cerebralvalley.ai/u/Sandn)
- **GitHub:** https://github.com/acastigliola/joinaroo
- **Demo video:** https://www.youtube.com/watch?v=Gi1A0Ddv7zY
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/510

Live in production: https://joinaroo.com

I built Joinaroo solo in Portland, Maine, and I am user #1. Live since October 30, 2026, built and run for ~$100 all-in. I built it to decide what is happening in Portland tonight. The Maine pilot covers Portland + Orono, with ~2,800 events in the catalog, gathered daily.

In Portland + Orono, other aggregators stop at the listing — title, time, ticket link; my pages add the people layer showing who is or was in the room. On Halloween night, Oct 31, 2026, the system placed 61 credited photos across 8 events. The Great Pumpkin Ball event page carries 26 photos from 5 identified people.

My photo-finding process scans public Instagram and Facebook posts — attendees, venues, professional photographers — for photos taken at each event, during and after it. I place those photos on the event page with a full credit stack: photographer name, studio/business, Instagram/Facebook links. People whose public photos were found appear checked in live right now by first name + last initial, with links to their profile, their post, and a Message action through Instagram/Facebook.

I shipped LIVE event cards where events happening now show real crowd photos and how many people are there. I shipped The Peephole, a strip on event pages showing tonight's other rooms in the same city through real crowd photos. Archived photos resurface on the next occurrence of a recurring act/event, when a user is deciding whether to go.

Muse Spark, via the Meta Model API, runs in production screening found photos, extracting photographer credits from captions, and stamping crowd photos every 15 minutes — 88 logged production runs, ~5.5M tokens, every run in a spend ledger, see Meta sections below.

Every aggregator has the same listings; I added the people layer — seeing who is or was at an event and messaging them — because that is what turns one-time visitors in Portland into regulars who meet people.

## More from Meta Global AI Developer Hackathon

- [FriendlyPear](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/507)
- [RAN Sentinel](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/508)
- [ashthejuan](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/509)
- [Shawn Ramey](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/511)
- [Camtoshop](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/512)
- [Smarties](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/513)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/510. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt