Nicolas Peters Saa · Hackathon · Boston AI Week 2026

Test Flight: what I learned without winning

In September 2026 I took part in Test Flight, the Glasswing Ventures Enterprise-Ready AI Hackathon in Boston, a two-day event. We built and demoed a working product. We didn't win, and the reasons why taught me more than the build itself.

Event
Glasswing Ventures
Dates
Sept 26–27, 2026
Duration
Two days
Theme
Enterprise-ready AI
Result
Shipped, didn't place

What we built

Phision: Pharma Quality Intelligence. When something goes wrong in a medicine factory, an AI agent rebuilds what happened across the plant's systems, stress-tests the suspected root cause, and gives QA an investigation where every claim is backed by cited evidence. Work that normally takes days takes minutes.

My part was the company itself: a fictional generics manufacturer with its own records, quality events and a planted root cause for the agent to find.

Try the demo (runs in your browser on the fictional case data) · Original hackathon deployment

What I learned

  1. Bring your own idea. Teams that arrived with a problem they already understood spent the weekend building. Picking an idea on the day eats the hours you need to make it work and make it convincing.
  2. Come with your own team. Being matched with strangers works, and I met great people. But in 36 hours, learning how each other works pulls against the task. A team that already knows its roles starts shipping on hour one.
  3. A random team is still worth it. I'd do it again for the people alone. Next time I'll do it with an idea and a team I've already lined up.

The team