24 Hours of FSDC: Building Under Pressure
At 3 AM, twelve hours into my first hackathon, I was debugging a Supabase Row Level Security error I couldn't parse. I didn't know what RLS was six hours earlier. My teammates had given up and were face-down on their laptops.
At 3 AM, twelve hours into my first hackathon, I was debugging a Supabase Row Level Security error I couldn't parse. I didn't know what RLS was six hours earlier. My teammates had given up and were face-down on their laptops.
I'd signed up for FSDC with two friends, mostly for fun. No serious expectations about winning. My entire development experience at that point was Python and JavaScript. Discord bots, scrapers. I'd never built a web app, never touched a frontend framework, never dealt with database-level authentication.
We decided to build a medical report aggregation platform. Users upload their reports, the system compiles everything into one PDF with an AI-generated summary, hosts it securely, and gives them a shareable link for doctors. We planned it out, looked at each other, and agreed: how hard could this possibly be.
V0 made it possible. It did the heavy lifting on everything. Frontend, backend structure, the Gemini API integration for summaries. We scaffolded the whole thing in Next.js, plugged in Supabase for auth and storage, and wired up a straightforward Gemini API call. A standard AI web app on paper.
The authentication nearly killed us.
Supabase auth flows, RLS policies, session management, permissions. Things that matter when you're shipping a real product but feel like pointless torture during a 24-hour hackathon where all you need is a working demo. We burned hours on edge cases that wouldn't have mattered for a prototype. By the time the login flow actually worked, it was 3 AM and we hadn't started on the core feature of our entire product.
The AI summary sounded simple. Take uploaded reports, compile into a PDF, send to Gemini, get a summary back. File size limits broke it first. Then rate limits. Then formatting came back mangled. Each fix surfaced the next problem, and we were debugging on zero sleep with caffeine that had stopped working around midnight.
We got it functional at 8 AM. Thirty minutes before the demo. No sleep, no prep, a product held together with hope.
Judges went after us hard for building everything in V0. No "real" backend, they said. We were too tired and too frustrated to push back. Just left. Went home. Crashed.
I woke up six hours later to five missed calls.
The organizers had been trying to reach us. We'd qualified for the finals. Every call, missed because we were unconscious. By the time we saw the messages, the response window had closed. Disqualified for not picking up the phone.
That still gets me. We walked out of that demo room convinced the judges hated our project, and they'd actually put us through. There's a lesson in there about sleep deprivation and snap judgments. Mostly I just wish I'd left my ringer on.
The repo is a mess and I won't pretend otherwise. My friend posted a demo video on LinkedIn if you want to see the product in action.
V0 made the impossible happen. Three people with zero web development experience built a working medical report platform in 24 hours. We didn't win. Didn't even get to compete in the finals. But I walked away knowing I could build something real under pressure, which was more than I expected going in.
Links: Demo here