The part of the hackathon we got wrong
Jan 2026 / 3 min
StudyQuest was a 24-hour build: an AI study planner with spaced repetition and gamification, on React Native with Expo, the Gemini API, and Zustand for state. Four of us, and only one had done a hackathon before. We picked React Native even though our shared background was React on the web, so we knew going in it would cost us time on unfamiliar problems. I still think choosing it on purpose beat defaulting to what we already knew, and I'd defend the time we lost to it as tuition.
The mistake that actually mattered was cheaper to avoid, and I made it anyway. We didn't read the challenge track closely enough, and the project drifted from the brief. Teams that stayed tight to the track scored better with less interesting builds than ours. That's the real lesson of a judged hackathon, and it's an uncomfortable one for anyone who likes building: the score rewards fit to the brief before it rewards the strongest thing you can make.
What I held was the order of work, the timeline, and the conversations with the judges and with other teams. The others used AI freely on their own parts and I kept the shape of the thing. If I ran it again I'd spend the first hour re-reading the track out loud with the whole team before a line of code, because that hour was the cheapest leverage we had and we skipped it.