Hackathon MVP: Gemini-Guess with Gemini 3 AI

Introducing Gemini-Guess: A Multi-Turn AI Reasoning Game

I’m excited to share the MVP of my hackathon project, Gemini-Guess

You can try it live here: https://gemini-guess.qz-l.com/


it is my first time to join a public hackathon, Gemini 3 Hackathon

https://devpost.com/software/gemini-guess

What is Gemini-Guess?

Gemini-Guess is an open-ended game where players challenge Gemini 3 to guess any idea, object, or concept. Players can provide clues via text, voice, or video, and Gemini 3 responds with deep, multi-turn reasoning, showing its thought process at every step.

Inspiration

Inspired by the timeless fun of guessing games, I wanted to see how far AI can go in reasoning about open-ended ideas. Gemini 3’s ability to follow complex, multi-step reasoning made it the perfect AI for this challenge.

What it does

  • Players think of a concept, object, or idea — or say/show it.
  • Gemini 3 attempts to guess it using multi-turn reasoning.
  • The game tracks rounds, AI guesses, and allows replaying the AI’s thought process.
  • Supports text, voice, and video input for natural interaction.

How we built it

  • Next.js for the frontend web app — fast and hackathon-friendly.
  • Gemini 3 API powers the reasoning and multi-turn guessing.
  • Web APIs handle microphone and camera input for voice/video clues.
  • Game state stored in memory for simplicity; ready for leaderboard tracking.
  • Replays rendered visually and with audio playback for judges and players.

Challenges we ran into

  • Designing prompts that work for unlimited concepts.
  • Handling voice and video input smoothly without latency.
  • Keeping the game fun while allowing AI to sometimes fail in interesting ways.
  • Managing multi-turn reasoning so it can be replayed concisely.

Accomplishments

  • Showcased Gemini 3’s depth and nuance in reasoning.
  • Created a replayable game visualizing the AI’s thought process.
  • Enabled text, voice, and video inputs in a single interactive app.
  • Made it accessible — fun for anyone, not just kids.

What we learned

  • Gemini 3 handles deep, multi-turn reasoning across open-ended challenges.
  • Multimodal input (voice/video) makes AI interaction feel natural.
  • Prompt design is both an art and a science.
  • Watching AI think is almost as fun as playing the game!

Next steps for Gemini-Guess

  • Add leaderboards and global scoring.
  • Expand voice/video AI interaction, possibly with live-stream challenges.
  • Optimize prompts for faster AI guesses without losing nuance.
  • Create a mobile-friendly version using React Native for on-the-go play.

Try it yourself!

Experience Gemini 3’s reasoning firsthand: https://gemini-guess.qz-l.com/

I can’t wait to see what concepts you challenge Gemini with — it’s surprising, fun, and sometimes hilarious!

❤️ Support This Blog


If this post helped you, you can support my writing with a small donation. Thank you for reading.


Comments

Popular posts from this blog

fixed: embedded-redis: Unable to run on macOS Sonoma

Copying MDC Context Map in Web Clients: A Comprehensive Guide

Reset user password for your own Ghost blog