What problem does it solve?
In an age where AI is everywhere but most people don't really understand it, OCTOBLUFF turns the "Can you tell if it's AI?" question into a shared experience that gets people talking, laughing, and learning together.
Who is it for?
- Friend groups looking for something more interesting than another Zoom call
- Students wanting to understand AI capabilities in a hands-on way
- Workplaces doing team building (way more fun than trust falls)
- Families bridging the tech knowledge gap between generations
- Gaming communities who want social games that spark actual conversations
How does it bring people together?
Real interaction, not just gameplay:
- You actually have to talk to each other and pay attention to how people communicate
- Creates hilarious moments when someone realises they've been arguing with an AI for 5 minutes
- Works equally well in-person or remote (doesn't require everyone in same room)
Lowers barriers:
- No gaming experience needed - if you can have a conversation, you can play
- Non-competitive option (everyone can just try to fool each other for fun)
- Educational without being a lecture - you learn what AI can/can't do by playing
Builds community skills:
- Active listening (you have to actually pay attention to spot patterns)
- Critical thinking (analysing conversation styles)
- Tech literacy (understanding AI capabilities through experience)
Why it matters:
Most people interact with AI daily (ChatGPT, customer service bots, social media) but don't really understand how it works. OCTOBLUFF makes AI tangible and understandable through play. You stop seeing AI as magic or a threat and start seeing it as a tool you can recognise and work with.
How we built it
Frontend: React + Vite with Tailwind CSS for the UI. Socket.io client handles real-time communication. We built a responsive game interface with player grids, chat system, voting mechanics, and custom SVG octopus avatars.
Backend: Node.js + Express with Socket.io for WebSocket connections. Integrated the Gemini API to generate unique AI personalities and conversation responses.
Voice Synthesis: ElevenLabs text-to-speech is used for ALL players (both human and AI). This levels the playing field - everyone has the same voice, so the only way to detect AI is through writing style and conversational patterns, not vocal cues.
Real-time Sync: Everything from chat messages to player shuffling to vote timers is synchronised across all clients via Socket.io events.
Challenges we ran into
AI believability: Tuning Gemini prompts so the AI sounds human enough to be challenging, since too obvious = boring.
ElevenLabs API rate limits: Managing voice generation requests for multiple AI agents without hitting API limits during active gameplay.
Real-time synchronisation: Making sure all 8 players see the same shuffled player order and colour assignments simultaneously was tricky. We had to carefully manage socket events and state updates.
Accomplishments that we're proud of
Polished UI in limited time: The cyberpunk aesthetic with gradient effects, glowing borders, and smooth animations came together better than expected for a hackathon timeline.
Built a complete full-stack multiplayer game from scratch in a hackathon timeframe - frontend, backend, real-time networking, API integrations, and game logic all working together.
Log in or sign up for Devpost to join the conversation.