Fatebound

Inspiration

Fatebound: Where Words Decide is inspired by the emergent storytelling of Death by AI and the classic tension of Wheel of Fortune. Gemini generates original fantasy scenarios that place players in dangerous, often morally difficult situations. To earn a chance at solving the word puzzle, players must first survive these AI-driven encounters by making smart, creative choices. Every scenario is different, every decision carries consequences, and once the puzzle begins, each guessed letter can shift the balance between survival and defeat.

What it does

Fatebound: Where Words Decide is a multiplayer fantasy party game that combines AI-generated storytelling with classic word-guessing mechanics. Each round begins with Gemini generating a unique, high-stakes adventure scenario that places the players in danger, such as moral dilemmas, environmental threats, or encounters with strange creatures. Players must describe how they respond to the situation, and the game evaluates those choices to determine who survives the encounter.

Surviving the scenario earns players the right to participate in the puzzle phase, where they work through a Wheel-of-Fortune-style word puzzle by guessing letters. Correct guesses reveal parts of the phrase and award points, while failed scenarios cost lives. As the game progresses, tension increases as players balance risky decisions, dwindling lives, and strategic letter choices.

Because Gemini generates both the scenarios and evaluates player actions, no two games are ever the same. The experience blends improvisational storytelling, player creativity, and puzzle-solving into a fast-paced game where survival is never guaranteed and the words themselves ultimately decide your fate.

How we built it

Fatebound was built using Next.js for the game logic, API routes, and real-time state management. Gemini 3 Flash Preview powers the dynamic storytelling, generating unique adventure scenarios and evaluating player choices so that each round feels unpredictable and alive. GitHub Copilot was used throughout development to accelerate iteration, refactor complex logic, and explore different approaches to AI integration.

The game was tested extensively with friends and family, who helped uncover edge cases, balance difficulty, and refine the overall flow. Their feedback played a key role in shaping the pacing, clarity of scenarios, and fairness of outcomes, ensuring the experience feels fun, understandable, and replayable.

Challenges we ran into

One of the biggest challenges was getting Gemini to consistently generate unique, engaging scenarios without repeating themes or falling into predictable patterns. Balancing creativity with reliability required careful tuning of temperature, token limits, and prompt structure to prevent truncation while still encouraging variety.

Another major challenge was managing AI output constraints, especially ensuring structured responses for game logic while working within strict token budgets. This involved building robust validation, repair, and fallback systems to keep the game running smoothly even when AI responses were imperfect.

Finally, creating a fair and enjoyable multiplayer experience added complexity. Player state, turn order, elimination rules, and pacing all had to work seamlessly together. Ensuring the game felt balanced, responsive, and fun across different group sizes required multiple iterations and real-world testing.

Accomplishments that we're proud of

Successfully built a fully functional multiplayer experience with synchronized game state, turn-based play, and clear win and elimination conditions. Players can seamlessly move between AI-driven scenarios and puzzle rounds without interruptions or inconsistencies.

Implemented multiple difficulty levels that meaningfully change the complexity of puzzles and challenges, making the game accessible to casual players while still offering depth and challenge for more experienced groups. These systems work together to keep the game engaging, replayable, and adaptable to different play styles and group dynamics.

What we learned

Building Fatebound provided hands-on experience with using AI as an active part of game logic rather than just a content generator. Designing prompts and tuning temperature and token limits revealed how small changes can dramatically affect creativity, reliability, and structure in AI-generated text.

This project was also a first step into using AI to both generate narrative content and evaluate player input in meaningful ways. Learning how to safely parse, validate, and respond to AI output taught valuable lessons about error handling, fallback systems, and designing around uncertainty. Overall, it deepened my understanding of how AI can drive interactive systems where player choices directly influence outcomes.

What's next for Fatebound

The game has already proven itself as a fun experience among friends and testers, and it will continue to be something I can bring out for casual game nights and shared play sessions. Their enthusiasm confirmed that the core idea works and that the blend of AI storytelling and puzzle solving is genuinely engaging.

Looking ahead, the ideal next step would be for Fatebound to win the hackathon and evolve into a more polished, publicly available game. With additional refinement, balancing, and user feedback, it has the potential to be expanded for wider audiences and long-term replayability.

Built With

• gemini3
• mongodb
• next.js
• typescript