Configurable rules
Board size, win length, and dimensionality can be adjusted so each match can be tuned from fast and familiar to complex and strategic.
A configurable 2D and 3D strategy game that expands tic-tac-toe into a deeper spatial system, built as an AI-native development workflow, with execution across design, gameplay, and technical systems.
Players can move from classic grid logic into larger boards, longer win conditions, gravity-enabled play, and real-time multiplayer. The result is a polished strategy experience that feels approachable on first play and progressively more tactical as the board opens up.
About the game
3D Connect-K starts from familiar tic-tac-toe rules and stretches them into a configurable connect-style experience. Players can choose compact or large boards, define how many pieces are needed to win, and switch between flat and full 3D play depending on how strategic or readable they want the board to feel.
The design goal was to make the game accessible without flattening its depth. That meant pairing intuitive setup options with richer mechanics like gravity mode, isometric viewing, customizable player identity, and multiple ways to play, so the game can feel casual at first and much more tactical over repeated sessions.
Board size, win length, and dimensionality can be adjusted so each match can be tuned from fast and familiar to complex and strategic.
Supports local multiplayer, computer opponents with multiple difficulty levels, and online multiplayer through room codes.
Players can use full 3D presentation or a more approachable isometric view to make larger boards easier to interpret.
Technical details
Built with React 19, TypeScript, and Vite to keep iteration fast while preserving a strong typed foundation for gameplay logic and UI state.
Three.js and React Three Fiber power the board visualization, piece placement, and perspective changes that make the 3D mode legible and engaging.
Zustand is used to separate gameplay state from player and configuration preferences, which keeps mode switching and rule changes manageable.
A reusable win-checking and board-line generation system allows the same rules engine to support both 2D and 3D boards across multiple sizes and win conditions.
AI behavior uses difficulty-based logic, and heavier decision work is pushed off the main interaction path to keep the interface responsive.
The project was built through natural-language prompting in Cursor, orchestrating implementation and iteration through Claude and ChatGPT as part of the delivery process.