Back to ANDR3A.OSwor2Duel
A fast-paced word game created as a thesis experiment to investigate human-AI collaboration in web development.
Overview
wor2Duel is a short-session word game built around alternating turns, rapid decisions and a visible score. Its development also served as the real-world test environment for my Bachelor's thesis on AI-assisted web development.
Game Design
Players build words one letter at a time during a timed match. Valid words, round progression and the final score create a compact gameplay loop designed around pressure and immediate feedback.
Architecture
The application uses React and Vite for the frontend, with game state separated into manageable UI and gameplay responsibilities. The word-selection logic is designed around an in-memory dictionary and prefix-based lookup.
Human + AI
The project was used to compare AI-assisted development approaches involving ChatGPT, GitHub Copilot and Tabnine. The experiment focused not only on whether AI could generate working code, but also on quality, integration, reliability and the amount of human intervention required.
Implementation
The interface was implemented with React, Vite, JavaScript, HTML and CSS, with particular attention to responsive behaviour, keyboard interaction, game-state feedback and readable round progression.
Challenges
The main challenge was balancing a fast game loop with enough visual feedback to keep the player oriented. The project also demonstrated that AI-generated code still requires human evaluation, debugging and architectural decisions.
Result
A playable web game that became more than a software project: it provided a concrete environment for studying the changing relationship between developers and AI coding assistants.
