Zombie Road Rampage Project Report
1. The Concept
Zombie Road Rampage is a browser-based 2D/2.5D physics driving game inspired by the flow of side-scrolling zombie vehicle games. The player drives through dangerous roads, crushes zombies, earns Scrap, upgrades vehicles, and retries runs to travel farther and unlock harder levels. I chose this project for a one-day sprint because it was a good test of how quickly AI-assisted development could handle physics, progression, responsive UI, animated enemies, vehicle upgrades, and game polish in a single web project.
2. The Stack
The Stack: HTML5, Vanilla CSS, JavaScript, HTML5 Canvas, Matter.js, Web Audio API, and localStorage.
The AI Assist: ChatGPT handled the initial game analysis, gameplay flow, level structure, physics requirements, UI/UX planning, asset requirements, and technical handoff prompts. Codex/Cursor was then used for the coding and iteration phase, translating the plans into the working game and repeatedly improving the physics, responsive layout, Garage, vehicles, zombies, levels, and gameplay systems.
3. The 1 Day Sprint:
Morning (Hours 1–3): We started by analyzing the reference game and defining the core loop: drive, destroy zombies, earn Scrap, upgrade the vehicle, retry, travel farther, and finish the level. We then planned the game states, Garage, vehicle progression, zombie types, fuel, boost, physics, camera, and responsive landscape behavior. The first playable version was created quickly, but it still looked and felt like a physics prototype.
Afternoon (Hours 4–6): Most of the development time went into improving gameplay quality. The biggest focus was vehicle physics: wheel alignment, suspension, acceleration, air rotation, terrain contact, camera movement, and smooth rendering. We also improved zombie walking, running, hit reactions, ragdoll behavior, vehicle artwork, transparent WebP assets, the Garage carousel, Home screen, Loading screen, level selection, settings, and result screens. The game was expanded to 20 levels with different areas, obstacles, hazards, and zombie combinations.
The Finish Line (Hours 7–8): The final hours focused on QA and mobile landscape support. We adjusted button sizes, font sizes, panel spacing, safe areas, and responsive layouts for short-height phones. We also refined Home animations, popup presentation, level endings, obstacle variety, and performance. The final pass concentrated on keeping gameplay smooth while preserving the existing game logic and progression.
4. The Roadblocks
Roadblocks: The biggest roadblock was getting the gameplay to feel smooth instead of simply functional. AI-generated physics often produced vehicles that felt weak, floaty, unstable, or visually disconnected from their wheels. Wheel pivots, suspension, chassis alignment, driver animation, zombie movement, and terrain seams required repeated tuning. Another challenge was that broad prompts caused Cursor/Codex to spend too much time changing unrelated systems, so the workflow had to shift toward smaller, tightly scoped tasks with clear acceptance tests. Responsive landscape layouts were also difficult because desktop-sized buttons, panels, and text did not scale well to short mobile screens.
5. Key Takeaways
What was the part that worked the best? The planning and iteration structure. Breaking the game into clear systems—vehicle physics, zombies, Garage, levels, UI, responsive behavior, and QA—made it much easier to identify what needed improvement. Detailed acceptance checks also helped prevent later prompts from breaking already working systems.
What was the part that worked the worst? Gameplay feel and large AI coding requests. A feature could technically work while still feeling bad to play. Broad instructions such as “make everything premium” or “fix the full game” often caused unnecessary changes and long AI runs. The best results came from fixing one problem at a time, such as one vehicle, one wheel system, one popup, or one responsive layout.
My final thought: If I had one more day, I would spend most of it on gameplay tuning rather than adding more content. I would further refine vehicle suspension, wheel contact, acceleration, camera smoothing, zombie reactions, obstacle pacing, and mobile performance across all 20 levels. If I rebuilt the project, I would lock the core physics and responsive layout much earlier, then expand the levels and visual content only after the first level felt completely smooth and polished.