mylab

I Used AI to Code a Game in Unity — Here’s What I Learned (arquivo)

I Used AI to Code a Game in Unity — Here’s What I Learned, por thuy le

”
This case study is a record of my experience designing “Hedgietown,” a Unity game for my MDes capstone project. During a two-day hackathon, I “vibe-coded” a minigame — Hedgie’s Easter Egg Hunt — using only Claude to write code, despite having no programming or game design experience. The term “vibe coding” was coined by OpenAI co-founder Andrej Karpathy in February 2025, describing AI-assisted programming where developers “fully give in to the vibes, embrace exponentials, and forget that the code even exists” (The Week, 2025). As Karpathy put it, “It’s not really coding — I just see stuff, say stuff, run stuff and copy paste stuff, and it mostly works” (Nielsen, 2025). For designers, this shift means we “can now try lots of things quickly, rather than having to hand ideas off to software engineers to create prototypes, one by one” (NPR, cited in The Week, 2025). If you can communicate your ideas clearly, you can now code them.

This approach aligned perfectly with my goal to challenge myself outside my comfort zone of product design and apply UI/UX principles in an entirely new context. How did I go from concept to functional prototype using vibe coding alone? And what does this mean for the future of design? This case study explores both the practical journey from idea to playable prototype and the broader implications of AI-assisted design for creative professionals.

Background: The Evolving Role of Design
Traditionally, the design-to-development process follows a clear handoff model: designers create mockups, wireframes, and prototypes, then pass these blueprints to engineers for implementation. This collaborative approach has defined the industry for decades.

AI-assisted coding is fundamentally changing this dynamic. Now, designers can take their blueprints directly to AI tools like Claude, which can guide them through the implementation process. For designers willing to embrace a beginner’s mindset and learn how engineers think, this changes the game.

This capability doesn’t necessarily make designers into one-person teams, but it does provide the majority of what’s needed to create an MVP and gather user feedback. That’s exactly what I did — building a functional game in Unity without prior coding or game design experience.

Methodology: Research Approach and Tools
Research Questions
How do mobile tap-to-move mechanics compare to traditional desktop controls in player experience?
What are the practical limitations of “vibe coding” for game development?
I used AI in three distinct ways throughout the development process:

Natural Language Scripting: Describing desired game mechanics in plain language to generate movement scripts
Tutorial Generation: Having Claude create step-by-step implementation guides
Debugging Partner: Troubleshooting issues through conversational problem-solving
Tools and Technologies
Primary Development: Unity 2022.3 LTS with Claude AI (Anthropic) for code generation
Design Tools: Figma for concept sketches, Resprite for pixel art
Game Concept: “Hedgie’s Easter Egg Hunt” — a 2D top-down timed collection game
Target Platforms: Cross-platform (desktop/mobile) to test tap-to-move mechanics

Design Constraints
I intentionally limited player interactions to tap and swipe gestures, mirroring the ease of use found in popular mobile games like FarmVille or cooking games. This constraint would help me evaluate whether to implement similar mechanics in the full Hedgietown game and test the viability of creating customized player movement scripts using AI code generation.

Design Process: From Concept to Playable Prototype

Initial Concept Development
The minigame concept emerged from my desire to test mobile-friendly mechanics. I went with a simple collection game format to focus primarily on player movement mechanics — simple enough to complete within the hackathon timeframe while yielding valuable insights about player behavior. Key questions included: How would players feel while moving around the map? What would they try to do? Would they prefer keyboard controls or tap-to-move mechanics?

Day One: Scene Creation and Movement Systems
I began by sketching the game map in Figma, designing a test scene with various obstacles and buildings. My focus was on setting up the scene and “vibe coding” the movement script. I kept the scene compact and implemented a visual indicator so players could see where they clicked on screen. Surprisingly, the basic movement code worked almost immediately after just a couple of prompts to Claude.

Day Two: Game Mechanics and UI Implementation
The second day focused on the egg collection game mechanics and game UI. I described my vision to Claude: a simple Easter egg collection game where players approach an NPC to start, then have 60 seconds to collect as many eggs as possible before seeing their final score. Claude provided a guided tutorial for setting up the game manager and UI elements for displaying time remaining and current score.

Technical Challenges and Creative Solutions
My biggest challenge was debugging. Multiple instances arose where I couldn’t identify what was wrong with the code or why the game wasn’t functioning as expected.

For example, I encountered issues with egg spawn points and the quest system not triggering when talking to the NPC. Rather than spend more hours debugging, I adapted the game flow: players now start collecting eggs immediately after clicking “start.” This actually improved the user experience by displaying all instructions upfront without requiring players to sit through dialogue. Small things kept building up that I couldn’t vibe-code my way out of. This raised concerns about whether I could create truly polished games and at what level of complexity — it felt I was starting to see the limitations of this approach.

Outcomes: Testing Results and User Feedback
118 prompts later, I had a complete game playable on both mobile and desktop devices. User testing revealed significant differences in player experience depending on the platform.

Performance Metrics
Mobile players: Collected an average of 9/16 eggs
Desktop players: Collected an average of 15/16 eggs
Perfect scores: Only one player collected all 16 eggs (on desktop)

Platform-Specific Feedback
Desktop players consistently performed better and reported that movement felt “more fun” on desktop. While players said the tap-to-move mechanic felt good overall, mobile players frequently got stuck on obstacles because the pathfinding system chose the shortest route, causing the character to rub against objects.

Key Areas for Improvement
User testing uncovered several critical issues:

Collision detection: Players frequently got stuck on obstacles
Mobile movement patterns: Players needed longer-distance movement options instead of relying on small taps
Egg collection sensitivity: Collision detection felt either too sensitive or misaligned with visual feedback

Reflections: Transformation and Future Implications
I used to think using AI was cheating — that it meant I wasn’t the one truly creating. I always needed an engineer to bring my ideas to life. Yes, I tried to code before, but I always got frustrated or confused. Then AI arrived, and suddenly my ideas became real. I’m not an engineer, but I know how something should feel. I don’t always know what I’m doing, but I always know what I’m trying to make.

A New Design Process
This experience taught me that you still have to design the experience. With a solid foundation going into development, you can test your ideas and see in real-time the technical limitations developers often discuss. It’s an interesting time to be a designer — being multidisciplinary is becoming increasingly important.

The new workflow becomes: design, prototype, build, iterate. You can now position yourself as a design engineer. If you’re a designer with ideas who can execute them, you become remarkably well-rounded, combining empathy and human psychology with aesthetics and functionality.

Being human is more important than ever. Designers have always been the dreamers — we need crazy ideas. AI amplifies this role. Being a designer should be less about optimization and more about dreaming, about pushing boundaries and imagining new possibilities.

Looking Forward
Will I get close to the game I envision for Hedgietown? Who knows, but I’ll get myself as close as I can. If this hackathon taught me anything, it’s that the gap between imagination and implementation is rapidly shrinking. The question isn’t whether designers can learn to code with AI — it’s how quickly we can adapt to this new creative landscape.

For creative professionals, this represents both an opportunity and a responsibility. We can now test our wildest ideas, but we must also develop new skills and ways of thinking. The future belongs to designers who can dream big and execute effectively — and AI just gave us the tools to do both.

Link to game, have fun :)
https://trinatle.itch.io/hedgietown-easter-egg-hunt

References
Context Engineering. (2025). Context Engineering is the New Vibe Coding (Learn this Now). YouTube. https://www.youtube.com/watch?v=P9A2NP5DtAA
Nielsen, J. (2025). Vibe Coding and Vibe Design. Jakob Nielsen Ph.D. Substack. https://jakobnielsenphd.substack.com/p/vibe-coding-vibe-design
The Week. (2025, June 3). The rise of ‘vibe coding’. The Week. https://theweek.com/tech/the-rise-of-vibe-coding
Verified Insider. (2025). Product Design Hiring in 2025. Verified Insider Substack. https://verifiedinsider.substack.com/p/product-design-hiring-in-2025-b47

Assets: https://cupnooble.itch.io/sprout-lands-asset-pack
”
+infos(fonte): LINK