Inspiration
In late 2024, I found myself constantly second-guessing images online: "Is this AI or real?" As generative AI tools like Midjourney and DALL-E improved, distinguishing synthetic images from photographs became nearly impossible. I wasn't alone, family, and Reddit communities were having the same debates.
This sparked a question: What if i turned this uncertainty into a daily community game?
What it does
PicOrPixel is a daily community game built natively on Reddit's Devvit platform where players distinguish AI-generated images from real photographs.
Core Features
🎮 Daily Challenge Mode
- Every day , 5 new images are generated
- All players worldwide face the exact same challenges
- Categories: nature, art, portraits, architecture, food
⚡ Dynamic Scoring System
- Base points by difficulty: Easy (100), Medium (200), Hard (300)
- Immediate feedback after each selection
🏆 Competitive Leaderboards
- Real-time daily rankings (top 50 displayed)
- Personal statistics tracking (days played, best score, average accuracy)
- Community stats (total players, average accuracy)
🎯 Gameplay Flow
- Click "Play Today's Challenge" on the Reddit post
- See the image
- Click Real or AI-generated button
- Get instant feedback (✓ Correct! or ✗ Wrong)
- View your round score with time bonus
- Proceed through 5 rounds
- See final score and leaderboard position
- Come back tomorrow !
How I built it
The PicOrPixel game was created end-to-end using Kiro — from the first idea to a live, playable release.
I integrated the Devvit MCP server into Kiro, creating a unified workspace where Kiro could:
- Access Devvit APIs and documentation directly
- Generate platform-specific code
- Execute Devvit CLI commands
- Deploy to Reddit
I started by writing a high-level specification. From there, Kiro automatically generated requirements, produced a design blueprint, and broke everything into structured development tasks.
Each task was then executed directly by Kiro.
Two demo videos showcase this workflow:
🧩 Demo 1: From spec to tasks— Watch how Kiro breaks down requirements
🎮 Demo 2: Final gameplay
This project proves how Kiro can transform a simple concept into a complete, production-ready game — autonomously, efficiently, and fast.
Challenges I ran into
Translating creative ideas into structured prompts Kiro could fully understand.
Keeping control over the game’s direction while Kiro automated most of the process.
Ensuring smooth integration between generated assets, logic, and deployment. Despite the speed of automation, fine-tuning quality and consistency remained the main challenge.
What I learned
I learned that Kiro might be a little too good at its job 😅
From writing specs to generating tasks and shipping production code, Kiro handled everything so efficiently that i started questioning my job security.
When your AI teammate does the entire sprint in minutes, your new job is probably cheering it on.
What's next for PicOrPixel ?
Growing the community, refining difficulty based on player data, and exploring ways to make AI detection a skill everyone can learn through play.
Log in or sign up for Devpost to join the conversation.