Back to Journal

AI Game Dev Adventures

Building games with artificial intelligence

Day 52025-08-19
roadways-clone

Mini Motorways Clone - The Comprehensive PRD Approach

Attempting to create a successful one-shot clone of Mini Motorways with comprehensive requirements and multiple screenshots, only to discover the dangers of overwhelming AI with too much information at once.

Screenshots from Day 5
Screenshot 1 from Mini Motorways Clone - The Comprehensive PRD Approach

Mini Motorways Clone - The Comprehensive PRD Approach

Challenge:

Can I create a successful one-shot clone with comprehensive requirements?

Attempt 1: Mini Motorways Clone (Take 1)

Took multiple screenshots of the iOS game, created detailed PRD, told AI to make exactly this game.

  • Combined screenshots, PRD, and additional context
  • Told AI: "Make exactly this game"
  • Provided maximum possible information upfront

I was confident this would work perfectly.

The Spectacular Failure

Total disaster.

What Went Wrong

  • HUD didn't work - UI elements were completely broken
  • Mechanics completely wrong - cars behaved nothing like the original
  • Went completely off track - AI started implementing features that didn't exist in the original

The Core Problem

The AI got completely overwhelmed by the amount of information and started making decisions I never asked for. Instead of following the requirements, it began improvising wildly.

The Unsupervised Problem

I made a critical mistake: I let the AI run unsupervised for too long.

When you give AI extensive requirements and then let it implement freely, it doesn't stick to the spec. It starts making creative interpretations and winds up in completely the wrong place.

Key Lessons Learned

Information Overload

  • More requirements ≠ better results
  • AI can get overwhelmed by comprehensive specifications
  • Too much context leads to creative interpretation rather than precise implementation

Supervision is Critical

  • Can't let AI run unsupervised for extended periods
  • Need frequent check-ins and course corrections
  • Long autonomous implementation leads to complete derailment

The Goldilocks Problem

  • Too few requirements: AI fills gaps poorly
  • Too many requirements: AI gets overwhelmed and improvises
  • Need to find the "just right" level of specification

Incremental Approach Needed

Rather than one massive requirements dump, need to:

  • Break problems down into smaller pieces
  • Implement incrementally with frequent validation
  • Maintain tight feedback loops

What's Next

The comprehensive PRD approach was a swing and a miss. Tomorrow I'm going to try the opposite approach: break everything down into much smaller, manageable pieces.

Maybe the secret isn't in perfect requirements, but in perfect process.

Ingredients

  • AI Assistant: Claude Sonnet - Struggled with comprehensive requirements
  • Reference Game: Mini Motorways (iOS) - Multiple screenshots taken
  • Approach: Detailed PRD → comprehensive one-shot development
  • Screenshot Tools: iOS screen capture - Free
  • PRD Creation: Detailed requirements document with multiple game references
  • Time Investment: Several hours on requirements + implementation attempts
  • Cost: ~$5-8 in AI credits for multiple comprehensive attempts
  • Key Failure: Too many requirements at once = AI improvisation
  • Discovery: Comprehensive approach leads to feature creep and derailment

Community & Discussion

Join the conversation through comments or share your thoughts across the web. Discuss this journal entry and share your experiences.

Comments & Discussion

Join the conversation! Comments are powered by GitHub Discussions. Share your thoughts about this journal entry.

💡 Tip: Comments are stored as GitHub Discussions, while web mentions capture shares and mentions from Twitter, Mastodon, blogs, and other platforms. Together they provide a complete picture of community engagement.