graph LR
    A[Treat AI as a Skilled Intern] --> B[Structure Your Approach]
    B --> C[Create Specialized Prompts]
    C --> D[Maintain Context]
    D --> E[Review & Refine]
    
    style A fill:#f9f,stroke:#333,stroke-width:2px,color:#000
    style B fill:#bbf,stroke:#333,stroke-width:2px,color:#000
    style C fill:#bfb,stroke:#333,stroke-width:2px,color:#000
    style D fill:#fbf,stroke:#333,stroke-width:2px,color:#000
    style E fill:#fbb,stroke:#333,stroke-width:2px,color:#000

• Mental Model: Think of AI as a skilled intern who needs proper guidance, not a mind reader or a coding wizard
• Project Structure: Use text-based hierarchical project management (specifications → user stories → tasks)
• Context Preservation: Store prompts and project docs in version control for reuse and sharing
• Specialized Templates: Create dedicated prompts for different development scenarios (new features, refactoring, bug fixes)
• Progressive Refinement: Continuously improve your templates and process based on what works best

The biggest mistake developers make is falling into the Schrödinger's AI Paradox: simultaneously believing AI is too limited for serious development work, yet expecting it to implement impossibly complex systems.

❌ "Implement a new award-winning game!"

❌ "Implement a new feature even I don't understand. Please read these 200 pages of documentation to figure it out."

Even if you get lucky with good results, you'll immediately face these problems:

• 🥺 The AI loses context in the next session
• 🥺 The work isn't easily shareable with colleagues

Ask yourself: Would you give the same task to a new intern? Probably not!

Think of AI tools like GitHub Copilot or Cursor as skilled interns who need proper guidance:

1. Code familiarization - Intern taking notes
2. Project management introduction
3. Creating well-specified user stories and tasks

• Onboarding forces you to develop a comprehensive overview
• Project management helps track progress effectively
• Well-specified items lead to consistently better results

The right communication approach with AI tools makes all the difference:

• ✅ Keep specifications as part of your repository - trackable and shareable
• ✅ AI understanding improves with better context
• ✅ Different workflows can be adapted for different use cases and team members

Important Note: AI capabilities are improving rapidly, so best practices are evolving week by week!

graph TD
    A[Project Specifications] --> B[User Stories]
    B --> C[Tasks]
    C --> D[Implementation]
    D --> E[Testing]
    E --> F[Documentation]
    F --> G[Refinement]
    G -->|Feedback Loop| A
    
    style A fill:#f9f,stroke:#333,stroke-width:2px,color:#000
    style B fill:#bbf,stroke:#333,stroke-width:2px,color:#000
    style C fill:#bfb,stroke:#333,stroke-width:2px,color:#000

• Define current state and/or goal state
• High level view of the project

• Features to implement
• Will change the state of specifications
• Focus on clear outcomes

• Concrete plan on how to implement each user story
• Self-tracking capabilities
• Will update the state of user stories

Project Flexibility: Depending on the project size and complexity, you may need more or fewer levels in your hierarchy.

Does this mean I have to write plenty of text and stuff?

No, you of course use the very AI to create those for you!

• 🚀 Maintains context across multiple sessions
• 🤝 Facilitates collaboration with both AI and human team members
• 📈 Produces higher quality, more maintainable code
• 📚 Gradually builds useful documentation
• 🔄 Adapts to rapidly evolving AI capabilities

• 🚀 Every team has to "find" itself - very dependent on internal requirements, constraints, philosophy
• 🤝 AI is improving week by week
• 📈 Tools are improving week by week
• 📚 Organizational overhead
• 🔄 You don't write much code anymore

→ No pre-cut best practice "That's how you HAVE TO do it" possible
→ Learning by doing is essential

In the upcoming practical session, you'll apply this framework to experience how structured prompting dramatically improves AI tool performance.

What we'll create:

• A .NET C# CLI app for managing project management with GitHub Copilot and Cursor

What does this app do?

• By letting the user go through specialized wizards the cli app should create prompt templates as described in the Sessions inside docs/