You are an expert learning architect who expands small seed content into comprehensive, structured learning hierarchies.
You specialize in writing “can-do” statements and organizing them into clear learning sequences.

# TASK OVERVIEW
Your task is to take the provided roadmap, and fill in the missing pieces so that it matches the output format required.

# INPUTS
- Topic: {topic}
- Audience: {audience}
- Purpose: {purpose}
- Language style: {style}
- Roadmap:
```
# Roadmap: Using AI in Software Engineering — Integrated Can-Do Statements

## Track 1: Overview & Benefits

### Topic: Foundations (definitions of AI/ML, LLMs, traditional vs AI approaches, `llms.txt`)
- I know the definitions of AI, machine learning, and large language models.
- I understand the difference between traditional programming and AI-driven approaches.
- I know what `llms.txt` is and why it matters for transparency and permissions.
- I understand that AI is a tool to support, not replace, engineering judgment.
- I understand that AI predicts the next likely token/value, not “thinking.”

### Topic: Opportunities and Risks (benefits vs limits)
- I achieve faster development by applying AI effectively.
- I deliver cost savings by using AI in the right contexts.
- I enhance creativity by exploring ideas with AI.
- I improve quality and documentation with AI assistance.
- I recognize that AI outputs may be inaccurate or biased.
- I balance benefits and risks when deciding how to use AI.
- I know the limits of AI in logic, reasoning, and novelty.
- I can evaluate when AI is the right tool versus traditional methods.
- I can adapt as AI capabilities evolve (e.g., improvements in logic/reasoning).
- I have awareness of emerging ethical and societal debates around AI use.

---

## Track 2: Theory of Machine Learning

### Topic: Core Concepts
- I know the difference between supervised, unsupervised, and reinforcement learning.
- I understand the difference between classification and regression.
- I know the main types of ML models (regression, classification, clustering, generative).
- I understand the role of datasets, features, and labels in training.
- I know how training, validation, and test sets are used to evaluate models.
- I understand the concept of overfitting and underfitting.

### Topic: Neural Networks & Deep Learning
- I understand how neural networks use layers of connected nodes to learn patterns.
- I know how weights, biases, and other parameters affect a model’s predictions.
- I understand the importance of hyperparameters such as learning rate and batch size.
- I understand what transfer learning is and when to use it.
- I know how temperature settings influence variability in outputs. [new]

### Topic: Practical Exploration
- I can use **playground.tensorflow.org** to explain how machines learn.
- I compare supervised and unsupervised approaches on simple datasets.
- I visualize model behavior to understand how it learns and adapts.

---

## Track 3: Practical Usage & Good Practices

### Topic: Effective Use & Iteration
- I give AI clear and specific instructions.
- I iterate on prompts instead of expecting one-shot answers.
- I use adversarial queries to identify weaknesses in AI outputs.
- I balance simplicity and complexity when integrating AI into systems.
- I am aware of costs, rate limits, and efficiency concerns when using AI.
- I frame clear, specific, and context-rich prompts.
- I reframe or simplify when outputs are unclear.
- I use structured prompts to explore multiple solution paths.
- I iterate step by step to refine outputs.
- I restart with clarified instructions when AI outputs fail. [duplicate]
- I have developed resilience in iterating with AI tools.

### Topic: Engineering Practices
- I use version control to safely track AI-assisted contributions.
- I recover easily from AI-influenced changes using rollback and history.
- I keep AI-assisted outputs accountable through commits and documentation.
- I provide AI with the right context to improve output quality.
- I am mindful of what gets committed to source control when using AI outputs.
- I can manage project context to get relevant results.

### Topic: Collaboration & Transparency
- I am transparent about when and how I use AI.
- I take responsibility for the outputs I generate with AI.
- I refine AI outputs thoughtfully before sharing them.
- I communicate clearly with colleagues and clients when AI has been used.
- I can explain AI tool outputs to team members and stakeholders.
- I can collaborate with colleagues to refine and validate AI results.

### Topic: Practical Applications
- I use AI to generate boilerplate or repetitive code.
- I use AI for debugging assistance.
- I use AI to improve code comments and documentation.
- I use AI to support testing and test case generation.
- I use AI to refine requirements and acceptance criteria.
- I use AI to summarize and improve text.
- I use AI to provide fresh perspectives in code reviews.
- I can use AI to draft documentation, design notes, or meeting summaries.

---

## Track 4: Safety, Ethics & Security

### Topic: Risk Awareness
- I recognize that AI outputs may be inaccurate or hallucinated.
- I understand that AI systems may reflect or amplify bias.
- I consider intellectual property concerns when using AI-generated content.
- I identify and mitigate security vulnerabilities that could arise from AI use.

### Topic: Responsible Practice
- I apply ethical principles when deciding how and when to use AI.
- I avoid using AI as a substitute for developing my own engineering skills.
- I align my use of AI with professional and organizational standards.
- I read model responses carefully instead of jumping straight to the result.
- I ask AI to consider performance, scalability, and security in solutions.
- I follow organizational guidelines for safe AI usage.

### Topic: Privacy & Compliance
- I anonymize sensitive data before sharing with AI tools.
- I minimize unnecessary data sharing with AI services.
- I use only approved AI tools for my projects.
- I validate AI outputs before relying on them.
- I comply with legal, contractual, and organizational AI usage policies.
- I understand the risks of exposing sensitive or proprietary information.
- I can configure AI tools with privacy and security settings.
- I can distinguish between safe local deployment and SaaS AI tools.
- I am aware of compliance requirements (data governance, IP, security policies).

---

## Track 5: Prompting

### Topic: Fundamentals
- I write clear and precise prompts.
- I use role prompting to guide AI behavior.
- I know the difference between instruction-style and completion-style prompts.

### Topic: Iteration
- I refine prompts iteratively to improve results.
- I use multi-turn prompting to develop ideas in a conversational flow.
- I reframe or simplify when outputs are unclear.
- I restart with clarified instructions when AI outputs fail.

### Topic: Advanced Techniques
- I use few-shot prompting to provide examples for AI.
- I apply chain-of-thought prompting for reasoning tasks.
- I use adversarial prompting to test AI outputs.
- I configure and apply system prompts effectively.
- I instruct AI using Markdown, code blocks, or structured formats effectively.

### Topic: Prompt Patterns
- I design reusable prompt templates.
- I control the style and tone of AI outputs through prompt design.
- I use structured patterns to achieve consistent outputs.
- I can create reusable prompt templates and knowledge bases.

### Topic: Evaluation
- I evaluate prompt effectiveness by comparing outputs.
- I reduce fragility in prompts to improve reliability.
- I ensure reproducibility by documenting effective prompts.

---

## Track 6: Models & Ecosystem

### Topic: Providers & Types
- I know the leading AI model providers.
- I understand the differences between open-source and closed-source models.
- I know when to use offline/local models instead of cloud-based ones.

### Topic: Modalities
- I know the difference between LLMs and other AI modalities.
- I use text-to-speech and speech-to-text models.
- I use AI for image and video generation.

### Topic: Access & Integration
- I use APIs and SDKs to access AI models programmatically.
- I chain inputs and outputs between models for complex workflows.

### Topic: Model Evaluation
- I know how to compare models using benchmarks and leaderboards.
- I select models based on performance and suitability for my task.
- I can explain trade-offs between lightweight vs. large models in practice.
- I can choose appropriate models balancing cost, speed, and accuracy.
- I understand the cost and energy implications of large AI models.
- I can optimize prompts and workflows to reduce unnecessary compute usage.

### Topic: Language & Domain Strengths
- I know that different models excel in different languages and domains.
- I choose models suited to the language or domain of my project.

---

## Track 7: Tools & Integration

### Topic: Development Tools
- I use AI-powered IDE extensions to improve coding productivity.
- I integrate automation helpers into my development workflow.

### Topic: Pipelines & Agents
- I use LangChain or similar frameworks to build AI agents.
- I integrate AI into pipelines with tools like Dagger.io.

### Topic: RAG (Retrieval-Augmented Generation)
- I understand the concept of retrieval-augmented generation (RAG).
- I know which models, datasets, and tools can be used for RAG.
- I use RAG to integrate external knowledge into AI-assisted tasks.

### Topic: MCP (Model Context Protocol)
- I understand the purpose of the Model Context Protocol.
- I use MCP to connect apps, tools, and AI agents.

### Topic: Setup & Sharing
- I configure AI tools with correct privacy and security settings.
- I share tool configurations with my team for consistent usage.

---

## Track 8: Teams & Collaboration

### Topic: Shared Practices
- I use AI to capture and refine meeting notes.
- I use AI to assist with code reviews while keeping team accountability.
- I use AI to help create and refine tickets from requirements.
- I use AI to support shared documentation for teams.

### Topic: Collaboration & Alignment
- I share AI configurations with my team.
- I align with my team on consistent AI workflows.
- I follow communication standards when AI has been used.
- I contribute to team knowledge by sharing AI learnings.
- I encourage adoption while addressing reliability and security concerns.

### Topic: Governance
- I follow team-defined rules for allowed and restricted AI tasks.
- I ensure compliance with project and client policies at the team level.
- I review team AI usage for security and privacy risks.
```

# OUTPUT FORMAT
Return JSON with:
```json
{{
  "slug": "{slug}",                  // provided value
  "id": "{slug}",                    // must be identical to slug
  "title": "Short human-friendly title (2–6 words)",
  "description": "8–20 word summary of the roadmap purpose",
  "expanded_statements": [ ... all of the can-do statements, matching the step titles ... ],
  "tracks": [
    {{
      "slug": "track-slug",
      "id": "track-slug",            // identical to slug
      "title": "Track Title",
      "description": "Track description",
      "order": 1,
      "paths": [
        {{
          "slug": "path-slug",
          "id": "path-slug",         // identical to slug
          "title": "Path Title",
          "description": "Path description",
          "order": 1,
          "steps": [
            {{
              "slug": "i-can-do-something",
              "id": "i-can-do-something",   // identical to slug
              "title": "I can do something",
              "description": "Short explanation",
              "level": "beginner",
              "order": 1
            }}
          ]
        }}
      ]
    }}
  ],
  "hierarchy_summary": "Explain how statements were expanded and organized"
}}