Skip to content

Chapter 5 Quiz: Leveraging Generative AI

Test your understanding of working with LLMs for MicroSim development, including prompt engineering, iterative refinement, and AI limitations.

Instructions: Select the best answer for each question. After answering, click "Show Answer" to check your response and read the explanation.

1. What is the most accurate description of working with an LLM for code generation?

  1. The LLM replaces the need for human developers
  2. The LLM is a partner that requires clear communication and human judgment
  3. The LLM can execute and test the code it generates
  4. The LLM always produces perfect code on the first attempt
Show Answer

The correct answer is B. An LLM is a partner, not a replacement. Success depends on understanding each other's strengths and limitations, communicating clearly, and developing shared patterns of collaboration. LLMs cannot execute code or guarantee correctness.

Concept Tested: Large Language Models

Bloom's Level: Understand

2. What does a "token" represent in the context of LLMs?

  1. A complete JavaScript function
  2. A single character of text
  3. Roughly 4 characters or 0.75 words on average
  4. An error message from the model
Show Answer

The correct answer is C. A token is roughly 4 characters or 0.75 words on average. However, tokenization varies—common words may be single tokens, code symbols may be individual tokens, and unusual words may be split into multiple tokens.

Concept Tested: Token Limits

Bloom's Level: Remember

3. Which technique involves providing examples of desired output in your prompt?

  1. Chain of thought
  2. Few-shot learning
  3. Role prompting
  4. Structured output
Show Answer

The correct answer is B. Few-shot learning involves providing 2-3 examples of similar MicroSims or desired output in your prompt. This helps the LLM understand the pattern and format you expect.

Concept Tested: Prompt Engineering

Bloom's Level: Remember

4. What is an "AI hallucination" in code generation?

  1. When the AI takes too long to respond
  2. When the AI generates confident-sounding but incorrect code or information
  3. When the AI refuses to generate code
  4. When the AI produces duplicate output
Show Answer

The correct answer is B. AI hallucinations occur when an LLM generates confident-sounding but incorrect information—like inventing functions that don't exist, using incorrect syntax, or fabricating API methods. Always verify generated code against documentation.

Concept Tested: AI Hallucinations

Bloom's Level: Understand

5. What is the primary purpose of iterative refinement when working with AI?

  1. To increase token usage
  2. To improve AI-generated code through successive feedback cycles
  3. To bypass context window limits
  4. To create multiple versions for comparison
Show Answer

The correct answer is B. Iterative refinement is the process of improving AI-generated code through successive conversations. You generate, test, identify issues, specify fixes, and regenerate—typically 2-4 cycles per MicroSim.

Concept Tested: Iterative Refinement

Bloom's Level: Understand

6. When debugging with AI, what information should you include in your prompt?

  1. Only the error message
  2. The error message, relevant code, expected behavior, and actual behavior
  3. Your entire codebase
  4. A screenshot of the error
Show Answer

The correct answer is B. Effective debugging prompts include the exact error message, the relevant code section, a description of expected behavior, and what actually happens. This gives the AI complete context to diagnose and fix the problem.

Concept Tested: Code Debugging with AI

Bloom's Level: Apply

7. What is the "context window" in an LLM?

  1. The visual display of the chat interface
  2. The total amount of text (in tokens) the LLM can process at once
  3. The time limit for generating a response
  4. The number of conversations you can have per day
Show Answer

The correct answer is B. The context window is the total amount of text (in tokens) that an LLM can "see" at once—including your prompt, conversation history, and any included files. When exceeded, the AI may lose earlier context.

Concept Tested: Context Window

Bloom's Level: Remember

8. Which is an example of a common p5.js hallucination?

  1. Using createCanvas(400, 400) to create a canvas
  2. Using circle(x, y, radius) instead of circle(x, y, diameter)
  3. Using background(255) for a white background
  4. Using fill() before drawing shapes
Show Answer

The correct answer is B. A common hallucination is confusing radius with diameter in the circle() function. The correct syntax is circle(x, y, diameter), not radius. This subtle error can cause shapes to be half their intended size.

Concept Tested: AI Hallucinations

Bloom's Level: Apply

9. What is the purpose of a "rules file" when working with AI?

  1. To limit the AI's capabilities
  2. To provide consistent context about project standards and preferences
  3. To record all conversations with the AI
  4. To track token usage
Show Answer

The correct answer is B. Rules files are documents that provide consistent context about your project's standards, patterns, and preferences—like canvas dimensions, variable naming conventions, and code structure. They serve as an instruction manual for your AI partner.

Concept Tested: Rules Files

Bloom's Level: Understand

10. What is a key limitation of LLMs that affects MicroSim development?

  1. They can only generate JavaScript code
  2. They cannot execute code or see visual output
  3. They cannot understand natural language
  4. They can only work with short prompts
Show Answer

The correct answer is B. LLMs cannot execute code or see visual output. They predict what code should look like based on patterns, but they cannot run it, view your canvas, or detect visual bugs. You must test all code and evaluate visual results yourself.

Concept Tested: AI Limitations

Bloom's Level: Understand

11. What makes a "strong prompt" more effective than a "weak prompt"?

  1. Using more tokens
  2. Including specific, actionable details like dimensions, colors, and behavior
  3. Making it longer
  4. Using technical jargon
Show Answer

The correct answer is B. Strong prompts include specific, actionable details—exact canvas sizes, RGB colors, slider ranges, behavior descriptions, and code standards. Vague requests like "make a bouncing ball" produce unpredictable results compared to detailed specifications.

Concept Tested: AI Prompting

Bloom's Level: Analyze

12. What does Claude Code enable that differs from using a chat interface?

  1. Faster response times
  2. Direct reading and writing of project files
  3. Free unlimited usage
  4. Automatic error correction
Show Answer

The correct answer is B. Claude Code integrates directly into your development workflow, allowing Claude to read your existing files, understand your project structure, and write changes directly. This eliminates copying code between chat and editor, dramatically accelerating the iteration cycle.

Concept Tested: Claude Code

Bloom's Level: Remember

Quiz Summary

Concepts Tested:

Concept Questions
Large Language Models 1
Token Limits 2
Prompt Engineering 3
AI Hallucinations 4, 8
Iterative Refinement 5
Code Debugging with AI 6
Context Window 7
Rules Files 9
AI Limitations 10
AI Prompting 11
Claude Code 12

Bloom's Taxonomy Distribution:

  • Remember: 4 questions (33%)
  • Understand: 5 questions (42%)
  • Apply: 2 questions (17%)
  • Analyze: 1 question (8%)