AI coding: debugging from a stack trace without guessing

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-ai-coding-debug-from-stack-trace-r7a1-creators

1. Why is it problematic to simply ask an AI coding assistant to 'fix this bug' when providing a stack trace?

   1. The AI lacks the authority to modify your codebase without permission
   2. The AI cannot read code files directly from your computer
   3. Stack traces contain sensitive information that should never be shared with AI
   4. The AI will likely guess at a solution rather than systematically analyze the problem

2. What is the most effective way to use AI when debugging a runtime error?

   1. Provide only the error message and wait for the AI to find the bug in your entire codebase
   2. Provide a screenshot of the error and ask the AI to explain what went wrong
   3. Provide the entire project and ask AI to rewrite all related functions to prevent future errors
   4. Provide the stack trace, the failing input, and the relevant function code, then ask for ranked hypotheses with experiments to verify each

3. According to the core concept being taught, what should you ask AI to provide for each hypothesis about a bug?

   1. A single minimal experiment that could confirm or disprove that hypothesis
   2. A summary of similar bugs from other projects
   3. The exact line number where the bug exists
   4. A complete rewritten function that fixes the issue

4. What limitation of AI debugging assistants is most likely to result in a fix that compiles but doesn't actually solve the root cause?

   1. AI cannot observe your actual runtime state and may mistake symptoms for root causes
   2. AI intentionally provides incorrect solutions to encourage learning
   3. AI cannot access external libraries to check for known bugs
   4. AI will always choose the simplest solution regardless of correctness

5. Why should you run an experiment to confirm a hypothesis before implementing an AI-suggested fix?

   1. Experiments are required by most coding style guides
   2. Experiments make your code run faster
   3. The AI might charge you for incorrect code suggestions
   4. Without verification, you risk fixing a symptom while the actual bug remains, leading to recurring issues

6. When analyzing a stack trace with AI, which task is AI particularly good at performing?

   1. Running your code to see what happens
   2. Compiling your code and reporting errors
   3. Mapping the stack trace to suspect lines and the sequence of function calls that led to the error
   4. Detecting security vulnerabilities in your codebase

7. What information should you ALWAYS provide to AI when debugging to get useful hypotheses?

   1. A video recording of the error occurring
   2. Your entire GitHub repository URL
   3. The stack trace, the failing input, and the relevant function code
   4. The names of all variables in your program

8. A student provides a stack trace to AI and asks for the three most likely root causes. The AI suggests three hypotheses. What should the student do next?

   1. Implement all three fixes simultaneously to see which one works
   2. Ask the AI to write unit tests for each hypothesis
   3. Pick one hypothesis and design a minimal experiment to confirm it before fixing anything
   4. Submit the hypotheses to their teacher for approval

9. Why might an AI-suggested fix compile and appear reasonable but still be incorrect?

   1. The AI addressed a symptom one layer above the actual bug, not the true root cause
   2. AI intentionally writes code with syntax errors to test you
   3. The AI ran out of memory while generating the fix
   4. The compiler may have a bug that prevents valid fixes from compiling

10. What does it mean for a debugging approach to be 'hypothesis-driven'?

    1. You only trust debugging results that come from automated testing frameworks
    2. You form testable explanations for why an error occurs and verify them before implementing fixes
    3. You write hypotheses in your code comments before running anything
    4. You guess wildly at possible solutions until something works

11. What can AI NOT do, even with a perfect stack trace and code?

    1. AI cannot generate hypotheses about potential causes
    2. AI cannot analyze code for potential errors
    3. AI cannot read code that uses certain programming languages
    4. AI cannot confirm a fix is correct without you running it

12. What is a 'minimal experiment' in the context of debugging with AI assistance?

    1. An experiment that runs in under one second
    2. A comprehensive test suite that covers all edge cases
    3. A experiment that modifies as few lines of code as possible
    4. The simplest possible test that can confirm or disprove a specific hypothesis about the bug

13. When AI provides ranked hypotheses about a bug, what does the ranking represent?

    1. The order in which you should implement the fixes
    2. The severity of each bug from least to most dangerous
    3. The alphabetical order of potential causes
    4. The AI's assessment of likelihood based on the evidence provided

14. What should you do if an AI-suggested fix compiles successfully but doesn't resolve the error when you run it?

    1. Blame the AI for providing bad advice
    2. Go back to your hypotheses and run additional experiments to identify the true root cause
    3. Give up on debugging
    4. Ask the AI for another fix without additional context

15. Why is describing your runtime state important when debugging with AI?

    1. Runtime state information makes your code run faster
    2. AI requires runtime state to compile your code
    3. AI uses runtime state to generate revenue
    4. Without describing runtime state, AI cannot distinguish between multiple possible causes that produce similar symptoms