Write Architecture Decision Records With AI

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-coder-adr-with-agent-creators

1. What is the primary purpose of an Architecture Decision Record (ADR) in an AI-agent project?

   1. To track which AI model version is currently deployed
   2. To replace traditional code comments in the repository
   3. To log every code change made by the AI agent
   4. To capture the reasoning behind architectural choices so future agents don't undo them casually

2. Which principle should be applied FIRST when starting an AI agent coding task?

   1. Define the failure path in detail
   2. Name the job the agent needs to accomplish before choosing the tool
   3. Select the most capable AI model available
   4. Write comprehensive test cases upfront

3. What does the lesson mean by writing the 'smallest useful scope' the agent can finish?

   1. Write the minimal code possible, even if it doesn't solve the core problem
   2. Have the agent work on one small file at a time regardless of the overall goal
   3. Define a focused, completable objective that delivers value without unnecessary features
   4. Minimize the number of test cases to speed up development

4. Why does the lesson advise running the result as a user rather than as a 'fan of the tool'?

   1. Because the tester's job is to find bugs, not validate the tool's capabilities
   2. Because tool fanboys write biased reviews
   3. To evaluate whether the solution actually solves the user's problem, not whether the AI tool performed well
   4. Because AI tools are unreliable and should not be trusted

5. Before sharing AI-generated code, which three things should be inspected?

   1. Syntax, variable names, and comment quality
   2. The diff, data access patterns, and failure path
   3. Test coverage, documentation, and CI pipeline status
   4. File size, line count, and execution speed

6. What is the relationship between a working AI demo and a production-ready system, as described in the lesson?

   1. AI demos cannot be made production-ready and should always be rewritten from scratch
   2. The real skill is turning a demo into something observable, reversible, and safe for others to use
   3. Production systems require removing all AI-generated code
   4. A working demo is always production-ready because it functions correctly

7. What does 'observable' mean in the context of turning an AI demo into production code?

   1. The system's behavior and state must be measurable and visible to operators
   2. The code must have extensive console logging statements
   3. The code must be open source and publicly viewable
   4. The AI must explain its reasoning for every decision it makes

8. What is the 'rollback path' and why is it important when deploying AI-generated changes?

   1. A documented procedure to undo changes if they cause problems
   2. A way to revert to an older AI model version
   3. A backup of the original codebase
   4. The path the AI takes when it fails to generate correct code

9. Which type of documentation helps context survive across different AI agent sessions?

   1. AGENTS.md, ADRs, and handoff docs
   2. Inline code comments only
   3. README.md files in each directory
   4. Git commit messages exclusively

10. Why is it risky to allow AI to make architectural changes without documentation?

    1. Documentation slows down development too much
    2. AI is always wrong about architecture
    3. Architectural changes don't need to be documented
    4. Future agents or developers may unknowingly undo important decisions because they don't understand the reasoning

11. What question should you answer to determine if an AI architectural change is successful?

    1. What rollback path exists if the output is wrong?
    2. Which AI model was used to generate the change?
    3. What test proves the change works?
    4. What data should the app or agent never expose?

12. What does 'reversible' mean in the context of AI-generated architectural changes?

    1. The AI can generate both forward and backward compatible code
    2. Changes can be undone easily if they cause problems
    3. The system can run in both directions
    4. The code can be compiled on any platform

13. What is a 'decision log' in the context of AI agent development?

    1. A list of all code decisions the AI made during a session
    2. A debugging output showing the AI's thought process
    3. A chronological record of architectural decisions and their reasoning
    4. A log file generated by the AI during execution

14. When inspecting 'data access' before sharing AI-generated code, what are you checking for?

    1. Whether the code uses an ORM or raw SQL
    2. How many rows the code can process at once
    3. What data the code can read, write, or expose—and whether that's appropriate
    4. Which database vendor is being used

15. The lesson emphasizes that AI can quickly create a working demo. What distinguishes a real developer skill from basic AI usage?

    1. Knowing which AI model to select
    2. Turning the demo into something observable, reversible, and safe for others to use
    3. Writing more demo code than the AI can generate
    4. Making the demo run faster