Context Window Discipline: What Fits in AI's Memory

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-builders-prompting-context-window-discipline-r8a8-teen

1. What is a context window in an AI language model?

   1. The maximum amount of text the model can read and consider at one time when generating a response
   2. The amount of memory required to install the model on a computer
   3. The time period during which the model was trained on data
   4. The space where the model stores its training data for reference

2. A student pastes an entire 300-page history textbook into a prompt and asks when World War II ended. The AI gives a confusing or incorrect answer. What is the most likely reason?

   1. The textbook contains contradictory information that confused the AI
   2. The AI was not trained on history topics
   3. The relevant information is buried in too much irrelevant text, making it harder for the model to identify
   4. The model cannot read more than 100 pages at once

3. What skill does the lesson compare to effective prompting with context curation?

   1. Being a translator, who converts information between languages
   2. Being a librarian, who carefully selects relevant materials for a query
   3. Being a packrat, who saves everything just in case
   4. Being a detective, who investigates all available evidence

4. What is meant by 'context curation' in AI prompting?

   1. Expanding your context window by upgrading your AI subscription
   2. Writing longer and more detailed prompts to give the AI more information
   3. Storing useful AI responses for future conversations
   4. Carefully selecting only the most relevant information to include in your prompt

5. You have a 50-page research paper and want to ask about the methodology section. What is the best prompting approach?

   1. Ask the question first without any document, then paste the paper
   2. Paste only the methodology section along with your question
   3. Paste all 50 pages so the AI has full context
   4. Copy the abstract and conclusion, skipping the middle sections

6. What does the term 'needle in a haystack' refer to in AI prompting?

   1. When important information gets lost in a large amount of irrelevant text
   2. When an AI is searching for a specific word in its training data
   3. When you ask an AI to find a physical needle in real life
   4. When the AI provides an answer that is completely wrong

7. A developer includes 10 different PDF documentation files in their prompt but gets confused answers from the AI. What likely went wrong?

   1. The AI cannot read PDF files properly
   2. The documentation files are too old for the AI to understand
   3. The AI needs all 10 files to give any answer at all
   4. The AI got confused trying to determine which documentation was relevant to the question

8. Modern AI models like Claude and GPT can handle 200,000+ tokens. Why is still important to limit what you paste?

   1. Modern models cannot actually read that many tokens accurately
   2. Tokens beyond 200,000 cause the AI to crash
   3. The 200,000 token limit only applies to output, not input
   4. Even within the token limit, performance degrades when irrelevant information dilutes the relevant content

9. What happens when you ask an AI about a document after pasting only the relevant section versus pasting the entire document?

   1. The answers are identical because the AI reads everything equally well
   2. The relevant section alone never produces correct answers
   3. The answer is usually sharper and more accurate with just the relevant section
   4. The entire document always produces better answers because it has more information

10. If you paste a whole textbook to ask one question, what is the AI likely to do?

    1. Refuse to answer because the textbook is too long
    2. Miss the answer that was in an earlier section because it got lost in the text
    3. Read the entire textbook perfectly and find the exact answer
    4. Only answer questions about the first chapter

11. Which statement best describes the relationship between context size and answer quality in AI prompting?

    1. Answer quality is completely unrelated to how much context you provide
    2. More context always produces better answers regardless of relevance
    3. Less context always produces wrong answers
    4. More context only helps when it is all relevant; mixed relevant and irrelevant context often hurts

12. What should you do before pasting a long document into an AI prompt?

    1. Paste everything because the AI will automatically find what matters
    2. Identify and paste only the sections most likely to contain the answer to your question
    3. Delete all formatting and only paste plain text
    4. Copy only the title and first paragraph

13. Why might an AI give a worse answer when you include documentation files that don't relate to your question?

    1. The unrelated files create noise that makes it harder for the model to identify what actually matters
    2. The AI cannot read more than one file at a time
    3. The AI becomes overloaded and stops processing entirely
    4. Including multiple files always makes the AI slower but not less accurate

14. What is the key difference between a librarian approach and a packrat approach to AI prompting?

    1. A librarian selects only relevant information; a packrat includes everything available
    2. A librarian uses technical terms; a packrat uses simple language
    3. A librarian writes longer prompts; a packrat writes shorter ones
    4. A librarian asks yes/no questions; a packrat asks open-ended questions

15. When would pasting an entire document be appropriate?

    1. Only when the document is shorter than 10 pages
    2. When your question genuinely requires understanding the whole document, such as summarizing it
    3. Never, because you should always paste only partial sections
    4. Always, because more information is better