quiz.json

{
  "lesson": "15-prompt-caching",
  "title": "Prompt Caching and Context Caching",
  "questions": [
    {
      "stage": "post",
      "question": "What discount does Anthropic apply to cache reads versus the base input rate?",
      "options": [
        "25% off",
        "50% off",
        "75% off",
        "90% off"
      ],
      "correct": 3,
      "explanation": ""
    },
    {
      "stage": "post",
      "question": "Why must dynamic timestamps go below the cache breakpoint, not above it?",
      "options": [
        "Caches only hit when the prefix is byte-identical; a changing timestamp breaks the match for everything after it",
        "Timestamps confuse the tokenizer",
        "They cost more tokens than static text",
        "Anthropic explicitly rejects timestamps in cached blocks"
      ],
      "correct": 0,
      "explanation": ""
    },
    {
      "stage": "post",
      "question": "OpenAI's prompt caching is configured how?",
      "options": [
        "Explicit cache_control markers",
        "A CachedContent API you create and reference",
        "Automatic prefix matching with no configuration",
        "A system-level flag you toggle per project"
      ],
      "correct": 2,
      "explanation": ""
    },
    {
      "stage": "post",
      "question": "For Anthropic, what write premium does the 1-hour extended TTL cost vs the 5-minute default?",
      "options": [
        "Same",
        "2x the write premium (50% over baseline)",
        "4x the write premium",
        "No write premium"
      ],
      "correct": 1,
      "explanation": ""
    },
    {
      "stage": "post",
      "question": "How many reuses are needed to break even on Anthropic's 25% write premium?",
      "options": [
        "1",
        "2",
        "5",
        "10"
      ],
      "correct": 1,
      "explanation": ""
    }
  ]
}