quiz.json

{
  "lesson": "27-finops-llms",
  "title": "FinOps for LLMs — Unit Economics and Multi-Tenant Attribution",
  "questions": [
    {
      "stage": "pre",
      "question": "Why does traditional FinOps break on LLM spend?",
      "options": [
        "LLMs don't cost money",
        "Cloud providers refuse to itemize",
        "LLM bills are always free",
        "Costs are token-transactions rather than resource-uptime; tags don't auto-propagate from API calls and you must stamp user/task/tenant at the call site"
      ],
      "correct": 3,
      "explanation": ""
    },
    {
      "stage": "check",
      "question": "Which three attribution dimensions does the lesson require instrumenting on day one?",
      "options": [
        "Provider, model, API version",
        "Region, AZ, datacenter",
        "Per-user (user_id), per-task (task_id + route), per-tenant (tenant_id)",
        "GPU, CPU, RAM"
      ],
      "correct": 2,
      "explanation": ""
    },
    {
      "stage": "check",
      "question": "Which four token layers should be broken out in cost attribution?",
      "options": [
        "GPU, CPU, RAM, storage",
        "Cache, model, gateway, observability",
        "Input, output, network, disk",
        "Prompt, tool, memory, response"
      ],
      "correct": 3,
      "explanation": ""
    },
    {
      "stage": "check",
      "question": "What is the kill-switch trigger in the enforcement ladder?",
      "options": [
        "Spend over $1 in a minute",
        "Any 5xx response",
        "Tenant spend z-score > 4 relative to baseline; auto-pause tenant and page on-call",
        "Latency P50 > 2s"
      ],
      "correct": 2,
      "explanation": ""
    },
    {
      "stage": "post",
      "question": "Which unit metric does the lesson recommend instead of $/M tokens?",
      "options": [
        "Cost per second",
        "Cost per GPU-hour",
        "Cost per product outcome (e.g. cost per resolved support ticket, cost per generated article, cost per successful agent task)",
        "Cost per gateway"
      ],
      "correct": 2,
      "explanation": ""
    },
    {
      "stage": "post",
      "question": "Which attribution pattern does the lesson call the highest-accuracy one mature teams use?",
      "options": [
        "Tag-and-aggregate only",
        "Model-based allocation",
        "Telemetry joiner — join traces to billing via trace IDs",
        "Sampling and extrapolation"
      ],
      "correct": 2,
      "explanation": ""
    }
  ]
}