{
  "version": 1,
  "language": "en",
  "title": "Select the next test for the bottleneck",
  "description": "Predict the changed resource. Compare one intervention under fixed conditions. Then select the next test from the measured result.",
  "caption": "Give the name of the resource that the intervention changes. Predict the new bottleneck. Compare one change under the same workload. Use the outcome to select the next measurement, not a fixed model-to-scheduler checklist. Author-created explanatory diagram. The playback is an example. Animation time is not measured runtime.",
  "alt": "Experiment loop: predict the changed resource. Do a test with one changed variable. Measure quality and service. Select the next hypothesis.",
  "sources": [
    {
      "path": "content/blog/2026-06-12_token-journey-across-llm-stack/post.md",
      "sha256": "66e83fe38f5b4077f193d245eae19f962b670be4bf39ee807dc5f8c02367c861",
      "description": "Current article explanation, equations, assumptions and retained citations."
    },
    {
      "path": "research/dynamic-mechanisms-2026-10-21/before/content/blog/2026-06-12_token-journey-across-llm-stack/illustrations/06-bottleneck-experiment-loop-algorithm.svg",
      "sha256": "32906ff7d9e21bddadf77ff9ae1a51f52ab1fdce9227568550bb0f8a5b7555a6",
      "description": "Frozen prior illustration; exact pre-migration bytes retained."
    },
    {
      "path": "content/blog/2026-06-12_token-journey-across-llm-stack/illustrations/source-notes.json",
      "description": "Existing article illustration provenance, retained unchanged."
    }
  ],
  "limitations": [
    "The execution is an example. Playback duration is not measured runtime."
  ],
  "targets": {
    "observe": {
      "id": "mechanism-observe",
      "kind": "node",
      "label": "Examine symptom"
    },
    "resource": {
      "id": "mechanism-resource",
      "kind": "node",
      "label": "Give resource name"
    },
    "predict": {
      "id": "mechanism-predict",
      "kind": "node",
      "label": "Predict change"
    },
    "change": {
      "id": "mechanism-change",
      "kind": "node",
      "label": "Change one variable"
    },
    "measure": {
      "id": "mechanism-measure",
      "kind": "node",
      "label": "Measure outcome"
    },
    "follow": {
      "id": "mechanism-follow",
      "kind": "node",
      "label": "Follow bottleneck"
    },
    "name": {
      "id": "mechanism-name",
      "kind": "edge",
      "label": "Give resource name"
    },
    "predict-next": {
      "id": "mechanism-predict-next",
      "kind": "edge",
      "label": "Predict the effect"
    },
    "intervene": {
      "id": "mechanism-intervene",
      "kind": "edge",
      "label": "Change one variable"
    },
    "measure-next": {
      "id": "mechanism-measure-next",
      "kind": "edge",
      "label": "Measure the outcome"
    },
    "interpret": {
      "id": "mechanism-interpret",
      "kind": "edge",
      "label": "Interpret the result"
    },
    "repeat": {
      "id": "mechanism-repeat",
      "kind": "edge",
      "label": "Select the next test"
    }
  },
  "stateLabels": {
    "stage": "Stage",
    "artifact": "Artifact"
  },
  "scenarios": [
    {
      "id": "controlled-experiment",
      "title": "One controlled iteration",
      "description": "One controlled iteration",
      "initial": {
        "stage": "Prepared",
        "artifact": "Inputs kept"
      },
      "unselected": [],
      "steps": [
        {
          "title": "Give the symptom",
          "description": "Give the symptom and the workload, quality and service target.",
          "duration": 3000,
          "active": [
            "observe"
          ],
          "paths": [],
          "state": {
            "stage": "Give the symptom",
            "artifact": "Workload and service/quality target"
          }
        },
        {
          "title": "Give changed resource name",
          "description": "Give the name of the resource that a change affects.",
          "duration": 3000,
          "active": [
            "resource"
          ],
          "paths": [
            "name"
          ],
          "state": {
            "stage": "Give changed resource name",
            "artifact": "Possible work, traffic or byte-seconds limit"
          }
        },
        {
          "title": "Predict a consequence",
          "description": "Predict a testable consequence. It is not a speedup claim.",
          "duration": 3000,
          "active": [
            "predict"
          ],
          "paths": [
            "predict-next"
          ],
          "state": {
            "stage": "Predict a consequence",
            "artifact": "A testable consequence, not a speedup claim"
          }
        },
        {
          "title": "Change one variable",
          "description": "Change one variable. Use matched conditions for comparison.",
          "duration": 3000,
          "active": [
            "change"
          ],
          "paths": [
            "intervene"
          ],
          "state": {
            "stage": "Change one variable",
            "artifact": "Matched conditions for comparison"
          }
        },
        {
          "title": "Measure the outcome",
          "description": "Measure quality, service and resources.",
          "duration": 3000,
          "active": [
            "measure"
          ],
          "paths": [
            "measure-next"
          ],
          "state": {
            "stage": "Measure the outcome",
            "artifact": "Quality, service and resource observations"
          }
        },
        {
          "title": "Select the next hypothesis",
          "description": "Select the next hypothesis. Use the observed bottleneck to select the next test.",
          "duration": 3000,
          "active": [
            "follow"
          ],
          "paths": [
            "interpret"
          ],
          "state": {
            "stage": "Select the next hypothesis",
            "artifact": "The observed bottleneck helps select the next test."
          }
        },
        {
          "title": "Do the steps again for the observed limit",
          "description": "Do the steps again for the observed limit. Keep the next resource hypothesis testable.",
          "duration": 3000,
          "active": [
            "resource"
          ],
          "paths": [
            "repeat"
          ],
          "state": {
            "stage": "Do the steps again for the observed limit",
            "artifact": "Keep the next resource hypothesis testable."
          }
        }
      ]
    }
  ]
}