{
  "schema_version": 1,
  "kind": "offline_local_browser_observation_action_ab",
  "timestamp_utc": "2026-10-04T05:45:30.058342+00:00",
  "version": "0.4.0",
  "python": "3.14.2",
  "platform": "Darwin",
  "baseline_commit": "feb2bc21f34e5a7c918ae6dde8b35780c16e197a",
  "baseline_page_js_sha256": "24b2289731daa9eb166e4222fad0c5320539eaa3f55d2bc0f143953fc0436dd7",
  "treatment_page_js_sha256": "4a577c21af7a75d5a0945b2dfc488bdf6d086bf665bd08f4c20649fa67ee1f67",
  "fixture_sha256": "0e18064442115fb678662ff0621716df31387e6eefd448ce3de965842aca3a31",
  "result_cache": false,
  "model_usage": [],
  "scope": "Pinned old and current DOM observer with identical current CDP transport, fixed program-owned action policy, synthetic about:blank HTML; no Jev inference or external accounts.",
  "cost_assumption": "No inference requests: zero model tokens and inference cost, not a hosted end-to-end model cost estimate.",
  "timing_boundary": "Fixture navigation plus injection, observation, action and independent DOM checks; browser startup excluded.",
  "limitations": [
    "Typed-state probe requires boolean ARIA/native states and exact current SELECT value; it is an observation contract check, not model accuracy.",
    "More observed controls and successful multi-step scrolling increase context and runtime; medians combine different task completion paths.",
    "Roleless controls require explicit onclick or focusable pointer styling; transparent inputs supported only for labelled native checkbox/radio toggles.",
    "Nested scrolling is vertical; new-tab switching, closed shadow roots and cross-origin frames remain outside scope."
  ],
  "summary": {
    "baseline": {
      "passed_cases": 3,
      "cases": 15,
      "typed_state_recall": 0.2,
      "wrong_actions": 0,
      "freshness_rejections": 6,
      "freshness_checks": 6,
      "median_whole_operation_ms": 5.918,
      "mean_planned_context_bytes": 5423
    },
    "treatment": {
      "passed_cases": 15,
      "cases": 15,
      "typed_state_recall": 1.0,
      "wrong_actions": 0,
      "freshness_rejections": 6,
      "freshness_checks": 6,
      "median_whole_operation_ms": 11.99,
      "mean_planned_context_bytes": 9838
    }
  },
  "runs": [
    {
      "case": "typed_states",
      "arm": "baseline",
      "repeat": 0,
      "passed": false,
      "probes": [
        {
          "label": "Notifications",
          "field": "checked",
          "expected": false,
          "actual": "false",
          "match": false
        },
        {
          "label": "Pin item",
          "field": "pressed",
          "expected": false,
          "actual": "false",
          "match": false
        },
        {
          "label": "History",
          "field": "selected",
          "expected": false,
          "actual": "false",
          "match": false
        },
        {
          "label": "Sort",
          "field": "control_value",
          "expected": "recent",
          "actual": null,
          "match": false
        },
        {
          "label": "Note",
          "field": "value",
          "expected": "",
          "actual": "",
          "match": true
        }
      ],
      "typed_state_recall": 0.2,
      "missing_or_mismatched_controls": 4,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 30.292,
      "planned_context_bytes": 8053,
      "context_samples_bytes": [
        8053
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "typed_states",
      "arm": "treatment",
      "repeat": 0,
      "passed": true,
      "probes": [
        {
          "label": "Notifications",
          "field": "checked",
          "expected": false,
          "actual": false,
          "match": true
        },
        {
          "label": "Pin item",
          "field": "pressed",
          "expected": false,
          "actual": false,
          "match": true
        },
        {
          "label": "History",
          "field": "selected",
          "expected": false,
          "actual": false,
          "match": true
        },
        {
          "label": "Sort",
          "field": "control_value",
          "expected": "recent",
          "actual": "recent",
          "match": true
        },
        {
          "label": "Note",
          "field": "value",
          "expected": "",
          "actual": "",
          "match": true
        }
      ],
      "typed_state_recall": 1.0,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 7.548,
      "planned_context_bytes": 8200,
      "context_samples_bytes": [
        8200
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "roleless_click",
      "arm": "baseline",
      "repeat": 0,
      "passed": false,
      "probes": [
        {
          "noninteractive_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 1,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 5.141,
      "planned_context_bytes": 2840,
      "context_samples_bytes": [
        2840
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "roleless_click",
      "arm": "treatment",
      "repeat": 0,
      "passed": true,
      "probes": [
        {
          "noninteractive_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 1,
      "whole_operation_ms": 30.659,
      "planned_context_bytes": 5242,
      "context_samples_bytes": [
        5242
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "transparent_toggle",
      "arm": "baseline",
      "repeat": 0,
      "passed": false,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 1,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 7.673,
      "planned_context_bytes": 2845,
      "context_samples_bytes": [
        2845
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "transparent_toggle",
      "arm": "treatment",
      "repeat": 0,
      "passed": true,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 1,
      "whole_operation_ms": 15.45,
      "planned_context_bytes": 5319,
      "context_samples_bytes": [
        5319
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "nested_scroll",
      "arm": "baseline",
      "repeat": 0,
      "passed": false,
      "probes": [],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 1,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 5.343,
      "planned_context_bytes": 2873,
      "context_samples_bytes": [
        2873
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "nested_scroll",
      "arm": "treatment",
      "repeat": 0,
      "passed": true,
      "probes": [],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 6,
      "executed_observed_actions": 6,
      "whole_operation_ms": 27.429,
      "planned_context_bytes": 19925,
      "context_samples_bytes": [
        2909,
        2926,
        2926,
        2926,
        2926,
        5312
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "freshness_occlusion",
      "arm": "baseline",
      "repeat": 0,
      "passed": true,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 2,
      "freshness_checks": 2,
      "observations": 2,
      "executed_observed_actions": 0,
      "whole_operation_ms": 11.366,
      "planned_context_bytes": 10504,
      "context_samples_bytes": [
        5252,
        5252
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "freshness_occlusion",
      "arm": "treatment",
      "repeat": 0,
      "passed": true,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 2,
      "freshness_checks": 2,
      "observations": 2,
      "executed_observed_actions": 0,
      "whole_operation_ms": 11.619,
      "planned_context_bytes": 10504,
      "context_samples_bytes": [
        5252,
        5252
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "typed_states",
      "arm": "treatment",
      "repeat": 1,
      "passed": true,
      "probes": [
        {
          "label": "Notifications",
          "field": "checked",
          "expected": false,
          "actual": false,
          "match": true
        },
        {
          "label": "Pin item",
          "field": "pressed",
          "expected": false,
          "actual": false,
          "match": true
        },
        {
          "label": "History",
          "field": "selected",
          "expected": false,
          "actual": false,
          "match": true
        },
        {
          "label": "Sort",
          "field": "control_value",
          "expected": "recent",
          "actual": "recent",
          "match": true
        },
        {
          "label": "Note",
          "field": "value",
          "expected": "",
          "actual": "",
          "match": true
        }
      ],
      "typed_state_recall": 1.0,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 11.99,
      "planned_context_bytes": 8200,
      "context_samples_bytes": [
        8200
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "typed_states",
      "arm": "baseline",
      "repeat": 1,
      "passed": false,
      "probes": [
        {
          "label": "Notifications",
          "field": "checked",
          "expected": false,
          "actual": "false",
          "match": false
        },
        {
          "label": "Pin item",
          "field": "pressed",
          "expected": false,
          "actual": "false",
          "match": false
        },
        {
          "label": "History",
          "field": "selected",
          "expected": false,
          "actual": "false",
          "match": false
        },
        {
          "label": "Sort",
          "field": "control_value",
          "expected": "recent",
          "actual": null,
          "match": false
        },
        {
          "label": "Note",
          "field": "value",
          "expected": "",
          "actual": "",
          "match": true
        }
      ],
      "typed_state_recall": 0.2,
      "missing_or_mismatched_controls": 4,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 6.132,
      "planned_context_bytes": 8053,
      "context_samples_bytes": [
        8053
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "roleless_click",
      "arm": "treatment",
      "repeat": 1,
      "passed": true,
      "probes": [
        {
          "noninteractive_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 1,
      "whole_operation_ms": 24.225,
      "planned_context_bytes": 5242,
      "context_samples_bytes": [
        5242
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "roleless_click",
      "arm": "baseline",
      "repeat": 1,
      "passed": false,
      "probes": [
        {
          "noninteractive_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 1,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 5.918,
      "planned_context_bytes": 2840,
      "context_samples_bytes": [
        2840
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "transparent_toggle",
      "arm": "treatment",
      "repeat": 1,
      "passed": true,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 1,
      "whole_operation_ms": 12.203,
      "planned_context_bytes": 5319,
      "context_samples_bytes": [
        5319
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "transparent_toggle",
      "arm": "baseline",
      "repeat": 1,
      "passed": false,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 1,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 5.24,
      "planned_context_bytes": 2845,
      "context_samples_bytes": [
        2845
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "nested_scroll",
      "arm": "treatment",
      "repeat": 1,
      "passed": true,
      "probes": [],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 6,
      "executed_observed_actions": 6,
      "whole_operation_ms": 26.286,
      "planned_context_bytes": 19925,
      "context_samples_bytes": [
        2909,
        2926,
        2926,
        2926,
        2926,
        5312
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "nested_scroll",
      "arm": "baseline",
      "repeat": 1,
      "passed": false,
      "probes": [],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 1,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 5.907,
      "planned_context_bytes": 2873,
      "context_samples_bytes": [
        2873
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "freshness_occlusion",
      "arm": "treatment",
      "repeat": 1,
      "passed": true,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 2,
      "freshness_checks": 2,
      "observations": 2,
      "executed_observed_actions": 0,
      "whole_operation_ms": 7.772,
      "planned_context_bytes": 10504,
      "context_samples_bytes": [
        5252,
        5252
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "freshness_occlusion",
      "arm": "baseline",
      "repeat": 1,
      "passed": true,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 2,
      "freshness_checks": 2,
      "observations": 2,
      "executed_observed_actions": 0,
      "whole_operation_ms": 9.046,
      "planned_context_bytes": 10504,
      "context_samples_bytes": [
        5252,
        5252
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "typed_states",
      "arm": "baseline",
      "repeat": 2,
      "passed": false,
      "probes": [
        {
          "label": "Notifications",
          "field": "checked",
          "expected": false,
          "actual": "false",
          "match": false
        },
        {
          "label": "Pin item",
          "field": "pressed",
          "expected": false,
          "actual": "false",
          "match": false
        },
        {
          "label": "History",
          "field": "selected",
          "expected": false,
          "actual": "false",
          "match": false
        },
        {
          "label": "Sort",
          "field": "control_value",
          "expected": "recent",
          "actual": null,
          "match": false
        },
        {
          "label": "Note",
          "field": "value",
          "expected": "",
          "actual": "",
          "match": true
        }
      ],
      "typed_state_recall": 0.2,
      "missing_or_mismatched_controls": 4,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 6.953,
      "planned_context_bytes": 8053,
      "context_samples_bytes": [
        8053
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "typed_states",
      "arm": "treatment",
      "repeat": 2,
      "passed": true,
      "probes": [
        {
          "label": "Notifications",
          "field": "checked",
          "expected": false,
          "actual": false,
          "match": true
        },
        {
          "label": "Pin item",
          "field": "pressed",
          "expected": false,
          "actual": false,
          "match": true
        },
        {
          "label": "History",
          "field": "selected",
          "expected": false,
          "actual": false,
          "match": true
        },
        {
          "label": "Sort",
          "field": "control_value",
          "expected": "recent",
          "actual": "recent",
          "match": true
        },
        {
          "label": "Note",
          "field": "value",
          "expected": "",
          "actual": "",
          "match": true
        }
      ],
      "typed_state_recall": 1.0,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 7.026,
      "planned_context_bytes": 8200,
      "context_samples_bytes": [
        8200
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "roleless_click",
      "arm": "baseline",
      "repeat": 2,
      "passed": false,
      "probes": [
        {
          "noninteractive_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 1,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 4.83,
      "planned_context_bytes": 2840,
      "context_samples_bytes": [
        2840
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "roleless_click",
      "arm": "treatment",
      "repeat": 2,
      "passed": true,
      "probes": [
        {
          "noninteractive_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 1,
      "whole_operation_ms": 10.098,
      "planned_context_bytes": 5242,
      "context_samples_bytes": [
        5242
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "transparent_toggle",
      "arm": "baseline",
      "repeat": 2,
      "passed": false,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 1,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 5.536,
      "planned_context_bytes": 2845,
      "context_samples_bytes": [
        2845
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "transparent_toggle",
      "arm": "treatment",
      "repeat": 2,
      "passed": true,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 1,
      "whole_operation_ms": 11.594,
      "planned_context_bytes": 5319,
      "context_samples_bytes": [
        5319
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "nested_scroll",
      "arm": "baseline",
      "repeat": 2,
      "passed": false,
      "probes": [],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 1,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 1,
      "executed_observed_actions": 0,
      "whole_operation_ms": 5.664,
      "planned_context_bytes": 2873,
      "context_samples_bytes": [
        2873
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "nested_scroll",
      "arm": "treatment",
      "repeat": 2,
      "passed": true,
      "probes": [],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 0,
      "freshness_checks": 0,
      "observations": 6,
      "executed_observed_actions": 6,
      "whole_operation_ms": 25.24,
      "planned_context_bytes": 19925,
      "context_samples_bytes": [
        2909,
        2926,
        2926,
        2926,
        2926,
        5312
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "freshness_occlusion",
      "arm": "baseline",
      "repeat": 2,
      "passed": true,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 2,
      "freshness_checks": 2,
      "observations": 2,
      "executed_observed_actions": 0,
      "whole_operation_ms": 8.345,
      "planned_context_bytes": 10504,
      "context_samples_bytes": [
        5252,
        5252
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    },
    {
      "case": "freshness_occlusion",
      "arm": "treatment",
      "repeat": 2,
      "passed": true,
      "probes": [
        {
          "disabled_excluded": true
        }
      ],
      "typed_state_recall": null,
      "missing_or_mismatched_controls": 0,
      "wrong_actions": 0,
      "freshness_rejections": 2,
      "freshness_checks": 2,
      "observations": 2,
      "executed_observed_actions": 0,
      "whole_operation_ms": 7.562,
      "planned_context_bytes": 10504,
      "context_samples_bytes": [
        5252,
        5252
      ],
      "model_usage": [],
      "model_requests": 0,
      "inference_cost_usd": 0
    }
  ]
}
