{
  "schema_version": "local-compare-1.0",
  "outcome_source": "local-verify",
  "verification": {
    "completed": 2,
    "passed": 2,
    "exit_code_counts": {
      "0": 2
    },
    "arms": {
      "before": {
        "completed": 1,
        "passed": 1,
        "exit_code_counts": {
          "0": 1
        }
      },
      "after": {
        "completed": 1,
        "passed": 1,
        "exit_code_counts": {
          "0": 1
        }
      }
    }
  },
  "signals": [],
  "registration": {
    "outcome_source": "local-verify",
    "intervention_id": "intervention_05573134be99472b9fcbacf7",
    "applied_at": "2030-01-02T00:00:00+00:00",
    "registered_at": "2029-12-31T00:00:00+00:00",
    "preregistered": true,
    "predictions": [
      {
        "metric": "tokens_per_success",
        "direction": "decrease",
        "rough_size_percent": 50
      }
    ],
    "non_inferiority_margin_pp": 50,
    "windows": {
      "before": {
        "since": "2030-01-01T00:00:00+00:00",
        "until": "2030-01-02T00:00:00+00:00",
        "bounds": "[since,until)"
      },
      "after": {
        "since": "2030-01-02T00:00:00+00:00",
        "until": "2030-01-03T00:00:00+00:00",
        "bounds": "[since,until)"
      }
    }
  },
  "sample_size": {
    "registered_per_arm": 1,
    "needed_per_arm": 0,
    "actual": {
      "before": 1,
      "after": 1
    }
  },
  "bootstrap": {
    "seed": 1729,
    "resamples": 5000,
    "unit": "worker_session"
  },
  "arms": {
    "before": {
      "n": 1,
      "success_rate": {
        "numerator": 1,
        "denominator": 1,
        "rate": 1.0,
        "wilson_95": [
          0.20654931437723745,
          1.0
        ]
      },
      "cost": {
        "new_input": {
          "numerator": 100,
          "denominator": 1,
          "value": 100.0,
          "interval_95": [
            100.0,
            100.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "cache_write": {
          "numerator": 0,
          "denominator": 1,
          "value": 0.0,
          "interval_95": [
            0.0,
            0.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "cache_read": {
          "numerator": 20,
          "denominator": 1,
          "value": 20.0,
          "interval_95": [
            20.0,
            20.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "output": {
          "numerator": 10,
          "denominator": 1,
          "value": 10.0,
          "interval_95": [
            10.0,
            10.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "reasoning_output": {
          "numerator": 2,
          "denominator": 1,
          "value": 2.0,
          "interval_95": [
            2.0,
            2.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "total": {
          "numerator": 130,
          "denominator": 1,
          "value": 130.0,
          "interval_95": [
            130.0,
            130.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "time": {
          "numerator": 6.0,
          "denominator": 1,
          "value": 6.0,
          "interval_95": [
            6.0,
            6.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        }
      },
      "candidate_states": {
        "success": 1,
        "failed": 0,
        "unverified": 0,
        "in_progress": 0,
        "no_change": 0
      },
      "units": [
        {
          "id": "session_76eb20bb68f2e2a16e7fa740",
          "agent": "sumbi-events:agent_9e260e54ae0e8fd149c80532",
          "parent_id": "session_76adbe992952c2c7b488f1be",
          "dispatched_at": "2030-01-01T01:00:00+00:00",
          "last_at": "2030-01-01T01:00:06+00:00",
          "start_scope": "repository_root",
          "state": "success",
          "reason": "last_check_passed",
          "task_type": null,
          "tokens": {
            "new_input": 100,
            "cache_write": 0,
            "cache_read": 20,
            "output": 10,
            "reasoning_output": 2,
            "total": 130,
            "evidence_incomplete": false
          },
          "tokens_complete": true,
          "observed_total": 130,
          "elapsed_seconds": 6.0,
          "model": [
            "model_99051c601e033380bf6cd425"
          ],
          "effort": [
            "effort_79ae4a0c1a7a14f3fbe5b7aa"
          ],
          "cli_version": [
            "version_951ecbfab63ab4a551bcdd71"
          ],
          "command_coverage": {
            "matched": 1,
            "unknown_execution_cwd": 0,
            "unknown_execution_start": 0,
            "unknown_exit_code": 0
          },
          "metadata_incomplete": {
            "model": false,
            "effort": false,
            "cli_version": false
          }
        }
      ]
    },
    "after": {
      "n": 1,
      "success_rate": {
        "numerator": 1,
        "denominator": 1,
        "rate": 1.0,
        "wilson_95": [
          0.20654931437723745,
          1.0
        ]
      },
      "cost": {
        "new_input": {
          "numerator": 100,
          "denominator": 1,
          "value": 100.0,
          "interval_95": [
            100.0,
            100.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "cache_write": {
          "numerator": 0,
          "denominator": 1,
          "value": 0.0,
          "interval_95": [
            0.0,
            0.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "cache_read": {
          "numerator": 20,
          "denominator": 1,
          "value": 20.0,
          "interval_95": [
            20.0,
            20.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "output": {
          "numerator": 10,
          "denominator": 1,
          "value": 10.0,
          "interval_95": [
            10.0,
            10.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "reasoning_output": {
          "numerator": 2,
          "denominator": 1,
          "value": 2.0,
          "interval_95": [
            2.0,
            2.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "total": {
          "numerator": 130,
          "denominator": 1,
          "value": 130.0,
          "interval_95": [
            130.0,
            130.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        },
        "time": {
          "numerator": 6.0,
          "denominator": 1,
          "value": 6.0,
          "interval_95": [
            6.0,
            6.0
          ],
          "interval_method": "percentile_bootstrap",
          "defined_resamples": 5000
        }
      },
      "candidate_states": {
        "success": 1,
        "failed": 0,
        "unverified": 0,
        "in_progress": 0,
        "no_change": 0
      },
      "units": [
        {
          "id": "session_d44acaa7b2346dd0f370478e",
          "agent": "sumbi-events:agent_9e260e54ae0e8fd149c80532",
          "parent_id": "session_f176352aa2bc1aeb178d930e",
          "dispatched_at": "2030-01-02T01:00:00+00:00",
          "last_at": "2030-01-02T01:00:06+00:00",
          "start_scope": "repository_root",
          "state": "success",
          "reason": "last_check_passed",
          "task_type": null,
          "tokens": {
            "new_input": 100,
            "cache_write": 0,
            "cache_read": 20,
            "output": 10,
            "reasoning_output": 2,
            "total": 130,
            "evidence_incomplete": false
          },
          "tokens_complete": true,
          "observed_total": 130,
          "elapsed_seconds": 6.0,
          "model": [
            "model_99051c601e033380bf6cd425"
          ],
          "effort": [
            "effort_79ae4a0c1a7a14f3fbe5b7aa"
          ],
          "cli_version": [
            "version_951ecbfab63ab4a551bcdd71"
          ],
          "command_coverage": {
            "matched": 1,
            "unknown_execution_cwd": 0,
            "unknown_execution_start": 0,
            "unknown_exit_code": 0
          },
          "metadata_incomplete": {
            "model": false,
            "effort": false,
            "cli_version": false
          }
        }
      ]
    }
  },
  "success_difference": {
    "before": {
      "numerator": 1,
      "denominator": 1,
      "rate": 1.0,
      "wilson_95": [
        0.20654931437723745,
        1.0
      ]
    },
    "after": {
      "numerator": 1,
      "denominator": 1,
      "rate": 1.0,
      "wilson_95": [
        0.20654931437723745,
        1.0
      ]
    },
    "difference": 0.0,
    "interval_95": [
      -0.7934506856227626,
      0.7934506856227626
    ],
    "interval_method": "newcombe_method_10",
    "non_inferiority": "inconclusive"
  },
  "ratios": {
    "new_input": {
      "numerator": 100.0,
      "denominator": 100.0,
      "value": 1.0,
      "interval_95": [
        1.0,
        1.0
      ],
      "interval_method": "percentile_bootstrap",
      "defined_resamples": 5000,
      "classification": "uncertain"
    },
    "cache_write": {
      "numerator": 0.0,
      "denominator": 0.0,
      "value": null,
      "interval_95": null,
      "interval_method": "percentile_bootstrap",
      "defined_resamples": 0,
      "classification": "uncertain"
    },
    "cache_read": {
      "numerator": 20.0,
      "denominator": 20.0,
      "value": 1.0,
      "interval_95": [
        1.0,
        1.0
      ],
      "interval_method": "percentile_bootstrap",
      "defined_resamples": 5000,
      "classification": "uncertain"
    },
    "output": {
      "numerator": 10.0,
      "denominator": 10.0,
      "value": 1.0,
      "interval_95": [
        1.0,
        1.0
      ],
      "interval_method": "percentile_bootstrap",
      "defined_resamples": 5000,
      "classification": "uncertain"
    },
    "reasoning_output": {
      "numerator": 2.0,
      "denominator": 2.0,
      "value": 1.0,
      "interval_95": [
        1.0,
        1.0
      ],
      "interval_method": "percentile_bootstrap",
      "defined_resamples": 5000,
      "classification": "uncertain"
    },
    "total": {
      "numerator": 130.0,
      "denominator": 130.0,
      "value": 1.0,
      "interval_95": [
        1.0,
        1.0
      ],
      "interval_method": "percentile_bootstrap",
      "defined_resamples": 5000,
      "classification": "uncertain"
    },
    "time": {
      "numerator": 6.0,
      "denominator": 6.0,
      "value": 1.0,
      "interval_95": [
        1.0,
        1.0
      ],
      "interval_method": "percentile_bootstrap",
      "defined_resamples": 5000,
      "classification": "uncertain"
    }
  },
  "exclusions": {
    "before": {
      "share": {
        "numerator": 0,
        "denominator": 1,
        "value": 0.0,
        "interval_95": null,
        "interval_method": "not_applicable_census"
      },
      "risky_share": {
        "numerator": 0,
        "denominator": 1,
        "value": 0.0,
        "interval_95": null,
        "interval_method": "not_applicable_census"
      },
      "counts": {},
      "units": []
    },
    "after": {
      "share": {
        "numerator": 0,
        "denominator": 1,
        "value": 0.0,
        "interval_95": null,
        "interval_method": "not_applicable_census"
      },
      "risky_share": {
        "numerator": 0,
        "denominator": 1,
        "value": 0.0,
        "interval_95": null,
        "interval_method": "not_applicable_census"
      },
      "counts": {},
      "units": []
    }
  },
  "mixes": {
    "agent": {
      "before": {
        "values": {
          "sumbi-events:agent_9e260e54ae0e8fd149c80532": {
            "numerator": 1,
            "denominator": 1,
            "value": 1.0,
            "interval_95": null,
            "interval_method": "not_applicable_census"
          }
        },
        "dominant": [
          "sumbi-events:agent_9e260e54ae0e8fd149c80532"
        ]
      },
      "after": {
        "values": {
          "sumbi-events:agent_9e260e54ae0e8fd149c80532": {
            "numerator": 1,
            "denominator": 1,
            "value": 1.0,
            "interval_95": null,
            "interval_method": "not_applicable_census"
          }
        },
        "dominant": [
          "sumbi-events:agent_9e260e54ae0e8fd149c80532"
        ]
      },
      "total_variation_distance": 0.0,
      "dominant_changed": false
    },
    "model": {
      "before": {
        "values": {
          "model_99051c601e033380bf6cd425": {
            "numerator": 1,
            "denominator": 1,
            "value": 1.0,
            "interval_95": null,
            "interval_method": "not_applicable_census"
          }
        },
        "dominant": [
          "model_99051c601e033380bf6cd425"
        ]
      },
      "after": {
        "values": {
          "model_99051c601e033380bf6cd425": {
            "numerator": 1,
            "denominator": 1,
            "value": 1.0,
            "interval_95": null,
            "interval_method": "not_applicable_census"
          }
        },
        "dominant": [
          "model_99051c601e033380bf6cd425"
        ]
      },
      "total_variation_distance": 0.0,
      "dominant_changed": false
    },
    "effort": {
      "before": {
        "values": {
          "effort_79ae4a0c1a7a14f3fbe5b7aa": {
            "numerator": 1,
            "denominator": 1,
            "value": 1.0,
            "interval_95": null,
            "interval_method": "not_applicable_census"
          }
        },
        "dominant": [
          "effort_79ae4a0c1a7a14f3fbe5b7aa"
        ]
      },
      "after": {
        "values": {
          "effort_79ae4a0c1a7a14f3fbe5b7aa": {
            "numerator": 1,
            "denominator": 1,
            "value": 1.0,
            "interval_95": null,
            "interval_method": "not_applicable_census"
          }
        },
        "dominant": [
          "effort_79ae4a0c1a7a14f3fbe5b7aa"
        ]
      },
      "total_variation_distance": 0.0,
      "dominant_changed": false
    },
    "cli_version": {
      "before": {
        "values": {
          "version_951ecbfab63ab4a551bcdd71": {
            "numerator": 1,
            "denominator": 1,
            "value": 1.0,
            "interval_95": null,
            "interval_method": "not_applicable_census"
          }
        },
        "dominant": [
          "version_951ecbfab63ab4a551bcdd71"
        ]
      },
      "after": {
        "values": {
          "version_951ecbfab63ab4a551bcdd71": {
            "numerator": 1,
            "denominator": 1,
            "value": 1.0,
            "interval_95": null,
            "interval_method": "not_applicable_census"
          }
        },
        "dominant": [
          "version_951ecbfab63ab4a551bcdd71"
        ]
      },
      "total_variation_distance": 0.0,
      "dominant_changed": false
    }
  },
  "flags": [
    {
      "name": "dispatch_overhead_excluded",
      "blocking": false,
      "evidence": {
        "sessions": 2,
        "observed_total": 80
      }
    }
  ],
  "coverage": {
    "adapters": {
      "sumbi-events": {
        "files_scanned": 3,
        "lines_read": 22,
        "broken_lines": 0,
        "duplicate_events": 2,
        "unreadable_files": 0,
        "invalid_timestamps": 0,
        "invalid_token_records": 0,
        "inherited_session_meta": 0,
        "unknown_record_types": {},
        "sessions_read": 4
      }
    },
    "commands": {
      "matched": 2,
      "unknown_execution_cwd": 0,
      "unknown_execution_start": 0,
      "unknown_exit_code": 0
    },
    "start_scope_counts": {
      "repository_root": 2
    },
    "excluded_scope": {},
    "incomplete_reasons": []
  },
  "dispatch_overhead": {
    "sessions": 2,
    "rows": [
      {
        "id": "session_f176352aa2bc1aeb178d930e",
        "agent": "sumbi-events:agent_dafcddd422574acb5f48fb2b",
        "tokens": {
          "new_input": 30,
          "cache_write": 0,
          "cache_read": 0,
          "output": 10,
          "reasoning_output": null,
          "total": 40,
          "evidence_incomplete": false
        },
        "tokens_complete": true,
        "observed_total": 40
      },
      {
        "id": "session_76adbe992952c2c7b488f1be",
        "agent": "sumbi-events:agent_dafcddd422574acb5f48fb2b",
        "tokens": {
          "new_input": 30,
          "cache_write": 0,
          "cache_read": 0,
          "output": 10,
          "reasoning_output": null,
          "total": 40,
          "evidence_incomplete": false
        },
        "tokens_complete": true,
        "observed_total": 40
      }
    ],
    "observed_total": 80
  },
  "excluded_worker_spend": {
    "observed_total": 0
  },
  "cost_basis": "retained_worker_lifetime_only",
  "scan_until": "2030-01-04T00:00:00+00:00",
  "pseudonyms": {
    "method": "hmac_sha256",
    "salted": true,
    "local_only": true
  },
  "active_minutes": 5.0,
  "limitations": [
    "agent_triggered_machine_check",
    "no_human_acceptance_or_revert_window",
    "shell_edits_invisible",
    "worker_costs_exclude_dispatch_overhead",
    "last_observed_time_is_not_acceptance",
    "deleted_logs_undetectable"
  ],
  "decision_order": [
    "coverage",
    "comparability",
    "success_non_inferiority",
    "time_and_tokens"
  ],
  "verdict": {
    "label": "proposal",
    "proposal": "withhold",
    "reasons": [
      "inconclusive_success"
    ]
  }
}
