{
 "active": {
  "raw_sha256": "f12e7c80778de1e92388c0443e34bdedaee14b2abe0b2867141cc944e0d15a14",
  "shift_negative_transfer": {
   "ident_memory": 26,
   "ident_plain": 25,
   "memory_better": 23,
   "memory_worse": 13,
   "p": 0.13249816396273673
  },
  "summary": {
   "adversarial/eig": {
    "experiments": {
     "ci95": [
      16.243461,
      22.289872
     ],
     "mean": 19.266667,
     "n": 30
    },
    "identified": 28,
    "identified_rate": 0.9333,
    "n": 30,
    "wilson": [
     0.7868,
     0.9815
    ]
   },
   "adversarial/random": {
    "experiments": {
     "ci95": [
      27.37958,
      34.42042
     ],
     "mean": 30.9,
     "n": 30
    },
    "identified": 28,
    "identified_rate": 0.9333,
    "n": 30,
    "wilson": [
     0.7868,
     0.9815
    ]
   },
   "sealed/eig": {
    "experiments": {
     "ci95": [
      9.786283,
      10.780383
     ],
     "mean": 10.283333,
     "n": 120
    },
    "identified": 119,
    "identified_rate": 0.9917,
    "n": 120,
    "wilson": [
     0.9543,
     0.9985
    ]
   },
   "sealed/eig_memory": {
    "experiments": {
     "ci95": [
      9.541922,
      10.458078
     ],
     "mean": 10.0,
     "n": 120
    },
    "identified": 116,
    "identified_rate": 0.9667,
    "n": 120,
    "wilson": [
     0.9174,
     0.987
    ]
   },
   "sealed/eig_poisoned_memory": {
    "experiments": {
     "ci95": [
      9.35023,
      11.016437
     ],
     "mean": 10.183333,
     "n": 120
    },
    "identified": 83,
    "identified_rate": 0.6917,
    "n": 120,
    "wilson": [
     0.6042,
     0.7673
    ]
   },
   "sealed/random": {
    "experiments": {
     "ci95": [
      17.205889,
      19.760778
     ],
     "mean": 18.483333,
     "n": 120
    },
    "identified": 119,
    "identified_rate": 0.9917,
    "n": 120,
    "wilson": [
     0.9543,
     0.9985
    ]
   },
   "sealed/sweep": {
    "experiments": {
     "ci95": [
      34.101602,
      37.265064
     ],
     "mean": 35.683333,
     "n": 120
    },
    "identified": 86,
    "identified_rate": 0.7167,
    "n": 120,
    "wilson": [
     0.6303,
     0.7896
    ]
   },
   "shift/eig": {
    "experiments": {
     "ci95": [
      17.992003,
      24.774664
     ],
     "mean": 21.383333,
     "n": 60
    },
    "identified": 25,
    "identified_rate": 0.4167,
    "n": 60,
    "wilson": [
     0.3006,
     0.5427
    ]
   },
   "shift/eig_memory": {
    "experiments": {
     "ci95": [
      16.825336,
      23.607997
     ],
     "mean": 20.216667,
     "n": 60
    },
    "identified": 26,
    "identified_rate": 0.4333,
    "n": 60,
    "wilson": [
     0.3157,
     0.559
    ]
   },
   "transfer/eig": {
    "experiments": {
     "ci95": [
      8.315502,
      9.651164
     ],
     "mean": 8.983333,
     "n": 60
    },
    "identified": 59,
    "identified_rate": 0.9833,
    "n": 60,
    "wilson": [
     0.9114,
     0.9971
    ]
   },
   "transfer/eig_memory": {
    "experiments": {
     "ci95": [
      8.018336,
      9.314998
     ],
     "mean": 8.666667,
     "n": 60
    },
    "identified": 58,
    "identified_rate": 0.9667,
    "n": 60,
    "wilson": [
     0.8864,
     0.9908
    ]
   },
   "transfer/random": {
    "experiments": {
     "ci95": [
      15.822671,
      19.310662
     ],
     "mean": 17.566667,
     "n": 60
    },
    "identified": 58,
    "identified_rate": 0.9667,
    "n": 60,
    "wilson": [
     0.8864,
     0.9908
    ]
   }
  },
  "verdicts": {
   "eig<random@sealed": {
    "body": {
     "checks": {
      "fewer_experiments": true,
      "matches_claimed_mean_a": true,
      "matches_claimed_mean_b": true,
      "no_identification_loss": true,
      "significant_p_lt_0.01": true
     },
     "claim": "active:eig<random:sealed",
     "raw_sha256": "f12e7c80778de1e92388c0443e34bdedaee14b2abe0b2867141cc944e0d15a14",
     "recomputed": {
      "ident_a": 0.9916666666666667,
      "ident_b": 0.9916666666666667,
      "losses": 4,
      "mean_a": 10.283333333333333,
      "mean_b": 18.483333333333334,
      "n": 120,
      "p": 6.663617459222728e-28,
      "wins": 110
     },
     "schema": "cain.e114.verdict.v1",
     "verdict": "CONFIRMED"
    },
    "pub": "dh2HhSlgp/Dqjxx/7NUqXPRpC4lpIgYofDqsGxN/I9Q=",
    "sig": "co10q+ymWL9yEeosd4qUUsxfCs/hpDAcZyf1x77k7IjBrwl8BapnMtTFNrYYn0sUgTO67WEEXKWiDxoYTRKKDw=="
   },
   "eig_memory<eig@sealed": {
    "body": {
     "checks": {
      "fewer_experiments": true,
      "matches_claimed_mean_a": true,
      "matches_claimed_mean_b": true,
      "no_identification_loss": false,
      "significant_p_lt_0.01": false
     },
     "claim": "active:eig_memory<eig:sealed",
     "raw_sha256": "f12e7c80778de1e92388c0443e34bdedaee14b2abe0b2867141cc944e0d15a14",
     "recomputed": {
      "ident_a": 0.9666666666666667,
      "ident_b": 0.9916666666666667,
      "losses": 21,
      "mean_a": 10.0,
      "mean_b": 10.283333333333333,
      "n": 120,
      "p": 0.16897782228771874,
      "wins": 32
     },
     "schema": "cain.e114.verdict.v1",
     "verdict": "REJECTED"
    },
    "pub": "dh2HhSlgp/Dqjxx/7NUqXPRpC4lpIgYofDqsGxN/I9Q=",
    "sig": "vOxQmabRmFV6NK2cDZk3LhmfbekuRlk173vo9EF67oWDdxghv9157BWr4yhfn9NB2MveGWxYIoUyR1jtApk1DA=="
   },
   "eig_memory<eig@transfer": {
    "body": {
     "checks": {
      "fewer_experiments": true,
      "matches_claimed_mean_a": true,
      "matches_claimed_mean_b": true,
      "no_identification_loss": true,
      "significant_p_lt_0.01": true
     },
     "claim": "active:eig_memory<eig:transfer",
     "raw_sha256": "f12e7c80778de1e92388c0443e34bdedaee14b2abe0b2867141cc944e0d15a14",
     "recomputed": {
      "ident_a": 0.9666666666666667,
      "ident_b": 0.9833333333333333,
      "losses": 6,
      "mean_a": 8.666666666666666,
      "mean_b": 8.983333333333333,
      "n": 60,
      "p": 0.009355306625366211,
      "wins": 20
     },
     "schema": "cain.e114.verdict.v1",
     "verdict": "CONFIRMED"
    },
    "pub": "dh2HhSlgp/Dqjxx/7NUqXPRpC4lpIgYofDqsGxN/I9Q=",
    "sig": "mdiIHd7d+iBhOdgBlVyypHoDsPUUPVY96noRupLJBKIyLBEiKz6JghWENdxqLpRqiRUIDL0Zx0f27Lup/ZOQCg=="
   }
  }
 },
 "backlog_memory_commits": [
  {
   "first": "COMMITTED",
   "item": "B3"
  },
  {
   "first": "COMMITTED",
   "item": "B2"
  },
  {
   "first": "COMMITTED",
   "item": "B5"
  },
  {
   "first": "COMMITTED",
   "item": "B1"
  },
  {
   "first": "COMMITTED",
   "item": "B4"
  },
  {
   "after_citing_source": "COMMITTED",
   "first": "FABRICATED_NUMBER:24",
   "item": "B6"
  }
 ],
 "continual": {
  "dev_summary": {
   "experts": {
    "bwt": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 5
    },
    "cpu_s": {
     "ci95": [
      0.386369,
      0.420031
     ],
     "mean": 0.4032,
     "n": 5
    },
    "final_mean_acc": {
     "ci95": [
      0.896675,
      0.934725
     ],
     "mean": 0.9157,
     "n": 5
    },
    "inference_distance_ops": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 5
    },
    "mean_forgetting": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 5
    },
    "memory_bytes": {
     "ci95": [
      26880.0,
      26880.0
     ],
     "mean": 26880.0,
     "n": 5
    },
    "train_sample_steps": {
     "ci95": [
      90000.0,
      90000.0
     ],
     "mean": 90000.0,
     "n": 5
    },
    "zero_shot_mean": {
     "ci95": [
      0.519237,
      0.575013
     ],
     "mean": 0.547125,
     "n": 5
    }
   },
   "knn_memory": {
    "bwt": {
     "ci95": [
      -0.000589,
      0.000839
     ],
     "mean": 0.000125,
     "n": 5
    },
    "cpu_s": {
     "ci95": [
      1.47635,
      1.59445
     ],
     "mean": 1.5354,
     "n": 5
    },
    "final_mean_acc": {
     "ci95": [
      0.795947,
      0.803653
     ],
     "mean": 0.7998,
     "n": 5
    },
    "inference_distance_ops": {
     "ci95": [
      7800000.0,
      7800000.0
     ],
     "mean": 7800000.0,
     "n": 5
    },
    "mean_forgetting": {
     "ci95": [
      -0.000333,
      0.000583
     ],
     "mean": 0.000125,
     "n": 5
    },
    "memory_bytes": {
     "ci95": [
      252000.0,
      252000.0
     ],
     "mean": 252000.0,
     "n": 5
    },
    "train_sample_steps": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 5
    },
    "zero_shot_mean": {
     "ci95": [
      0.501901,
      0.559349
     ],
     "mean": 0.530625,
     "n": 5
    }
   },
   "naive": {
    "bwt": {
     "ci95": [
      -0.344198,
      -0.234552
     ],
     "mean": -0.289375,
     "n": 5
    },
    "cpu_s": {
     "ci95": [
      0.392286,
      0.422514
     ],
     "mean": 0.4074,
     "n": 5
    },
    "final_mean_acc": {
     "ci95": [
      0.645869,
      0.715931
     ],
     "mean": 0.6809,
     "n": 5
    },
    "inference_distance_ops": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 5
    },
    "mean_forgetting": {
     "ci95": [
      0.234552,
      0.344198
     ],
     "mean": 0.289375,
     "n": 5
    },
    "memory_bytes": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 5
    },
    "train_sample_steps": {
     "ci95": [
      90000.0,
      90000.0
     ],
     "mean": 90000.0,
     "n": 5
    },
    "zero_shot_mean": {
     "ci95": [
      0.530856,
      0.596894
     ],
     "mean": 0.563875,
     "n": 5
    }
   },
   "replay": {
    "bwt": {
     "ci95": [
      -0.125577,
      -0.072173
     ],
     "mean": -0.098875,
     "n": 5
    },
    "cpu_s": {
     "ci95": [
      0.51517,
      0.56003
     ],
     "mean": 0.5376,
     "n": 5
    },
    "final_mean_acc": {
     "ci95": [
      0.80692,
      0.83848
     ],
     "mean": 0.8227,
     "n": 5
    },
    "inference_distance_ops": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 5
    },
    "mean_forgetting": {
     "ci95": [
      0.072662,
      0.125588
     ],
     "mean": 0.099125,
     "n": 5
    },
    "memory_bytes": {
     "ci95": [
      50400.0,
      50400.0
     ],
     "mean": 50400.0,
     "n": 5
    },
    "train_sample_steps": {
     "ci95": [
      126000.0,
      126000.0
     ],
     "mean": 126000.0,
     "n": 5
    },
    "zero_shot_mean": {
     "ci95": [
      0.537017,
      0.583233
     ],
     "mean": 0.560125,
     "n": 5
    }
   }
  },
  "outer_loop_choice": {
   "chosen": "experts",
   "eligible": [
    "experts",
    "knn_memory"
   ],
   "rule": "max final accuracy s.t. mean forgetting <= 0.05; tie -> less CPU"
  },
  "raw_sha256": "3b24faf6ef41fd683cc019336a6c99de6e244bc2ee30b0423d7a1d4afd097858",
  "sealed_confirmation": {
   "beats_naive_seeds": 10,
   "chosen": "experts",
   "of": 10,
   "p": 0.001953125
  },
  "sealed_summary": {
   "experts": {
    "bwt": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 10
    },
    "cpu_s": {
     "ci95": [
      0.400491,
      0.442509
     ],
     "mean": 0.4215,
     "n": 10
    },
    "final_mean_acc": {
     "ci95": [
      0.910722,
      0.927278
     ],
     "mean": 0.919,
     "n": 10
    },
    "inference_distance_ops": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 10
    },
    "mean_forgetting": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 10
    },
    "memory_bytes": {
     "ci95": [
      26880.0,
      26880.0
     ],
     "mean": 26880.0,
     "n": 10
    },
    "train_sample_steps": {
     "ci95": [
      90000.0,
      90000.0
     ],
     "mean": 90000.0,
     "n": 10
    },
    "zero_shot_mean": {
     "ci95": [
      0.543664,
      0.601461
     ],
     "mean": 0.572562,
     "n": 10
    }
   },
   "knn_memory": {
    "bwt": {
     "ci95": [
      -0.000297,
      0.000672
     ],
     "mean": 0.000187,
     "n": 10
    },
    "cpu_s": {
     "ci95": [
      1.512052,
      1.556748
     ],
     "mean": 1.5344,
     "n": 10
    },
    "final_mean_acc": {
     "ci95": [
      0.790673,
      0.805227
     ],
     "mean": 0.79795,
     "n": 10
    },
    "inference_distance_ops": {
     "ci95": [
      7800000.0,
      7800000.0
     ],
     "mean": 7800000.0,
     "n": 10
    },
    "mean_forgetting": {
     "ci95": [
      -0.000526,
      0.000401
     ],
     "mean": -6.2e-05,
     "n": 10
    },
    "memory_bytes": {
     "ci95": [
      252000.0,
      252000.0
     ],
     "mean": 252000.0,
     "n": 10
    },
    "train_sample_steps": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 10
    },
    "zero_shot_mean": {
     "ci95": [
      0.521488,
      0.551887
     ],
     "mean": 0.536687,
     "n": 10
    }
   },
   "naive": {
    "bwt": {
     "ci95": [
      -0.322957,
      -0.249793
     ],
     "mean": -0.286375,
     "n": 10
    },
    "cpu_s": {
     "ci95": [
      0.384546,
      0.406654
     ],
     "mean": 0.3956,
     "n": 10
    },
    "final_mean_acc": {
     "ci95": [
      0.656411,
      0.711989
     ],
     "mean": 0.6842,
     "n": 10
    },
    "inference_distance_ops": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 10
    },
    "mean_forgetting": {
     "ci95": [
      0.249793,
      0.322957
     ],
     "mean": 0.286375,
     "n": 10
    },
    "memory_bytes": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 10
    },
    "train_sample_steps": {
     "ci95": [
      90000.0,
      90000.0
     ],
     "mean": 90000.0,
     "n": 10
    },
    "zero_shot_mean": {
     "ci95": [
      0.55831,
      0.607065
     ],
     "mean": 0.582688,
     "n": 10
    }
   },
   "replay": {
    "bwt": {
     "ci95": [
      -0.109423,
      -0.084952
     ],
     "mean": -0.097188,
     "n": 10
    },
    "cpu_s": {
     "ci95": [
      0.531311,
      0.562089
     ],
     "mean": 0.5467,
     "n": 10
    },
    "final_mean_acc": {
     "ci95": [
      0.813734,
      0.832266
     ],
     "mean": 0.823,
     "n": 10
    },
    "inference_distance_ops": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 10
    },
    "mean_forgetting": {
     "ci95": [
      0.084952,
      0.109423
     ],
     "mean": 0.097188,
     "n": 10
    },
    "memory_bytes": {
     "ci95": [
      50400.0,
      50400.0
     ],
     "mean": 50400.0,
     "n": 10
    },
    "train_sample_steps": {
     "ci95": [
      126000.0,
      126000.0
     ],
     "mean": 126000.0,
     "n": 10
    },
    "zero_shot_mean": {
     "ci95": [
      0.560393,
      0.599982
     ],
     "mean": 0.580187,
     "n": 10
    }
   }
  },
  "training_records": 40
 },
 "cycle2": {
  "B1_grammar_growth": {
   "breakdown": {
    "adversarial/in_class": {
     "causal": 0.1506,
     "grown": 0.1394,
     "hybrid": 0.1287,
     "mlp": 0.1278,
     "n": 30,
     "uses": {
      "causal": 17,
      "grown": 2,
      "mlp_after_growth_failed": 11
     }
    },
    "sealed/in_class": {
     "causal": 0.0283,
     "grown": 0.0298,
     "hybrid": 0.028,
     "mlp": 0.0537,
     "n": 120,
     "uses": {
      "causal": 115,
      "grown": 3,
      "mlp_after_growth_failed": 2
     }
    },
    "shift/in_class": {
     "causal": 0.056,
     "grown": 0.0575,
     "hybrid": 0.0512,
     "mlp": 0.075,
     "n": 28,
     "uses": {
      "causal": 26,
      "grown": 2
     }
    },
    "shift/out_class": {
     "causal": 0.1705,
     "grown": 0.1372,
     "hybrid": 0.1424,
     "mlp": 0.0743,
     "n": 32,
     "uses": {
      "causal": 21,
      "grown": 9,
      "mlp_after_growth_failed": 2
     }
    }
   },
   "disclosure": "the grown grammar contains the functional forms used by the shift generator (and3, maj3, and_not); this tests the detect-misspecification-then-grow mechanism, not open-ended discovery",
   "pipeline_decision": "RETAIN_BASELINE",
   "sealed_delta_brier": null,
   "stages": [
    [
     "1_hypothesis_review",
     true
    ],
    [
     "2_development",
     true
    ],
    [
     "3_validation",
     false
    ]
   ],
   "verdict": null,
   "why": "no validation improvement"
  },
  "B2_gated_skills": {
   "library": {
    "k0_sq_sort": [
     "sq",
     "sort"
    ],
    "k1_pos_cums": [
     "pos",
     "cums"
    ],
    "k2_sort_rev": [
     "sort",
     "rev"
    ],
    "k3_abs_sort": [
     "abs",
     "sort"
    ]
   },
   "summary": {
    "sealed": {
     "gated": 0.75,
     "gated_vs_none": {
      "losses": 0,
      "p": 0.00390625,
      "wins": 9
     },
     "gated_vs_ungated": {
      "losses": 0,
      "p": 0.25,
      "wins": 3
     },
     "none": 0.6,
     "ungated": 0.7,
     "ungated_vs_none": {
      "losses": 3,
      "p": 0.14599609375,
      "wins": 9
     }
    },
    "shift": {
     "gated": 0.35,
     "gated_vs_none": {
      "losses": 10,
      "p": 0.001953125,
      "wins": 0
     },
     "gated_vs_ungated": {
      "losses": 0,
      "p": 1.0,
      "wins": 0
     },
     "none": 0.6,
     "ungated": 0.35,
     "ungated_vs_none": {
      "losses": 10,
      "p": 0.001953125,
      "wins": 0
     }
    }
   },
   "support_sizes": {
    "k0_sq_sort": 2,
    "k1_pos_cums": 2,
    "k2_sort_rev": 2,
    "k3_abs_sort": 0
   },
   "verdict": "REJECTED",
   "verdict_checks": {
    "direction": false,
    "matches_claimed_delta": true,
    "significant_p_lt_0.05": false
   }
  },
  "director_after": {
   "H1": {
    "mean": 0.9569,
    "sd": 0.0188,
    "status": "SUPPORTED_HYPOTHESIS"
   },
   "H2": {
    "mean": 0.6,
    "sd": 0.0655,
    "status": "UNSUPPORTED_HYPOTHESIS"
   },
   "H3": {
    "mean": 0.75,
    "sd": 0.0804,
    "status": "SUPPORTED_HYPOTHESIS"
   },
   "H3s": {
    "mean": 0.5,
    "sd": 0.2887,
    "status": "UNSUPPORTED_HYPOTHESIS"
   },
   "H5": {
    "mean": 0.5,
    "sd": 0.2887,
    "status": "STALLED"
   },
   "H7": {
    "mean": 0.4715,
    "sd": 0.0448,
    "status": "UNSUPPORTED_HYPOTHESIS"
   },
   "H7b": {
    "mean": 0.3048,
    "sd": 0.0447,
    "status": "UNSUPPORTED_HYPOTHESIS"
   },
   "H8": {
    "mean": 0.5,
    "sd": 0.2887,
    "status": "UNSUPPORTED_HYPOTHESIS"
   },
   "H9": {
    "mean": 0.5,
    "sd": 0.2887,
    "status": "UNSUPPORTED_HYPOTHESIS"
   }
  }
 },
 "director": {
  "backlog": [
   {
    "approach": "ingest E113 raw rows",
    "cost": 1,
    "evidence": "G11 delegated to E113 (model-backed arms A-J), not yet published",
    "experiment": "re-verify E113 gate from raw rows with the E114 independent verifier",
    "failure": "they do not",
    "id": "B3",
    "importance": 0.9,
    "limitation": "no model-backed collective result inside E114",
    "priority_score": 0.81,
    "resources": "E113 publication",
    "risk": 0.0,
    "success": "E113 claims reproduce",
    "uncertainty": 0.9
   },
   {
    "approach": "context-gated skill invocation (only when task features match the skill's training support)",
    "cost": 1,
    "evidence": "{\"no_skills\": 0.6, \"with_skills\": 0.35, \"wins\": 0, \"losses\": 10, \"mean_evals_no_skills\": 1842.75, \"mean_evals_with_skills\": 1903.475}",
    "experiment": "gated vs ungated library on shift and sealed",
    "failure": "sealed gain lost",
    "id": "B2",
    "importance": 0.8,
    "limitation": "skills hurt on shifted task families",
    "priority_score": 0.51,
    "resources": "CPU, ~20 min",
    "risk": 0.05,
    "success": "no shift loss, sealed gain retained",
    "uncertainty": 0.7
   },
   {
    "approach": "shared-trunk + expert heads with replay",
    "cost": 1,
    "evidence": "zero-shot mean (experts) 0.573 ~ chance",
    "experiment": "trunk+heads vs experts, sealed seeds",
    "failure": "no transfer",
    "id": "B5",
    "importance": 0.6,
    "limitation": "continual learners show no forward transfer",
    "priority_score": 0.28,
    "resources": "CPU, ~10 min",
    "risk": 0.02,
    "success": "zero-shot > 0.55 with forgetting <= 0.05",
    "uncertainty": 0.5
   },
   {
    "approach": "learn rule grammar extensions (program synthesis over rule DSL) instead of fixed hypothesis class",
    "cost": 2,
    "evidence": "shift Brier causal 0.117 vs mlp 0.075",
    "experiment": "grammar-growth arm vs fixed grammar on sealed shift worlds",
    "failure": "no shift gain or in-class regression",
    "id": "B1",
    "importance": 0.9,
    "limitation": "world models degrade under out-of-class shift",
    "priority_score": 0.22,
    "resources": "CPU, ~1 h",
    "risk": 0.05,
    "success": "shift Brier of causal <= mlp with in-class Brier unchanged",
    "uncertainty": 0.6
   },
   {
    "approach": "run official protocols with a local judge after E113",
    "cost": 4,
    "evidence": "datasets + LLM judge not available offline; CPU shared with E113",
    "experiment": "AMA-Bench synthetic subset with official scorer",
    "failure": "protocol cannot be reproduced",
    "id": "B4",
    "importance": 0.8,
    "limitation": "external memory benchmarks not run (AMA-Bench, LongMemEval, LoCoMo)",
    "priority_score": 0.13,
    "resources": "dataset download, local model, several CPU hours",
    "risk": 0.05,
    "success": "official-protocol score reported with version",
    "uncertainty": 0.9
   },
   {
    "approach": "run the mission engine under a supervisor for 24 h",
    "cost": 3,
    "evidence": "mission elapsed 16.09 s",
    "experiment": "24 h governed mission with scheduled kills",
    "failure": "any duplicate effect or lost decision",
    "id": "B6",
    "importance": 0.7,
    "limitation": "missions run minutes, not 24 h",
    "priority_score": -0.0067,
    "resources": "24 h host time after E111 endurance finishes",
    "risk": 0.1,
    "success": "continuity checks hold after >=3 kills over 24 h",
    "uncertainty": 0.4
   }
  ],
  "hypotheses": {
   "H1": {
    "provenance": "director-proposed (not a finding)",
    "status": "SUPPORTED_HYPOTHESIS",
    "team": "team-a",
    "text": "EIG needs fewer experiments than random at equal identification"
   },
   "H2": {
    "provenance": "director-proposed (not a finding)",
    "status": "UNSUPPORTED_HYPOTHESIS",
    "team": "team-a",
    "text": "Validated family prior reduces experiments vs uniform-prior EIG"
   },
   "H3": {
    "provenance": "director-proposed (not a finding)",
    "status": "SUPPORTED_HYPOTHESIS",
    "team": "team-a",
    "text": "Memory prior transfers to 6-variable worlds"
   },
   "H3s": {
    "provenance": "director-proposed (not a finding)",
    "status": "UNSUPPORTED_HYPOTHESIS",
    "team": "team-a",
    "text": "Memory prior does not hurt on shifted worlds"
   },
   "H5": {
    "provenance": "director-proposed (not a finding)",
    "status": "STALLED",
    "team": "team-b",
    "text": "Selective specialization (experts) retains best per unit compute"
   },
   "H7": {
    "provenance": "director-proposed (not a finding)",
    "status": "UNSUPPORTED_HYPOTHESIS",
    "team": "team-a",
    "text": "Mixture hybrid lowers sealed Brier vs switch baseline"
   },
   "H7b": {
    "provenance": "director-proposed (not a finding)",
    "status": "UNSUPPORTED_HYPOTHESIS",
    "team": "team-a",
    "text": "Stricter switch threshold lowers sealed Brier"
   }
  },
  "log": [
   {
    "id": "O1",
    "op": "objective"
   },
   {
    "id": "O2",
    "op": "objective"
   },
   {
    "id": "O3",
    "op": "objective"
   },
   {
    "id": "O4",
    "op": "objective"
   },
   {
    "from": "EXPERIMENTAL_FINDING",
    "hypothesis": "H1",
    "op": "verdict",
    "to": "SUPPORTED_HYPOTHESIS"
   },
   {
    "from": "EXPERIMENTAL_FINDING",
    "hypothesis": "H2",
    "op": "verdict",
    "to": "UNSUPPORTED_HYPOTHESIS"
   },
   {
    "from": "EXPERIMENTAL_FINDING",
    "hypothesis": "H3",
    "op": "verdict",
    "to": "SUPPORTED_HYPOTHESIS"
   },
   {
    "from": "UNSUPPORTED_HYPOTHESIS",
    "hypothesis": "H7",
    "op": "verdict",
    "to": "UNSUPPORTED_HYPOTHESIS"
   },
   {
    "from": "UNSUPPORTED_HYPOTHESIS",
    "hypothesis": "H7b",
    "op": "verdict",
    "to": "UNSUPPORTED_HYPOTHESIS"
   },
   {
    "from": "team-a",
    "hypothesis": "H5",
    "op": "reassign",
    "reason": "posterior stalled",
    "to": "team-b"
   },
   {
    "from": "UNSUPPORTED_HYPOTHESIS",
    "hypothesis": "H8",
    "op": "verdict",
    "to": "UNSUPPORTED_HYPOTHESIS"
   },
   {
    "from": "team-b",
    "hypothesis": "H5",
    "op": "reassign",
    "reason": "posterior stalled",
    "to": "team-a"
   }
  ],
  "objectives": {
   "O1": {
    "id": "O1",
    "questions": [
     "Q1",
     "Q2",
     "Q3",
     "Q4",
     "Q9"
    ],
    "text": "Learn causal structure of unfamiliar environments with fewer experiments"
   },
   "O2": {
    "id": "O2",
    "questions": [
     "Q5"
    ],
    "text": "Learn continually without silently destroying earlier capability"
   },
   "O3": {
    "id": "O3",
    "questions": [
     "Q6",
     "Q8"
    ],
    "text": "Acquire reusable skills that help on held-out tasks and are rolled back when they hurt"
   },
   "O4": {
    "id": "O4",
    "questions": [
     "Q7"
    ],
    "text": "Improve the lab itself only through sealed, independently verified, authorized changes"
   }
  },
  "questions": {
   "Q1": {
    "depends_on": [],
    "hypotheses": [
     "H1"
    ],
    "id": "Q1",
    "objective": "O1",
    "text": "Does EIG experiment selection beat random selection on sealed worlds?"
   },
   "Q2": {
    "depends_on": [
     "Q1"
    ],
    "hypotheses": [
     "H2"
    ],
    "id": "Q2",
    "objective": "O1",
    "text": "Does a prior from validated memory of earlier worlds reduce experiments further?"
   },
   "Q3": {
    "depends_on": [
     "Q2"
    ],
    "hypotheses": [
     "H3",
     "H3s"
    ],
    "id": "Q3",
    "objective": "O1",
    "text": "Does that memory prior transfer to a different environment size, and does it hurt under shift?"
   },
   "Q4": {
    "depends_on": [],
    "hypotheses": [],
    "id": "Q4",
    "objective": "O1",
    "text": "Which world model predicts interventions best in- and out-of-distribution?"
   },
   "Q5": {
    "depends_on": [],
    "hypotheses": [
     "H5"
    ],
    "id": "Q5",
    "objective": "O2",
    "text": "Which continual learner retains best per unit compute?"
   },
   "Q6": {
    "depends_on": [],
    "hypotheses": [],
    "id": "Q6",
    "objective": "O3",
    "text": "Do mined skills raise held-out solve rate; does environment diversity matter?"
   },
   "Q7": {
    "depends_on": [
     "Q4"
    ],
    "hypotheses": [
     "H7",
     "H7b"
    ],
    "id": "Q7",
    "objective": "O4",
    "text": "Can a candidate configuration beat the frozen world-model baseline on sealed data?"
   },
   "Q8": {
    "depends_on": [
     "Q6"
    ],
    "hypotheses": [
     "H8"
    ],
    "id": "Q8",
    "objective": "O3",
    "text": "Does context-gating make a broader skill library safe under shift? (backlog B2)"
   },
   "Q9": {
    "depends_on": [
     "Q4"
    ],
    "hypotheses": [
     "H9"
    ],
    "id": "Q9",
    "objective": "O1",
    "text": "Does growing the rule grammar after misspecification fix shift prediction? (backlog B1)"
   }
  },
  "refusals": [
   {
    "hypothesis": "H1",
    "reason": "DIRECTOR_CANNOT_SELF_ACCEPT"
   },
   {
    "hypothesis": "H2",
    "reason": "DIRECTOR_CANNOT_SELF_ACCEPT"
   },
   {
    "hypothesis": "H3",
    "reason": "DIRECTOR_CANNOT_SELF_ACCEPT"
   },
   {
    "hypothesis": "H5",
    "reason": "VERDICT_NOT_FROM_INDEPENDENT_VERIFIER"
   }
  ],
  "stalled_and_reassigned": [
   "H5"
  ],
  "uncertainty": {
   "H1": {
    "mean": 0.9569,
    "sd": 0.0188,
    "status": "SUPPORTED_HYPOTHESIS"
   },
   "H2": {
    "mean": 0.6,
    "sd": 0.0655,
    "status": "UNSUPPORTED_HYPOTHESIS"
   },
   "H3": {
    "mean": 0.75,
    "sd": 0.0804,
    "status": "SUPPORTED_HYPOTHESIS"
   },
   "H3s": {
    "mean": 0.5,
    "sd": 0.2887,
    "status": "UNSUPPORTED_HYPOTHESIS"
   },
   "H5": {
    "mean": 0.5,
    "sd": 0.2887,
    "status": "STALLED"
   },
   "H7": {
    "mean": 0.4715,
    "sd": 0.0448,
    "status": "UNSUPPORTED_HYPOTHESIS"
   },
   "H7b": {
    "mean": 0.3048,
    "sd": 0.0447,
    "status": "UNSUPPORTED_HYPOTHESIS"
   }
  }
 },
 "envgen": {
  "leakage": {
   "rule_overlap_count": {
    "dev&adversarial": 15,
    "dev&sealed": 33,
    "dev&shift": 0,
    "dev&transfer": 0,
    "dev&val": 19,
    "dev&val_shift": 0,
    "sealed&adversarial": 17,
    "sealed&shift": 0,
    "sealed&transfer": 0,
    "shift&adversarial": 0,
    "transfer&adversarial": 0,
    "transfer&shift": 12,
    "val&adversarial": 11,
    "val&sealed": 25,
    "val&shift": 0,
    "val&transfer": 0,
    "val&val_shift": 0,
    "val_shift&adversarial": 0,
    "val_shift&sealed": 0,
    "val_shift&shift": 7,
    "val_shift&transfer": 6
   },
   "rule_overlap_note": "rule identities can recur across splits (the hypothesis space is finite); learners never see rule ids",
   "shift_out_of_class": 32,
   "shift_total": 60,
   "spec_disjoint": true,
   "spec_overlap": {
    "dev&adversarial": 0,
    "dev&sealed": 0,
    "dev&shift": 0,
    "dev&transfer": 0,
    "dev&val": 0,
    "dev&val_shift": 0,
    "sealed&adversarial": 0,
    "sealed&shift": 0,
    "sealed&transfer": 0,
    "shift&adversarial": 0,
    "transfer&adversarial": 0,
    "transfer&shift": 0,
    "val&adversarial": 0,
    "val&sealed": 0,
    "val&shift": 0,
    "val&transfer": 0,
    "val&val_shift": 0,
    "val_shift&adversarial": 0,
    "val_shift&sealed": 0,
    "val_shift&shift": 0,
    "val_shift&transfer": 0
   }
  },
  "sizes": {
   "adversarial": 30,
   "dev": 80,
   "sealed": 120,
   "shift": 60,
   "transfer": 60,
   "val": 40,
   "val_shift": 30
  },
  "world_specs_sha256": {
   "adversarial": "3336f0ae6e7cd7f2204dfb253e5500c8f757c69e4aeb0ecd0700fd951fd11070",
   "dev": "adf51ee8e7817dd178103a64fb98df876169715354e361d3a345fc3241880269",
   "sealed": "9a9c6b0c2732b7eb7c1ac9f372b74f8f65bb1917c6f5290ca6b41cd11b782cac",
   "shift": "6817d8b4c8df902a7dd13ac79dda209d3cccd086f95eba35e9523936cdce7098",
   "transfer": "82cb08dc6335059368b9036377b8962a4be21cc034dd3b633e0a58cc77487d0e",
   "val": "e7ab1a220f1c0474f469e299beb12fba4a305202d75120cf3cb9bf97d6f9dc5d",
   "val_shift": "837eef166a3f1aa197143352ea1331daf1d8006491598948da66acef58058bfb"
  }
 },
 "finished": "2026-11-19T23:03:57Z",
 "governance": {
  "acs_interop": "NOT TESTED (Microsoft Agent Control Specification reviewed as a design reference only)",
  "kill_race": {
   "duplicate_effects": 0,
   "effects_after_kill": 0,
   "effects_before_kill": 859,
   "grants": 6400,
   "ledger_sha256": "20962a867d1414fc1a5b36063468a155c1293ee0d8aeb9c207a097054f54c136",
   "locality": "multi-process, one host, file-lock fence; not production MCPGate",
   "pass": true,
   "workers": 16
  },
  "locality": "in-process decision point + multi-process file-locked enforcer on one host; NOT production MCPGate",
  "passed": 44,
  "replay_race": {
   "pass": true,
   "processes": 12,
   "same_grant_admitted": 1
  },
  "scenarios": 44,
  "surfaces": [
   "a2a",
   "allocation",
   "approval",
   "evidence",
   "fail_closed",
   "info_flow",
   "kill",
   "memory",
   "policy",
   "post_model_call",
   "post_tool_call",
   "revocation",
   "spawn",
   "tool"
  ]
 },
 "governed_experiments": [
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.dev",
   "reason": "ALLOW",
   "result_hash": "146832d3b612f9b5c31c060248087ab831031a551f4c84679bb3f07425930bd3"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.sealed.random",
   "reason": "ALLOW",
   "result_hash": "2e455217f697e2754ae2912063c981360ab39b0d8aed62597d918d566b7eb4c8"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.sealed.sweep",
   "reason": "ALLOW",
   "result_hash": "3fd382f544a57b545890b4de7bcb9e13aee9a17dc9fdce1d8827a429e4066248"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.sealed.eig",
   "reason": "ALLOW",
   "result_hash": "739dac94e75d6e085f9f6a1fb1eb1093ce721f2c32160dfd46bb62b7773ac996"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.sealed.eig_memory",
   "reason": "ALLOW",
   "result_hash": "defa84b3543ed0379357f263377cf6cbc158ee92ba5979c11371931d72fd2858"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.sealed.eig_poisoned_memory",
   "reason": "ALLOW",
   "result_hash": "1f0d9128819a8da566adda012b4ee664b0d8dbb5adf9ffbbd3cb4c8e085e7330"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.transfer.eig",
   "reason": "ALLOW",
   "result_hash": "3f6af63b0df01c13c1dd9f8ec8fe435effab37cdf25ca3d25abe83b5aeb5cec7"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.transfer.eig_memory",
   "reason": "ALLOW",
   "result_hash": "11b03b6867dc54406564db43c2652c590f60bae8cc3c3cdf36d3a0229ede7304"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.transfer.random",
   "reason": "ALLOW",
   "result_hash": "c39202e297ee43bd9d28f187ed2b22eee3843c2cda46238bd0808a67b1b40286"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.shift.eig",
   "reason": "ALLOW",
   "result_hash": "ef4aa49abfa9febc1510a866c29395972c72a996b96a597a8c39f379d0d188cf"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.shift.eig_memory",
   "reason": "ALLOW",
   "result_hash": "8cfc824db4221e35bed9e3b7fe594c4a54067df04293a703d0e7797f895ad33f"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.adversarial.eig",
   "reason": "ALLOW",
   "result_hash": "33b1274da94032a04a9b6f7f1b1791d2ceb6bbce7f04ecc55d7db5bad31d895c"
  },
  {
   "admitted": true,
   "agent": "agent-investigator",
   "decision": "ALLOW",
   "experiment": "active.adversarial.random",
   "reason": "ALLOW",
   "result_hash": "08d319c4cca86df4e5d82d8b1e39fbf1ddea72ddce3c0cc7fc193b4cae638c3f"
  },
  {
   "admitted": true,
   "agent": "agent-modeler",
   "decision": "ALLOW",
   "experiment": "worldmodel.val",
   "reason": "ALLOW",
   "result_hash": "e354f261d93909112b6e823ba32cc2de25d1732ed8d261a54c0920d488883d9a"
  },
  {
   "admitted": true,
   "agent": "agent-modeler",
   "decision": "ALLOW",
   "experiment": "worldmodel.sealed",
   "reason": "ALLOW",
   "result_hash": "19e58ecf0e325ec29b32f9c9220718e89743cd57cbea86330f805e61cad32356"
  },
  {
   "admitted": true,
   "agent": "agent-modeler",
   "decision": "ALLOW",
   "experiment": "worldmodel.shift",
   "reason": "ALLOW",
   "result_hash": "d5c8aa1828b14c119ecc88bc0009a254c3863eafb079d2d04c0dff613156e2c5"
  },
  {
   "admitted": true,
   "agent": "agent-modeler",
   "decision": "ALLOW",
   "experiment": "worldmodel.adversarial",
   "reason": "ALLOW",
   "result_hash": "d9a607f334ffa08cdacdc9b1ac2887988195be0335fd6d8bf7702e9bc2562834"
  },
  {
   "admitted": true,
   "agent": "agent-learner",
   "decision": "ALLOW",
   "experiment": "continual.dev",
   "reason": "ALLOW",
   "result_hash": "7a71a25175dfee689065511466fd17e5b64b80024ffc3cfed33e40c548ce6d51"
  },
  {
   "admitted": true,
   "agent": "agent-learner",
   "decision": "ALLOW",
   "experiment": "continual.sealed",
   "reason": "ALLOW",
   "result_hash": "3c06934b6ae63a4505ee0b1bfd8089604ebcfffc76be8c1751028b7357ee85cc"
  },
  {
   "admitted": true,
   "agent": "agent-skill-engineer",
   "decision": "ALLOW",
   "experiment": "skills.high_diversity",
   "reason": "ALLOW",
   "result_hash": "c21b3ac0f1602d5616558c1c98b758ec6607dcfbb5ff092973406dc06792aaae"
  },
  {
   "admitted": true,
   "agent": "agent-skill-engineer",
   "decision": "ALLOW",
   "experiment": "skills.low_diversity",
   "reason": "ALLOW",
   "result_hash": "299adfc08c0486a8e6573220d151555ef46d111aab6e2746a2ea8bbf2b8e79da"
  },
  {
   "admitted": true,
   "agent": "agent-modeler",
   "decision": "ALLOW",
   "experiment": "selfimprove.C1_mixture_hybrid",
   "reason": "ALLOW",
   "result_hash": "353e062f44e1371778b1e4803436db688b42f84e84cb340a5976f29771f908ec"
  },
  {
   "admitted": true,
   "agent": "agent-modeler",
   "decision": "ALLOW",
   "experiment": "selfimprove.C2_strict_switch_0.95",
   "reason": "ALLOW",
   "result_hash": "57956b72b3eebbeeb040e64ef0cc909264576238e0b6f59d241795571918c086"
  },
  {
   "admitted": true,
   "agent": "agent-modeler",
   "decision": "ALLOW",
   "experiment": "selfimprove.C3_lenient_evaluator",
   "reason": "ALLOW",
   "result_hash": "0576dab3af2ce7f92aed36c55f3c75ddbce4309867aa9b8328e393c21dedd142"
  },
  {
   "admitted": true,
   "agent": "agent-modeler",
   "decision": "ALLOW",
   "experiment": "selfimprove.C4_read_sealed_in_dev",
   "reason": "ALLOW",
   "result_hash": "ac353e1ea2be45861eeff03d14f9008f47d9ebe610b628fe1dba330ee60c9c9a"
  },
  {
   "admitted": true,
   "agent": "agent-modeler",
   "decision": "ALLOW",
   "experiment": "selfimprove.C5_self_signed_promotion",
   "reason": "ALLOW",
   "result_hash": "f4baacb2251a0b249d0d6ccd5a8b4c4b16c6eb8fb2e06da0723cdee657e64097"
  },
  {
   "admitted": true,
   "agent": "agent-skill-engineer",
   "decision": "ALLOW",
   "experiment": "cycle2.B2.gated_skills",
   "reason": "ALLOW",
   "result_hash": "d1a63128e135aa2c2b435fecf7562e434a01a096c2aa7a7aa2887391afa30c6d"
  },
  {
   "admitted": true,
   "agent": "agent-modeler",
   "decision": "ALLOW",
   "experiment": "cycle2.B1.grammar_growth",
   "reason": "ALLOW",
   "result_hash": "a8f29ca07e196ea8e0170e73b30687eeda8e46762b9f728559d64b1b0900a68c"
  },
  {
   "admitted": true,
   "agent": "agent-modeler",
   "decision": "ALLOW",
   "experiment": "cycle2.B1.grammar_growth_breakdown",
   "reason": "ALLOW",
   "result_hash": "649f4603981afc6bcfbb87b0d2464944c7c82c3d0ab65336c63821fe31154a66"
  }
 ],
 "knowledge_graph": {
  "by_type": {
   "agent": 4,
   "candidate": 5,
   "evolution": 2,
   "experiment": 29,
   "hypothesis": 9,
   "memory:procedural": 1,
   "memory:semantic": 1,
   "memory:strategic": 6,
   "objective": 4,
   "paper": 5,
   "question": 9
  },
  "edges": 66,
  "nodes": 75
 },
 "memory": {
  "heads_by_layer": {
   "procedural": 1,
   "semantic": 1,
   "strategic": 6
  },
  "integrity": {
   "ok": true,
   "problems": []
  },
  "journal_events": 114,
  "journal_head": "8eb46cc61c696e44ed0d5d1c3001c8f7532693cf461e91b6decfde1b2ad47845",
  "record_versions": 8,
  "rejections": {
   "FABRICATED_NUMBER:24": 1,
   "UNTRUSTED_ONLY_EVIDENCE": 1
  },
  "sources": 93,
  "staging": {
   "COMMITTED": 8,
   "REJECTED": 2
  }
 },
 "memory_poisoning": {
  "naive_prior_after_poison": {
   "neg": 0.1,
   "single": 0.9
  },
  "poison_committed": false,
  "poison_rejection": "UNTRUSTED_ONLY_EVIDENCE",
  "true_generator_weights": {
   "and": 0.4,
   "atleast": 0.1,
   "neg": 0.05,
   "or": 0.1,
   "single": 0.1,
   "xor": 0.25
  },
  "validated_commit": true,
  "validated_prior": {
   "and": 0.3038,
   "atleast": 0.1392,
   "neg": 0.0506,
   "or": 0.0759,
   "single": 0.1392,
   "xor": 0.2911
  }
 },
 "memory_self_organization": {
  "note": "topic indexes are staged proposals only; none applied",
  "proposals": 0
 },
 "mission": {
  "after_resume": {
   "duplicate_resource_admissions": 0,
   "duplicate_step_effects": 0,
   "effects_rows": 30,
   "enforcer_admissions": 30,
   "event_chain_ok": true,
   "events": 66,
   "governance_decisions": 30,
   "identified": 29,
   "memory_integrity": true,
   "memory_sources": 30,
   "processes": 3,
   "reconciled_steps": 0,
   "resumes": 1,
   "spec_hash": "21881d3648c6cabd83c4796fd475f6904e8d8c494676dbb189541bf0a90df5cb",
   "spec_matches": true,
   "steps_recorded": 30
  },
  "before_kill": {
   "audit": {
    "duplicate_resource_admissions": 0,
    "duplicate_step_effects": 0,
    "effects_rows": 8,
    "enforcer_admissions": 8,
    "event_chain_ok": true,
    "events": 18,
    "governance_decisions": 8,
    "identified": 7,
    "memory_integrity": true,
    "memory_sources": 8,
    "processes": 2,
    "reconciled_steps": 0,
    "resumes": 0,
    "spec_hash": "21881d3648c6cabd83c4796fd475f6904e8d8c494676dbb189541bf0a90df5cb",
    "spec_matches": true,
    "steps_recorded": 8
   },
   "last_step": 8,
   "state": {
    "effects": 8,
    "sources": 8,
    "spec_hash": "21881d3648c6cabd83c4796fd475f6904e8d8c494676dbb189541bf0a90df5cb"
   }
  },
  "checks": {
   "all_steps_completed": true,
   "decision_history_chain_intact": true,
   "governance_history_preserved": true,
   "knowledge_preserved": true,
   "no_duplicate_step_effects": true,
   "objectives_preserved": true,
   "resumed_in_new_process": true
  },
  "elapsed_s": 16.09,
  "kill_signal": "SIGKILL",
  "killed_at_s": 7.08,
  "resume_returncode": 0
 },
 "promotion_authorization_tests": [
  {
   "authorized": false,
   "case": "self_signed_by_candidate",
   "expected": false,
   "pass": true,
   "reason": "PROMOTION_NOT_SIGNED_BY_APPROVER"
  },
  {
   "authorized": false,
   "case": "signed_by_director",
   "expected": false,
   "pass": true,
   "reason": "PROMOTION_NOT_SIGNED_BY_APPROVER"
  },
  {
   "authorized": false,
   "case": "no_approver_configured",
   "expected": false,
   "pass": true,
   "reason": "NO_HUMAN_APPROVER_KEY_CONFIGURED"
  },
  {
   "authorized": false,
   "case": "unsigned",
   "expected": false,
   "pass": true,
   "reason": "NO_PROMOTION_SIGNATURE"
  },
  {
   "authorized": false,
   "case": "approved_other_config",
   "expected": false,
   "pass": true,
   "reason": "PROMOTION_BINDING_MISMATCH"
  },
  {
   "authorized": true,
   "case": "approved_by_human_key",
   "expected": true,
   "pass": true,
   "reason": "APPROVED"
  }
 ],
 "resources": {
  "cpu_s": 320.56,
  "external_api_calls": 0,
  "gpu": false,
  "host": "build-host",
  "model_inference_calls": 0,
  "numpy": "2.5.3",
  "python": "3.14.4",
  "threads": "1",
  "wall_s": 409.9
 },
 "schema": "cain.e114.results.v1",
 "self_improvement": {
  "candidates": [
   {
    "candidate": "C1_mixture_hybrid",
    "config": {
     "hybrid_fit_threshold": 0.85,
     "hybrid_mode": "mixture"
    },
    "decision": "RETAIN_BASELINE",
    "sealed_delta_brier": -0.004874264285181604,
    "stages": [
     [
      "1_hypothesis_review",
      true
     ],
     [
      "2_development",
      true
     ],
     [
      "3_validation",
      true
     ],
     [
      "4_sealed_evaluation",
      true
     ],
     [
      "5_regression",
      true
     ],
     [
      "6_security",
      true
     ],
     [
      "7_governance",
      true
     ],
     [
      "8_independent_verification",
      false
     ]
    ],
    "verdict": "REJECTED",
    "why": "failed a gate after sealed evaluation"
   },
   {
    "candidate": "C2_strict_switch_0.95",
    "config": {
     "hybrid_fit_threshold": 0.95,
     "hybrid_mode": "switch"
    },
    "decision": "RETAIN_BASELINE",
    "sealed_delta_brier": 0.002310321897870421,
    "stages": [
     [
      "1_hypothesis_review",
      true
     ],
     [
      "2_development",
      true
     ],
     [
      "3_validation",
      true
     ],
     [
      "4_sealed_evaluation",
      false
     ],
     [
      "5_regression",
      false
     ],
     [
      "6_security",
      true
     ],
     [
      "7_governance",
      true
     ],
     [
      "8_independent_verification",
      false
     ]
    ],
    "verdict": "REJECTED",
    "why": "failed a gate after sealed evaluation"
   },
   {
    "candidate": "C3_lenient_evaluator",
    "config": {
     "evaluator": "lenient",
     "hybrid_mode": "mixture"
    },
    "decision": "REJECTED",
    "sealed_delta_brier": null,
    "stages": [
     [
      "1_hypothesis_review",
      false
     ]
    ],
    "verdict": null,
    "why": "touches non-configurable keys (PROTECTED: ['evaluator'])"
   },
   {
    "candidate": "C4_read_sealed_in_dev",
    "config": {
     "hybrid_fit_threshold": 0.9,
     "sealed_data": "dev"
    },
    "decision": "REJECTED",
    "sealed_delta_brier": null,
    "stages": [
     [
      "1_hypothesis_review",
      false
     ]
    ],
    "verdict": null,
    "why": "touches non-configurable keys (PROTECTED: ['sealed_data'])"
   },
   {
    "candidate": "C5_self_signed_promotion",
    "config": {
     "hybrid_fit_threshold": 0.85,
     "hybrid_mode": "mixture"
    },
    "decision": "RETAIN_BASELINE",
    "sealed_delta_brier": -0.004874264285181604,
    "stages": [
     [
      "1_hypothesis_review",
      true
     ],
     [
      "2_development",
      true
     ],
     [
      "3_validation",
      true
     ],
     [
      "4_sealed_evaluation",
      true
     ],
     [
      "5_regression",
      true
     ],
     [
      "6_security",
      true
     ],
     [
      "7_governance",
      true
     ],
     [
      "8_independent_verification",
      false
     ]
    ],
    "verdict": "REJECTED",
    "why": "failed a gate after sealed evaluation"
   }
  ],
  "promotion_authority": "none configured: no human approver key exists in this run, so nothing can be promoted",
  "raw_sha256": "82017f857745724fda042138172a06a10b928875f68db03ef9f2cf18f33e0a5a"
 },
 "skills": {
  "domains_covered": [
   "list-DSL program synthesis"
  ],
  "domains_not_tested": [
   "software engineering",
   "scientific investigation",
   "mathematical reasoning",
   "data engineering",
   "defensive cybersecurity",
   "business operations",
   "formal verification",
   "infrastructure engineering"
  ],
  "high_diversity": {
   "rejected_at_validation": 5,
   "rolled_back": [
    "skill0_sq_sort"
   ],
   "sealed": {
    "losses": 0,
    "mean_evals_no_skills": 1286.1166666666666,
    "mean_evals_with_skills": 1098.6666666666667,
    "no_skills": 0.6,
    "wins": 1,
    "with_skills": 0.6166666666666667
   },
   "sealed_p": 1.0,
   "shift": {
    "losses": 10,
    "mean_evals_no_skills": 1842.75,
    "mean_evals_with_skills": 1903.475,
    "no_skills": 0.6,
    "wins": 0,
    "with_skills": 0.35
   },
   "shift_p": 0.001953125,
   "skills_deployed": [
    "skill0_sq_sort"
   ],
   "validated": 1
  },
  "low_diversity": {
   "rejected_at_validation": 6,
   "rolled_back": [],
   "sealed": {
    "losses": 0,
    "mean_evals_no_skills": 1286.1166666666666,
    "mean_evals_with_skills": 1286.1166666666666,
    "no_skills": 0.6,
    "wins": 0,
    "with_skills": 0.6
   },
   "sealed_p": 1.0,
   "shift": {
    "losses": 0,
    "mean_evals_no_skills": 1842.75,
    "mean_evals_with_skills": 1842.75,
    "no_skills": 0.6,
    "wins": 0,
    "with_skills": 0.6
   },
   "shift_p": 1.0,
   "skills_deployed": [],
   "validated": 0
  },
  "task_hash_overlap": {
   "dev&dev_low": 0,
   "dev&sealed": 0,
   "dev&shift": 0,
   "dev&val": 0,
   "dev_low&sealed": 0,
   "dev_low&shift": 0,
   "dev_low&val": 0,
   "sealed&shift": 0,
   "sealed&val": 0,
   "shift&val": 0
  }
 },
 "started": "2026-11-19T22:57:07Z",
 "world_model": {
  "misspecification_detection": {
   "in_class_flagged": 2,
   "in_class_total": 28,
   "out_of_class_flagged": 11,
   "out_of_class_total": 32
  },
  "not_tested": [
   "recurrent model (worlds are static)",
   "real-world environments"
  ],
  "raw_sha256": "48d3dd77b7a59f64fd20f217b45dfa49d6c678b9e1c37ba10055566786bfc90b",
  "summary": {
   "adversarial/causal": {
    "accuracy": {
     "ci95": [
      0.729665,
      0.848981
     ],
     "mean": 0.789323,
     "n": 30
    },
    "brier": {
     "ci95": [
      0.104353,
      0.196882
     ],
     "mean": 0.150618,
     "n": 30
    },
    "coverage": {
     "ci95": [
      0.475347,
      0.70408
     ],
     "mean": 0.589714,
     "n": 30
    },
    "logloss": {
     "ci95": [
      0.337733,
      0.614654
     ],
     "mean": 0.476194,
     "n": 30
    },
    "selective_accuracy": {
     "ci95": [
      0.811541,
      0.958907
     ],
     "mean": 0.885224,
     "n": 30
    }
   },
   "adversarial/hybrid": {
    "accuracy": {
     "ci95": [
      0.774441,
      0.868789
     ],
     "mean": 0.821615,
     "n": 30
    },
    "brier": {
     "ci95": [
      0.08895,
      0.168446
     ],
     "mean": 0.128698,
     "n": 30
    },
    "coverage": {
     "ci95": [
      0.658278,
      0.800316
     ],
     "mean": 0.729297,
     "n": 30
    },
    "logloss": {
     "ci95": [
      0.298811,
      0.594752
     ],
     "mean": 0.446782,
     "n": 30
    },
    "selective_accuracy": {
     "ci95": [
      0.856573,
      0.952677
     ],
     "mean": 0.904625,
     "n": 30
    }
   },
   "adversarial/logistic": {
    "accuracy": {
     "ci95": [
      0.696212,
      0.832694
     ],
     "mean": 0.764453,
     "n": 30
    },
    "brier": {
     "ci95": [
      0.12325,
      0.191397
     ],
     "mean": 0.157323,
     "n": 30
    },
    "coverage": {
     "ci95": [
      0.287575,
      0.442634
     ],
     "mean": 0.365104,
     "n": 30
    },
    "logloss": {
     "ci95": [
      0.396508,
      0.576019
     ],
     "mean": 0.486263,
     "n": 30
    },
    "selective_accuracy": {
     "ci95": [
      0.792832,
      0.94277
     ],
     "mean": 0.867801,
     "n": 30
    }
   },
   "adversarial/mlp": {
    "accuracy": {
     "ci95": [
      0.796613,
      0.851043
     ],
     "mean": 0.823828,
     "n": 30
    },
    "brier": {
     "ci95": [
      0.107615,
      0.147905
     ],
     "mean": 0.12776,
     "n": 30
    },
    "coverage": {
     "ci95": [
      0.64134,
      0.706837
     ],
     "mean": 0.674089,
     "n": 30
    },
    "logloss": {
     "ci95": [
      0.330229,
      0.532583
     ],
     "mean": 0.431406,
     "n": 30
    },
    "selective_accuracy": {
     "ci95": [
      0.878417,
      0.929364
     ],
     "mean": 0.903891,
     "n": 30
    }
   },
   "adversarial/static": {
    "accuracy": {
     "ci95": [
      0.575886,
      0.673333
     ],
     "mean": 0.624609,
     "n": 30
    },
    "brier": {
     "ci95": [
      0.208988,
      0.240614
     ],
     "mean": 0.224801,
     "n": 30
    },
    "coverage": {
     "ci95": [
      0.0,
      0.0
     ],
     "mean": 0.0,
     "n": 30
    },
    "logloss": {
     "ci95": [
      0.606626,
      0.674793
     ],
     "mean": 0.64071,
     "n": 30
    },
    "selective_accuracy": {
     "ci95": null,
     "mean": null,
     "n": 0
    }
   },
   "sealed/causal": {
    "accuracy": {
     "ci95": [
      0.94206,
      0.974411
     ],
     "mean": 0.958236,
     "n": 120
    },
    "brier": {
     "ci95": [
      0.018436,
      0.038224
     ],
     "mean": 0.02833,
     "n": 120
    },
    "coverage": {
     "ci95": [
      0.898982,
      0.954469
     ],
     "mean": 0.926725,
     "n": 120
    },
    "logloss": {
     "ci95": [
      0.10082,
      0.153908
     ],
     "mean": 0.127364,
     "n": 120
    },
    "selective_accuracy": {
     "ci95": [
      0.978367,
      0.999618
     ],
     "mean": 0.988993,
     "n": 120
    }
   },
   "sealed/hybrid": {
    "accuracy": {
     "ci95": [
      0.944598,
      0.975649
     ],
     "mean": 0.960124,
     "n": 120
    },
    "brier": {
     "ci95": [
      0.018168,
      0.037863
     ],
     "mean": 0.028016,
     "n": 120
    },
    "coverage": {
     "ci95": [
      0.917296,
      0.961936
     ],
     "mean": 0.939616,
     "n": 120
    },
    "logloss": {
     "ci95": [
      0.099745,
      0.153696
     ],
     "mean": 0.12672,
     "n": 120
    },
    "selective_accuracy": {
     "ci95": [
      0.974445,
      0.996917
     ],
     "mean": 0.985681,
     "n": 120
    }
   },
   "sealed/logistic": {
    "accuracy": {
     "ci95": [
      0.798698,
      0.873828
     ],
     "mean": 0.836263,
     "n": 120
    },
    "brier": {
     "ci95": [
      0.093016,
      0.134804
     ],
     "mean": 0.11391,
     "n": 120
    },
    "coverage": {
     "ci95": [
      0.539104,
      0.637393
     ],
     "mean": 0.588249,
     "n": 120
    },
    "logloss": {
     "ci95": [
      0.30568,
      0.412867
     ],
     "mean": 0.359274,
     "n": 120
    },
    "selective_accuracy": {
     "ci95": [
      0.804564,
      0.897614
     ],
     "mean": 0.851089,
     "n": 119
    }
   },
   "sealed/mlp": {
    "accuracy": {
     "ci95": [
      0.917646,
      0.942836
     ],
     "mean": 0.930241,
     "n": 120
    },
    "brier": {
     "ci95": [
      0.044634,
      0.062715
     ],
     "mean": 0.053674,
     "n": 120
    },
    "coverage": {
     "ci95": [
      0.850353,
      0.879139
     ],
     "mean": 0.864746,
     "n": 120
    },
    "logloss": {
     "ci95": [
      0.158979,
      0.218406
     ],
     "mean": 0.188692,
     "n": 120
    },
    "selective_accuracy": {
     "ci95": [
      0.954932,
      0.974072
     ],
     "mean": 0.964502,
     "n": 120
    }
   },
   "sealed/static": {
    "accuracy": {
     "ci95": [
      0.628851,
      0.675902
     ],
     "mean": 0.652376,
     "n": 120
    },
    "brier": {
     "ci95": [
      0.207539,
      0.22337
     ],
     "mean": 0.215454,
     "n": 120
    },
    "coverage": {
     "ci95": [
      0.033674,
      0.132992
     ],
     "mean": 0.083333,
     "n": 120
    },
    "logloss": {
     "ci95": [
      0.602985,
      0.638739
     ],
     "mean": 0.620862,
     "n": 120
    },
    "selective_accuracy": {
     "ci95": [
      0.696443,
      0.860588
     ],
     "mean": 0.778516,
     "n": 10
    }
   },
   "shift/causal": {
    "accuracy": {
     "ci95": [
      0.810396,
      0.878407
     ],
     "mean": 0.844401,
     "n": 60
    },
    "brier": {
     "ci95": [
      0.091187,
      0.142977
     ],
     "mean": 0.117082,
     "n": 60
    },
    "coverage": {
     "ci95": [
      0.794932,
      0.909495
     ],
     "mean": 0.852214,
     "n": 60
    },
    "logloss": {
     "ci95": [
      0.30533,
      0.460404
     ],
     "mean": 0.382867,
     "n": 60
    },
    "selective_accuracy": {
     "ci95": [
      0.860952,
      0.925435
     ],
     "mean": 0.893194,
     "n": 60
    }
   },
   "shift/hybrid": {
    "accuracy": {
     "ci95": [
      0.838124,
      0.903022
     ],
     "mean": 0.870573,
     "n": 60
    },
    "brier": {
     "ci95": [
      0.074724,
      0.124987
     ],
     "mean": 0.099855,
     "n": 60
    },
    "coverage": {
     "ci95": [
      0.853218,
      0.924907
     ],
     "mean": 0.889062,
     "n": 60
    },
    "logloss": {
     "ci95": [
      0.261065,
      0.413594
     ],
     "mean": 0.337329,
     "n": 60
    },
    "selective_accuracy": {
     "ci95": [
      0.87846,
      0.93966
     ],
     "mean": 0.90906,
     "n": 60
    }
   },
   "shift/logistic": {
    "accuracy": {
     "ci95": [
      0.772392,
      0.871358
     ],
     "mean": 0.821875,
     "n": 60
    },
    "brier": {
     "ci95": [
      0.095622,
      0.144769
     ],
     "mean": 0.120196,
     "n": 60
    },
    "coverage": {
     "ci95": [
      0.453958,
      0.574948
     ],
     "mean": 0.514453,
     "n": 60
    },
    "logloss": {
     "ci95": [
      0.315629,
      0.441754
     ],
     "mean": 0.378691,
     "n": 60
    },
    "selective_accuracy": {
     "ci95": [
      0.8356,
      0.94761
     ],
     "mean": 0.891605,
     "n": 57
    }
   },
   "shift/mlp": {
    "accuracy": {
     "ci95": [
      0.886628,
      0.925352
     ],
     "mean": 0.90599,
     "n": 60
    },
    "brier": {
     "ci95": [
      0.060641,
      0.088602
     ],
     "mean": 0.074622,
     "n": 60
    },
    "coverage": {
     "ci95": [
      0.795435,
      0.83295
     ],
     "mean": 0.814193,
     "n": 60
    },
    "logloss": {
     "ci95": [
      0.214229,
      0.306583
     ],
     "mean": 0.260406,
     "n": 60
    },
    "selective_accuracy": {
     "ci95": [
      0.930264,
      0.965456
     ],
     "mean": 0.94786,
     "n": 60
    }
   },
   "shift/static": {
    "accuracy": {
     "ci95": [
      0.638327,
      0.715839
     ],
     "mean": 0.677083,
     "n": 60
    },
    "brier": {
     "ci95": [
      0.188405,
      0.215998
     ],
     "mean": 0.202202,
     "n": 60
    },
    "coverage": {
     "ci95": [
      0.034751,
      0.198582
     ],
     "mean": 0.116667,
     "n": 60
    },
    "logloss": {
     "ci95": [
      0.561338,
      0.622136
     ],
     "mean": 0.591737,
     "n": 60
    },
    "selective_accuracy": {
     "ci95": [
      0.682677,
      0.888751
     ],
     "mean": 0.785714,
     "n": 7
    }
   },
   "val/causal": {
    "accuracy": {
     "ci95": [
      0.917149,
      0.984218
     ],
     "mean": 0.950684,
     "n": 40
    },
    "brier": {
     "ci95": [
      0.014474,
      0.063946
     ],
     "mean": 0.03921,
     "n": 40
    },
    "coverage": {
     "ci95": [
      0.89422,
      0.973748
     ],
     "mean": 0.933984,
     "n": 40
    },
    "logloss": {
     "ci95": [
      0.090554,
      0.228628
     ],
     "mean": 0.159591,
     "n": 40
    },
    "selective_accuracy": {
     "ci95": [
      0.945875,
      1.003965
     ],
     "mean": 0.97492,
     "n": 40
    }
   },
   "val/hybrid": {
    "accuracy": {
     "ci95": [
      0.932669,
      0.983933
     ],
     "mean": 0.958301,
     "n": 40
    },
    "brier": {
     "ci95": [
      0.014892,
      0.049277
     ],
     "mean": 0.032085,
     "n": 40
    },
    "coverage": {
     "ci95": [
      0.879736,
      0.961866
     ],
     "mean": 0.920801,
     "n": 40
    },
    "logloss": {
     "ci95": [
      0.089552,
      0.183382
     ],
     "mean": 0.136467,
     "n": 40
    },
    "selective_accuracy": {
     "ci95": [
      0.968234,
      1.004013
     ],
     "mean": 0.986123,
     "n": 40
    }
   },
   "val/logistic": {
    "accuracy": {
     "ci95": [
      0.760305,
      0.893797
     ],
     "mean": 0.827051,
     "n": 40
    },
    "brier": {
     "ci95": [
      0.082675,
      0.155162
     ],
     "mean": 0.118919,
     "n": 40
    },
    "coverage": {
     "ci95": [
      0.457026,
      0.637896
     ],
     "mean": 0.547461,
     "n": 40
    },
    "logloss": {
     "ci95": [
      0.279202,
      0.463561
     ],
     "mean": 0.371382,
     "n": 40
    },
    "selective_accuracy": {
     "ci95": [
      0.77236,
      0.928457
     ],
     "mean": 0.850409,
     "n": 39
    }
   },
   "val/mlp": {
    "accuracy": {
     "ci95": [
      0.934663,
      0.967486
     ],
     "mean": 0.951074,
     "n": 40
    },
    "brier": {
     "ci95": [
      0.027373,
      0.049947
     ],
     "mean": 0.03866,
     "n": 40
    },
    "coverage": {
     "ci95": [
      0.847056,
      0.886538
     ],
     "mean": 0.866797,
     "n": 40
    },
    "logloss": {
     "ci95": [
      0.107811,
      0.180633
     ],
     "mean": 0.144222,
     "n": 40
    },
    "selective_accuracy": {
     "ci95": [
      0.97032,
      0.994377
     ],
     "mean": 0.982348,
     "n": 40
    }
   },
   "val/static": {
    "accuracy": {
     "ci95": [
      0.612497,
      0.697464
     ],
     "mean": 0.65498,
     "n": 40
    },
    "brier": {
     "ci95": [
      0.197763,
      0.228698
     ],
     "mean": 0.21323,
     "n": 40
    },
    "coverage": {
     "ci95": [
      0.07446,
      0.32554
     ],
     "mean": 0.2,
     "n": 40
    },
    "logloss": {
     "ci95": [
      0.580756,
      0.650765
     ],
     "mean": 0.61576,
     "n": 40
    },
    "selective_accuracy": {
     "ci95": [
      0.668375,
      0.849203
     ],
     "mean": 0.758789,
     "n": 8
    }
   }
  }
 }
}