{
  "schema": "gobstopper-public-retention-ablation-v1",
  "study_date": "2026-09-19",
  "completed_at": "2026-09-19T17:34:07.686013+00:00",
  "measurement": "Paired deterministic scored-policy replay on 73 high-context archived Codex roots and 41 below-trigger controls. Projected current-context tokens and literal-probe survival only.",
  "provenance": {
    "started_at": "2026-09-19T17:31:48.425818+00:00",
    "baseline_binary_sha256": "8cc622469d6904666d5e0135a717ccfadd6cf0f2f79543349661303040a71c1c",
    "candidate_binary_sha256": "0d924d0b8322bd58c8ab4cee7318c41177c809433c61d12aa0ddb7cd26f5b88a",
    "selected_samples": 114,
    "local_model_calls": 0,
    "remote_model_calls": 0,
    "candidate_base_commit": "ded1d0c9e645ea6644e326e632949225765d39da",
    "candidate_source_state": "Local build of the reviewed keep-score-cutoff patch on the stated main commit; a source-header comment was changed after the measured build. The candidate is distinct from the baseline installed when this experiment ran."
  },
  "summary": {
    "total_evaluations": 570,
    "failed_evaluations": 0,
    "disabled_equivalence": true,
    "wall_seconds": 139.259
  },
  "aggregates": [
    {
      "cohort": "high_context",
      "variant": "baseline_disabled",
      "selected": 73,
      "successful": 73,
      "failures": 0,
      "plans": 73,
      "no_plan": 0,
      "median_reduction_percent_including_noops": 36.436228842966436,
      "sum_context_before": 10797307,
      "sum_projected_reclaimed": 4012086,
      "literal_samples": 73,
      "literal_total": 4667,
      "literal_retained": 3596,
      "literal_unchanged_source_derived_samples": 0,
      "new_errors": 0,
      "new_warnings": 0,
      "disabled_equivalence_failures": 0,
      "median_wall_ms": 196.397,
      "paired_retention_improved_samples": 0,
      "paired_retention_decreased_samples": 0,
      "paired_retention_equal_samples": 73,
      "paired_reduction_increased_samples": 0
    },
    {
      "cohort": "high_context",
      "variant": "candidate_disabled",
      "selected": 73,
      "successful": 73,
      "failures": 0,
      "plans": 73,
      "no_plan": 0,
      "median_reduction_percent_including_noops": 36.436228842966436,
      "sum_context_before": 10797307,
      "sum_projected_reclaimed": 4012086,
      "literal_samples": 73,
      "literal_total": 4667,
      "literal_retained": 3596,
      "literal_unchanged_source_derived_samples": 0,
      "new_errors": 0,
      "new_warnings": 0,
      "disabled_equivalence_failures": 0,
      "median_wall_ms": 192.59,
      "paired_retention_improved_samples": 0,
      "paired_retention_decreased_samples": 0,
      "paired_retention_equal_samples": 73,
      "paired_reduction_increased_samples": 0
    },
    {
      "cohort": "high_context",
      "variant": "cutoff_0.35",
      "selected": 73,
      "successful": 73,
      "failures": 0,
      "plans": 71,
      "no_plan": 2,
      "median_reduction_percent_including_noops": 34.96093374439965,
      "sum_context_before": 10797307,
      "sum_projected_reclaimed": 3615788,
      "literal_samples": 73,
      "literal_total": 4667,
      "literal_retained": 3770,
      "literal_unchanged_source_derived_samples": 2,
      "new_errors": 0,
      "new_warnings": 0,
      "disabled_equivalence_failures": 0,
      "median_wall_ms": 201.54,
      "paired_retention_improved_samples": 40,
      "paired_retention_decreased_samples": 0,
      "paired_retention_equal_samples": 33,
      "paired_reduction_increased_samples": 0
    },
    {
      "cohort": "high_context",
      "variant": "cutoff_0.5",
      "selected": 73,
      "successful": 73,
      "failures": 0,
      "plans": 73,
      "no_plan": 0,
      "median_reduction_percent_including_noops": 36.35049846093552,
      "sum_context_before": 10797307,
      "sum_projected_reclaimed": 3973490,
      "literal_samples": 73,
      "literal_total": 4667,
      "literal_retained": 3614,
      "literal_unchanged_source_derived_samples": 0,
      "new_errors": 0,
      "new_warnings": 0,
      "disabled_equivalence_failures": 0,
      "median_wall_ms": 205.619,
      "paired_retention_improved_samples": 11,
      "paired_retention_decreased_samples": 0,
      "paired_retention_equal_samples": 62,
      "paired_reduction_increased_samples": 2
    },
    {
      "cohort": "high_context",
      "variant": "cutoff_0.65",
      "selected": 73,
      "successful": 73,
      "failures": 0,
      "plans": 73,
      "no_plan": 0,
      "median_reduction_percent_including_noops": 36.436228842966436,
      "sum_context_before": 10797307,
      "sum_projected_reclaimed": 4008273,
      "literal_samples": 73,
      "literal_total": 4667,
      "literal_retained": 3598,
      "literal_unchanged_source_derived_samples": 0,
      "new_errors": 0,
      "new_warnings": 0,
      "disabled_equivalence_failures": 0,
      "median_wall_ms": 199.742,
      "paired_retention_improved_samples": 1,
      "paired_retention_decreased_samples": 0,
      "paired_retention_equal_samples": 72,
      "paired_reduction_increased_samples": 1
    },
    {
      "cohort": "below_trigger_control",
      "variant": "baseline_disabled",
      "selected": 41,
      "successful": 41,
      "failures": 0,
      "plans": 0,
      "no_plan": 41,
      "median_reduction_percent_including_noops": 0.0,
      "sum_context_before": 2031198,
      "sum_projected_reclaimed": 0,
      "literal_samples": 0,
      "literal_total": 0,
      "literal_retained": 0,
      "literal_unchanged_source_derived_samples": 0,
      "new_errors": 0,
      "new_warnings": 0,
      "disabled_equivalence_failures": 0,
      "median_wall_ms": 42.739,
      "paired_retention_improved_samples": 0,
      "paired_retention_decreased_samples": 0,
      "paired_retention_equal_samples": 0,
      "paired_reduction_increased_samples": 0
    },
    {
      "cohort": "below_trigger_control",
      "variant": "candidate_disabled",
      "selected": 41,
      "successful": 41,
      "failures": 0,
      "plans": 0,
      "no_plan": 41,
      "median_reduction_percent_including_noops": 0.0,
      "sum_context_before": 2031198,
      "sum_projected_reclaimed": 0,
      "literal_samples": 0,
      "literal_total": 0,
      "literal_retained": 0,
      "literal_unchanged_source_derived_samples": 0,
      "new_errors": 0,
      "new_warnings": 0,
      "disabled_equivalence_failures": 0,
      "median_wall_ms": 46.961,
      "paired_retention_improved_samples": 0,
      "paired_retention_decreased_samples": 0,
      "paired_retention_equal_samples": 0,
      "paired_reduction_increased_samples": 0
    },
    {
      "cohort": "below_trigger_control",
      "variant": "cutoff_0.35",
      "selected": 41,
      "successful": 41,
      "failures": 0,
      "plans": 0,
      "no_plan": 41,
      "median_reduction_percent_including_noops": 0.0,
      "sum_context_before": 2031198,
      "sum_projected_reclaimed": 0,
      "literal_samples": 0,
      "literal_total": 0,
      "literal_retained": 0,
      "literal_unchanged_source_derived_samples": 0,
      "new_errors": 0,
      "new_warnings": 0,
      "disabled_equivalence_failures": 0,
      "median_wall_ms": 32.68,
      "paired_retention_improved_samples": 0,
      "paired_retention_decreased_samples": 0,
      "paired_retention_equal_samples": 0,
      "paired_reduction_increased_samples": 0
    },
    {
      "cohort": "below_trigger_control",
      "variant": "cutoff_0.5",
      "selected": 41,
      "successful": 41,
      "failures": 0,
      "plans": 0,
      "no_plan": 41,
      "median_reduction_percent_including_noops": 0.0,
      "sum_context_before": 2031198,
      "sum_projected_reclaimed": 0,
      "literal_samples": 0,
      "literal_total": 0,
      "literal_retained": 0,
      "literal_unchanged_source_derived_samples": 0,
      "new_errors": 0,
      "new_warnings": 0,
      "disabled_equivalence_failures": 0,
      "median_wall_ms": 47.956,
      "paired_retention_improved_samples": 0,
      "paired_retention_decreased_samples": 0,
      "paired_retention_equal_samples": 0,
      "paired_reduction_increased_samples": 0
    },
    {
      "cohort": "below_trigger_control",
      "variant": "cutoff_0.65",
      "selected": 41,
      "successful": 41,
      "failures": 0,
      "plans": 0,
      "no_plan": 41,
      "median_reduction_percent_including_noops": 0.0,
      "sum_context_before": 2031198,
      "sum_projected_reclaimed": 0,
      "literal_samples": 0,
      "literal_total": 0,
      "literal_retained": 0,
      "literal_unchanged_source_derived_samples": 0,
      "new_errors": 0,
      "new_warnings": 0,
      "disabled_equivalence_failures": 0,
      "median_wall_ms": 45.857,
      "paired_retention_improved_samples": 0,
      "paired_retention_decreased_samples": 0,
      "paired_retention_equal_samples": 0,
      "paired_reduction_increased_samples": 0
    }
  ],
  "limitations": [
    "The same 114 archived root sessions from one person and one Mac were replayed; this is not a new independent sample.",
    "This development comparison reuses an already studied corpus. The original baseline was already known when the three cutoffs were registered before this run; this is not held-out or external validation.",
    "All three cutoffs were registered before outcomes, but none is established as a tuned optimum on held-out tasks.",
    "Literal matching is not semantic importance, correct continuation, task success or automatic recovery.",
    "No provider continuation, billing, local model or remote model was evaluated; this is not a fast-jev-compaction comparison.",
    "No-plan cases remain in reduction denominators as zero. For two high-context no-plan cases at cutoff 0.35, retention is derived from leaving the source unchanged, not measured on a rewrite.",
    "The baseline and candidate use the corrected 64-probe cap. Probe totals differ from the original 729-session study; do not compare their retention ratios as a policy effect.",
    "Structural severity/code counts did not increase; that does not establish safe provider resume.",
    "Timing is descriptive under shared host load, not a latency benchmark.",
    "Private transcripts, session identifiers, paths, per-session hashes, corpus manifests and per-session results are excluded; the exact private corpus cannot be independently replayed from this download."
  ]
}