{
  "schema_version": "1.1",
  "id": "A-001",
  "date": "2026-10-23",
  "status": "unreviewed_editorial_assumption_scenario_NOT_statistically_calibrated",
  "display_purpose": "Sensitivity analysis only; not a locked or official Cortist forecast probability",
  "probability_percent": 86,
  "scenario_sensitivity_range_percent": [
    64,
    96
  ],
  "range_interpretation": "Judgmental / illustrative parameter variation; NOT a statistical confidence or credible interval",
  "model_type": "conditional assumption decomposition",
  "model_expression": "P(capability by 2028) x P(eligible independent observation | capability)",
  "assumptions": [
    {
      "label": "80%-reliable 8h capability by 2028",
      "p": 0.95,
      "basis": "Judgmental assumption informed by METR 2026 estimate of about 3.1h at 80% reliability and historical time-horizon growth; extrapolation may break."
    },
    {
      "label": "Comparable qualifying public independent evaluation, conditional",
      "p": 0.9,
      "basis": "Judgmental assumption. METR's previous public dashboard was no longer actively updated by September 2026."
    }
  ],
  "sensitivity": [
    {
      "case": "conservative",
      "factors": [
        0.8,
        0.8
      ],
      "probability": 64
    },
    {
      "case": "working",
      "factors": [
        0.95,
        0.9
      ],
      "probability": 85.5
    },
    {
      "case": "optimistic",
      "factors": [
        0.99,
        0.97
      ],
      "probability": 96.03
    }
  ],
  "limitations": [
    "The 80% task-success threshold in the question is not the same as 85% chance of the milestone being met by 2028.",
    "METR time horizons cover defined technical tasks, not full-day job substitution.",
    "The multiplicative decomposition is an editorial simplification, not a measured or calibrated statistical model.",
    "Continuation or comparability of independent evaluation is not assured."
  ],
  "sources": [
    {
      "title": "METR time-horizon methodology and updates",
      "url": "https://metr.org/time-horizons/"
    },
    {
      "title": "Forecasting Research Institute: 80% reliable task length forecaster survey",
      "url": "https://leap.forecastingresearch.org/forecasts/80-8-hour-task-horizon-1"
    },
    {
      "title": "METR limits of task horizons",
      "url": "https://metr.org/notes/2026-01-22-time-horizon-limitations/"
    }
  ],
  "prelock_requirements": [
    "Freeze benchmark suite, 80% horizon statistics, minimum number of tasks and human duration baseline; select successor evaluator if necessary.",
    "Specify how 'publicly available' and 'independent replication' are determined.",
    "Replace judgmental factors with a reproducible calibrated reference class and independent reviewer assessment."
  ],
  "forecaster_signoff": false,
  "independent_review_count": 0,
  "locked": false,
  "scoring_eligible": false,
  "automatic_score": false,
  "source_candidate": "/data/forecasts/research/a-001.json",
  "candidate_page": "/forecast/ai-agents-eight-hour-tasks-2028/",
  "version": "0.2",
  "audit_date": "2026-10-23",
  "pre_review_result": "HOLD_NOT_READY_FOR_INDEPENDENT_REVIEW_SIGNOFF",
  "precise_model_working_probability": 85.5,
  "rounding_rule": "nearest whole percentage point; exact .5 rounds upward",
  "previous_displayed_percentage": 85,
  "correction": "v0.1 displayed 85% and 45% for the exact 85.5% and 45.5% calculations; v0.2 corrects to 86% and 46% under the stated standard rounding rule. E-001 remains 54%.",
  "audit_record": "/data/forecasts/review/a-001-method-audit-20261023-v02.json",
  "previous_version": "/data/forecasts/calibration/a-001-working-20261023.json",
  "external_reviewer_signoffs": 0,
  "independent_validation_complete": false
}