{
  "schema_version": "2026-05-24",
  "suite_id": "claude-context-limit3",
  "source": "scripts/benchmark-repeat.sh aggregate.json",
  "reproducibility": {
    "manifest_sha256": "911824a2e8ef2751630029f253ee53ac158c747063ad01c6103bde7feef688e3",
    "manifest_schema_version": "2026-05-24",
    "harness": "claude",
    "required_repeats": 3,
    "fresh_context_contract": "Each repeat invokes one full baseline/optimized harness run. The harness creates fresh detached git worktrees and launches separate Claude/Codex task sessions for baseline and optimized runs.",
    "base_ref": "b96b8a7f5cc57c4335bc7bc85ec726c836ed0996",
    "quality_command": "go test ./..."
  },
  "required_repeats": 3,
  "completed_repeats": 3,
  "quality_passed": true,
  "failed_repeats": [],
  "delta_stats": {
    "avoidable_waste_high_points": {
      "n": 3,
      "mean": -0.6666666666666666,
      "median": 0,
      "min": -2,
      "max": 0,
      "stdev": 1.1547005383792515
    },
    "avoidable_waste_low_points": {
      "n": 3,
      "mean": -0.6666666666666666,
      "median": 0,
      "min": -2,
      "max": 0,
      "stdev": 1.1547005383792515
    },
    "claude_cache_creation_input_tokens": {
      "n": 3,
      "mean": 4330.333333333333,
      "median": 2834,
      "min": 2291,
      "max": 7866,
      "stdev": 3073.990294931546
    },
    "claude_cache_read_input_tokens": {
      "n": 3,
      "mean": 49340.333333333336,
      "median": 49157,
      "min": -38591,
      "max": 137455,
      "stdev": 88023.14319162509
    },
    "claude_output_tokens": {
      "n": 3,
      "mean": 1169.0,
      "median": 1374,
      "min": 677,
      "max": 1456,
      "stdev": 428.05256686533255
    },
    "claude_total_cost_usd": {
      "n": 3,
      "mean": 0.048434183333333346,
      "median": 0.047379599999999994,
      "min": 0.006412950000000001,
      "max": 0.09151000000000004,
      "stdev": 0.042558325732820305
    },
    "context_growth_events": {
      "n": 3,
      "mean": -0.3333333333333333,
      "median": 0,
      "min": -1,
      "max": 0,
      "stdev": 0.5773502691896257
    },
    "estimated_tokens": {
      "n": 3,
      "mean": 7327.0,
      "median": 5221,
      "min": 3424,
      "max": 13336,
      "stdev": 5280.943381631732
    },
    "failed_commands": {
      "n": 3,
      "mean": 0.0,
      "median": 0,
      "min": 0,
      "max": 0,
      "stdev": 0.0
    },
    "rereads": {
      "n": 3,
      "mean": 0.0,
      "median": 0,
      "min": 0,
      "max": 0,
      "stdev": 0.0
    },
    "tool_output_tokens": {
      "n": 3,
      "mean": 4169.666666666667,
      "median": 1492,
      "min": 1320,
      "max": 9697,
      "stdev": 4787.583558887859
    }
  },
  "run_dirs": [
    "run-01",
    "run-02",
    "run-03"
  ],
  "promotion_policy": "recommendation_evidence",
  "candidate_source_url": null,
  "candidate_issue": null,
  "published_api_cost_estimate": {
    "harness": "claude",
    "complete_cost_surface": true,
    "multi_session_runs_not_in_stdout_usage": [],
    "rates": {
      "model_assumption": "Claude Sonnet 4.6",
      "input_tokens_per_mtok_usd": 3.0,
      "cache_creation_input_tokens_per_mtok_usd": 6.0,
      "cache_read_input_tokens_per_mtok_usd": 0.3,
      "output_tokens_per_mtok_usd": 15.0
    },
    "baseline_mean_usd": 0.22295049999999997,
    "optimized_mean_usd": 0.2809886,
    "delta_mean_usd": 0.0580381,
    "savings_rate": -0.26031832177994674,
    "savings_percent": -26.031832177994673,
    "scaled_examples": {
      "savings_rate": -0.26031832177994674,
      "savings_percent": -26.031832177994673,
      "weekly_100_usd": -26.031832177994673,
      "weekly_500_usd": -130.15916088997338,
      "monthly_2000_usd": -520.6366435598935,
      "monthly_5000_usd": -1301.5916088997337,
      "monthly_10000_usd": -2603.1832177994675,
      "basis": "Scaled linearly from the suite's optimized-vs-baseline published API-rate delta: savings = comparable baseline spend * savings_percent."
    },
    "delta_min_usd": 0.011426700000000012,
    "delta_max_usd": 0.10907250000000002,
    "runs": [
      {
        "run": "run-01",
        "baseline_usd": 0.2276802,
        "optimized_usd": 0.3367527,
        "delta_usd": 0.10907250000000002
      },
      {
        "run": "run-02",
        "baseline_usd": 0.21886529999999998,
        "optimized_usd": 0.230292,
        "delta_usd": 0.011426700000000012
      },
      {
        "run": "run-03",
        "baseline_usd": 0.222306,
        "optimized_usd": 0.2759211,
        "delta_usd": 0.05361509999999997
      }
    ],
    "note": "Uses exposed Claude Code aggregate usage fields repriced at published Claude Sonnet 4.6 API rates. Claude Code native billing can differ because raw runs also report internal model usage."
  },
  "privacy_boundary": "Raw Claude/Codex logs, raw prompts, local paths, and secrets are not included in public aggregate artifacts."
}