{
  "schema": "feldra-ai-comparison-v1",
  "date": "2026-11-09",
  "scope": "Online automated correctness workflow for six instances in three known small RTL families, two repetitions. No PPA or total engineering-productivity comparison.",
  "source_files": [
    "sources/ai-summary.json",
    "sources/ai-attempts.json"
  ],
  "arms": [
    {
      "arm": "contract_generator_r1",
      "accepted": 6,
      "cases": 6,
      "generation_seconds": 0.00015940004959702492,
      "tool_seconds": 8.619129899539985,
      "total_online_seconds": 8.619289299589582,
      "tool_calls": 24,
      "estimated_gross_usd": 0,
      "statuses": {
        "products_4_s2": "accepted",
        "products_7_s4": "accepted",
        "contexts_2_w24": "accepted",
        "contexts_4_w32": "accepted",
        "buffer_w13": "accepted",
        "buffer_w29": "accepted"
      }
    },
    {
      "arm": "contract_generator_r2",
      "accepted": 6,
      "cases": 6,
      "generation_seconds": 0.00022380007430911064,
      "tool_seconds": 7.96597600041423,
      "total_online_seconds": 7.966199800488539,
      "tool_calls": 24,
      "estimated_gross_usd": 0,
      "statuses": {
        "products_4_s2": "accepted",
        "products_7_s4": "accepted",
        "contexts_2_w24": "accepted",
        "contexts_4_w32": "accepted",
        "buffer_w13": "accepted",
        "buffer_w29": "accepted"
      }
    },
    {
      "arm": "argon_source_only_r1",
      "accepted": 6,
      "cases": 6,
      "generation_seconds": 39.361268599983305,
      "tool_seconds": 6.3984364999923855,
      "total_online_seconds": 45.75970509997569,
      "tool_calls": 24,
      "estimated_gross_usd": 0.08745,
      "statuses": {
        "products_4_s2": "accepted",
        "products_7_s4": "accepted",
        "contexts_2_w24": "accepted",
        "contexts_4_w32": "accepted",
        "buffer_w13": "accepted",
        "buffer_w29": "accepted"
      }
    },
    {
      "arm": "argon_with_feedback_r1",
      "accepted": 6,
      "cases": 6,
      "generation_seconds": 135.78169919992797,
      "tool_seconds": 6.8534374999580905,
      "total_online_seconds": 142.63513669988606,
      "tool_calls": 24,
      "estimated_gross_usd": 0.0935264,
      "statuses": {
        "products_4_s2": "accepted",
        "products_7_s4": "accepted",
        "contexts_2_w24": "accepted",
        "contexts_4_w32": "accepted",
        "buffer_w13": "accepted",
        "buffer_w29": "accepted"
      }
    },
    {
      "arm": "flash_source_only_r1",
      "accepted": 6,
      "cases": 6,
      "generation_seconds": 6.706304399995133,
      "tool_seconds": 5.455519300303422,
      "total_online_seconds": 12.161823700298555,
      "tool_calls": 24,
      "estimated_gross_usd": 0.0189783,
      "statuses": {
        "products_4_s2": "accepted",
        "products_7_s4": "accepted",
        "contexts_2_w24": "accepted",
        "contexts_4_w32": "accepted",
        "buffer_w13": "accepted",
        "buffer_w29": "accepted"
      }
    },
    {
      "arm": "flash_with_feedback_r1",
      "accepted": 6,
      "cases": 6,
      "generation_seconds": 6.306676400010474,
      "tool_seconds": 5.714357899618335,
      "total_online_seconds": 12.021034299628809,
      "tool_calls": 24,
      "estimated_gross_usd": 0.0213675,
      "statuses": {
        "products_4_s2": "accepted",
        "products_7_s4": "accepted",
        "contexts_2_w24": "accepted",
        "contexts_4_w32": "accepted",
        "buffer_w13": "accepted",
        "buffer_w29": "accepted"
      }
    },
    {
      "arm": "argon_with_feedback_r2",
      "accepted": 6,
      "cases": 6,
      "generation_seconds": 40.45348559995182,
      "tool_seconds": 7.023685300257057,
      "total_online_seconds": 47.477170900208876,
      "tool_calls": 24,
      "estimated_gross_usd": 0.0935264,
      "statuses": {
        "products_4_s2": "accepted",
        "products_7_s4": "accepted",
        "contexts_2_w24": "accepted",
        "contexts_4_w32": "accepted",
        "buffer_w13": "accepted",
        "buffer_w29": "accepted"
      }
    },
    {
      "arm": "argon_source_only_r2",
      "accepted": 6,
      "cases": 6,
      "generation_seconds": 35.451363099971786,
      "tool_seconds": 5.662516599870287,
      "total_online_seconds": 41.11387969984207,
      "tool_calls": 24,
      "estimated_gross_usd": 0.054626,
      "statuses": {
        "products_4_s2": "accepted",
        "products_7_s4": "accepted",
        "contexts_2_w24": "accepted",
        "contexts_4_w32": "accepted",
        "buffer_w13": "accepted",
        "buffer_w29": "accepted"
      }
    },
    {
      "arm": "flash_with_feedback_r2",
      "accepted": 6,
      "cases": 6,
      "generation_seconds": 7.384333799942397,
      "tool_seconds": 5.559556599822827,
      "total_online_seconds": 12.943890399765223,
      "tool_calls": 24,
      "estimated_gross_usd": 0.021252,
      "statuses": {
        "products_4_s2": "accepted",
        "products_7_s4": "accepted",
        "contexts_2_w24": "accepted",
        "contexts_4_w32": "accepted",
        "buffer_w13": "accepted",
        "buffer_w29": "accepted"
      }
    },
    {
      "arm": "flash_source_only_r2",
      "accepted": 6,
      "cases": 6,
      "generation_seconds": 6.64607330004219,
      "tool_seconds": 5.430529600009322,
      "total_online_seconds": 12.076602900051512,
      "tool_calls": 24,
      "estimated_gross_usd": 0.01890405,
      "statuses": {
        "products_4_s2": "accepted",
        "products_7_s4": "accepted",
        "contexts_2_w24": "accepted",
        "contexts_4_w32": "accepted",
        "buffer_w13": "accepted",
        "buffer_w29": "accepted"
      }
    }
  ],
  "groups": [
    {
      "id": "contract_generator",
      "label": "Vertragsgenerator",
      "arm_ids": [
        "contract_generator_r1",
        "contract_generator_r2"
      ],
      "median_online_seconds": 8.29274455003906,
      "online_seconds": [
        8.619289299589582,
        7.966199800488539
      ]
    },
    {
      "id": "flash_source_only",
      "label": "Gemini 3.8 Flash · ohne Diagnose",
      "arm_ids": [
        "flash_source_only_r1",
        "flash_source_only_r2"
      ],
      "median_online_seconds": 12.119213300175034,
      "online_seconds": [
        12.161823700298555,
        12.076602900051512
      ]
    },
    {
      "id": "flash_with_feedback",
      "label": "Gemini 3.8 Flash · mit Diagnose",
      "arm_ids": [
        "flash_with_feedback_r1",
        "flash_with_feedback_r2"
      ],
      "median_online_seconds": 12.482462349697016,
      "online_seconds": [
        12.021034299628809,
        12.943890399765223
      ]
    },
    {
      "id": "argon_source_only",
      "label": "Gemini 4 Argon · ohne Diagnose",
      "arm_ids": [
        "argon_source_only_r1",
        "argon_source_only_r2"
      ],
      "median_online_seconds": 43.43679239990888,
      "online_seconds": [
        45.75970509997569,
        41.11387969984207
      ]
    },
    {
      "id": "argon_with_feedback",
      "label": "Gemini 4 Argon · mit Diagnose",
      "arm_ids": [
        "argon_with_feedback_r1",
        "argon_with_feedback_r2"
      ],
      "median_online_seconds": 95.05615380004747,
      "online_seconds": [
        142.63513669988606,
        47.477170900208876
      ]
    }
  ],
  "audit_excerpt": {
    "successful_vector_comparisons": 1474560,
    "scored_icarus_calls": 240,
    "common_control_icarus_calls": 36,
    "generation_attempts": 9,
    "count_tokens_calls": 4,
    "known_gross_provider_usd": 0.40963065,
    "unconfirmed_timeout_max_reserve_usd": 0.2094444,
    "conservative_incurred_upper_usd": 0.61907505,
    "actual_invoice_available": false,
    "observed_control_start_to_scoring_end_wallclock_seconds": 580.403354,
    "common_control_seconds": 12.974715600023046,
    "common_count_tokens_seconds": 3.2070717997848988,
    "continuation_authentication_seconds": 3.1029691000003368,
    "host_load_isolated": false,
    "human_or_model_assisted_content_interventions_after_freeze": 0,
    "transport_or_runner_intervention_count_after_freeze": 4,
    "postprocessing_audit_corrections": 1
  },
  "audit_original_sha256": "6b75e9392050ae2fbc1596c1bf4e6aa95311e55594b334f74b7e35edd59ff1b4",
  "excluded_costs": [
    "Generator, oracle and experiment authoring and review",
    "Model pretraining",
    "Codex subscription and token allocation",
    "Local electrical energy",
    "Actual invoice and credits"
  ],
  "tasks": [
    {
      "id": "buffer_w13",
      "family": "buffer",
      "traces": [
        4096,
        4096,
        4096
      ]
    },
    {
      "id": "buffer_w29",
      "family": "buffer",
      "traces": [
        4096,
        4096,
        4096
      ]
    },
    {
      "id": "contexts_2_w24",
      "family": "contexts",
      "traces": [
        4096,
        4096,
        4096
      ]
    },
    {
      "id": "contexts_4_w32",
      "family": "contexts",
      "traces": [
        4096,
        4096,
        4096
      ]
    },
    {
      "id": "products_4_s2",
      "family": "products",
      "traces": [
        32768,
        8192,
        8192
      ]
    },
    {
      "id": "products_7_s4",
      "family": "products",
      "traces": [
        32768,
        8192,
        8192
      ]
    }
  ]
}
