{
 "schema": "modelcensus/lab-run@1",
 "lab": "bs-detector",
 "runId": "2026-09-27-r2",
 "recordedAt": "2026-09-27T14:35:33.922Z",
 "commit": "a806f5d",
 "labSha256": "f38da8ce814f53e0ca03e6cd8afbf3c56feab73ad67b5a8500c319651efe6cf2",
 "node": "v22.20.0",
 "calls": [
  {
   "input": "agentic-hollow",
   "arm": "jev",
   "repeat": 1,
   "latencyMs": 191,
   "costUsd": 0.00002352,
   "inputTokens": 560,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.48,
    "substance_distribution": [
     0.67,
     0.24,
     0.08,
     0.01,
     0,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.81,
    "filler": 0.93,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.95,
    "flaw_distribution": {
     "none": 0,
     "vanity_metric": 0.02,
     "buzzword_stacking": 0.97,
     "hand_waving": 0.01
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "63da952e1cda6cf63ae6a79969502ab6aa7c96cd6286cbd59ca141dc016be198",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our AI-native platform leverages cutting-edge agentic workflows to unlock 10x productivity across the enterprise. Built for the future of work, it seamlessly transforms how teams collaborate, innovate and deliver value at scale."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.48,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.67,
       "1": 0.24,
       "2": 0.08,
       "3": 0.01,
       "4": 0,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.81
     },
     "filler": {
      "type": "noul",
      "noul": 0.93
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "none": 0,
       "vanity_metric": 0.02,
       "buzzword_stacking": 0.97,
       "hand_waving": 0.01
      },
      "confidence": 0.95
     }
    },
    "usage": {
     "input_tokens": 560,
     "output_tokens": 91,
     "cost": 0.00002352
    },
    "id": "gen-dec-1790519736-nD38DBa2mRQca0N20Wwv",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "agentic-hollow",
   "arm": "llm",
   "repeat": 1,
   "latencyMs": 779,
   "costUsd": 0.00007365,
   "inputTokens": 371,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 3,
    "filler": 0.8,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "64ef554d6fdc0baa1ef1daea9ac52d0b1d15945fbb1b76fff2c146abc059389b",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur AI-native platform leverages cutting-edge agentic workflows to unlock 10x productivity across the enterprise. Built for the future of work, it seamlessly transforms how teams collaborate, innovate and deliver value at scale."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519736-4bYdtaWW6T36JQqUIY7l",
    "object": "chat.completion",
    "created": 1790519736,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":3,\"filler\":0.8,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 371,
     "completion_tokens": 30,
     "total_tokens": 401,
     "cost": 0.00007365,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007365,
      "upstream_inference_prompt_cost": 0.00005565,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "agentic-backed",
   "arm": "jev",
   "repeat": 1,
   "latencyMs": 164,
   "costUsd": 0.000024528,
   "inputTokens": 584,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 7.39,
    "substance_distribution": [
     0,
     0.01,
     0.02,
     0.06,
     0.09,
     0.11,
     0.16,
     0.22,
     0.22,
     0.11
    ],
    "substance_confidence": 0.41,
    "filler": 0.15,
    "flaw": "none",
    "flaw_confidence": 0.86,
    "flaw_distribution": {
     "buzzword_stacking": 0,
     "none": 0.89,
     "vanity_metric": 0.08,
     "hand_waving": 0.03
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "83b552b87a442b248929456879f103598f01a46483f861b20f29a53f242c6593",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our agent drafts first-response replies for support tickets. In a six-week pilot with 14 agents at a 40-person SaaS company, median first-response time fell from 3h 10m to 41m; agents edited 62% of drafts before sending. It does not handle billing disputes."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 6.39,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0,
       "1": 0.01,
       "2": 0.02,
       "3": 0.06,
       "4": 0.09,
       "5": 0.11,
       "6": 0.16,
       "7": 0.22,
       "8": 0.22,
       "9": 0.11
      },
      "confidence": 0.41
     },
     "filler": {
      "type": "noul",
      "noul": 0.15
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "buzzword_stacking": 0,
       "none": 0.89,
       "vanity_metric": 0.08,
       "hand_waving": 0.03
      },
      "confidence": 0.86
     }
    },
    "usage": {
     "input_tokens": 584,
     "output_tokens": 86,
     "cost": 0.000024528
    },
    "id": "gen-dec-1790519737-f5VkPc5GkDNSjgdLL7c2",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "agentic-backed",
   "arm": "llm",
   "repeat": 1,
   "latencyMs": 731,
   "costUsd": 0.00007485,
   "inputTokens": 391,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "e421cf5e0d015b409956f270fe540c5f0f3120a96ea0ad9e871b03f2fe35e00b",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur agent drafts first-response replies for support tickets. In a six-week pilot with 14 agents at a 40-person SaaS company, median first-response time fell from 3h 10m to 41m; agents edited 62% of drafts before sending. It does not handle billing disputes."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519737-QbfZcOqTZ2ZtY7Exsn6f",
    "object": "chat.completion",
    "created": 1790519737,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 391,
     "completion_tokens": 27,
     "total_tokens": 418,
     "cost": 0.00007485,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007485,
      "upstream_inference_prompt_cost": 0.00005865,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "security-hollow",
   "arm": "jev",
   "repeat": 1,
   "latencyMs": 213,
   "costUsd": 0.000023562,
   "inputTokens": 561,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.61,
    "substance_distribution": [
     0.59,
     0.29,
     0.1,
     0.02,
     0,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.76,
    "filler": 0.94,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.23,
    "flaw_distribution": {
     "vanity_metric": 0.33,
     "buzzword_stacking": 0.42,
     "hand_waving": 0.25,
     "none": 0
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "d0ac5827a3b25d13348ed9777a2c32a48af33cb387c80ec52c5faf77838c61c0",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Introducing the world's most advanced AI-powered security platform. Our revolutionary engine stops 99.9% of threats before they happen, giving you total peace of mind in an ever-evolving threat landscape."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.61,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.59,
       "1": 0.29,
       "2": 0.1,
       "3": 0.02,
       "4": 0,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.76
     },
     "filler": {
      "type": "noul",
      "noul": 0.94
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "vanity_metric": 0.33,
       "buzzword_stacking": 0.42,
       "hand_waving": 0.25,
       "none": 0
      },
      "confidence": 0.23
     }
    },
    "usage": {
     "input_tokens": 561,
     "output_tokens": 91,
     "cost": 0.000023562
    },
    "id": "gen-dec-1790519738-EMynkFObJwIKELBB3rLz",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "security-hollow",
   "arm": "llm",
   "repeat": 1,
   "latencyMs": 655,
   "costUsd": 0.0000732,
   "inputTokens": 368,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 3,
    "filler": 0.8,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "6f0e1fd6c68c1ad53fb9621a785302c4c2b039e3d2ab5e1bd13c0d59a95ff4e9",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nIntroducing the world's most advanced AI-powered security platform. Our revolutionary engine stops 99.9% of threats before they happen, giving you total peace of mind in an ever-evolving threat landscape."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519738-6BpTQlvxjFP32WcsdnHk",
    "object": "chat.completion",
    "created": 1790519738,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":3,\"filler\":0.8,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 368,
     "completion_tokens": 30,
     "total_tokens": 398,
     "cost": 0.0000732,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.0000732,
      "upstream_inference_prompt_cost": 0.0000552,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "security-backed",
   "arm": "jev",
   "repeat": 1,
   "latencyMs": 235,
   "costUsd": 0.00002457,
   "inputTokens": 585,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 6.49,
    "substance_distribution": [
     0.01,
     0.05,
     0.05,
     0.06,
     0.14,
     0.12,
     0.19,
     0.22,
     0.13,
     0.03
    ],
    "substance_confidence": 0.25,
    "filler": 0.18,
    "flaw": "none",
    "flaw_confidence": 0.39,
    "flaw_distribution": {
     "hand_waving": 0.03,
     "none": 0.55,
     "buzzword_stacking": 0,
     "vanity_metric": 0.42
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "b6d5c5a3cb6bc7b6d2eb0e36a4763d6e34df7c59aa9d502a06a81d3484a32abc",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our scanner flags leaked API keys in pull requests before merge. On 2,300 public repositories with known leaks from 2025, it caught 1,941 (84%) with 37 false positives. It misses keys split across files, and we list the known gaps on our docs page."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 5.49,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.01,
       "1": 0.05,
       "2": 0.05,
       "3": 0.06,
       "4": 0.14,
       "5": 0.12,
       "6": 0.19,
       "7": 0.22,
       "8": 0.13,
       "9": 0.03
      },
      "confidence": 0.25
     },
     "filler": {
      "type": "noul",
      "noul": 0.18
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "hand_waving": 0.03,
       "none": 0.55,
       "buzzword_stacking": 0,
       "vanity_metric": 0.42
      },
      "confidence": 0.39
     }
    },
    "usage": {
     "input_tokens": 585,
     "output_tokens": 86,
     "cost": 0.00002457
    },
    "id": "gen-dec-1790519738-mGnL12g46xKw2ekAVuWc",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "security-backed",
   "arm": "llm",
   "repeat": 1,
   "latencyMs": 619,
   "costUsd": 0.00007455,
   "inputTokens": 389,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "e4eb040c67bee5b01347f6a7ea924fbb6a2daae61f9fb722d1e36dcc2d4e954c",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur scanner flags leaked API keys in pull requests before merge. On 2,300 public repositories with known leaks from 2025, it caught 1,941 (84%) with 37 false positives. It misses keys split across files, and we list the known gaps on our docs page."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519739-UWyM6j0a8PyXEHpp1vBX",
    "object": "chat.completion",
    "created": 1790519739,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 389,
     "completion_tokens": 27,
     "total_tokens": 416,
     "cost": 0.00007455,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007455,
      "upstream_inference_prompt_cost": 0.00005835,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "funding-hollow",
   "arm": "jev",
   "repeat": 1,
   "latencyMs": 168,
   "costUsd": 0.000023562,
   "inputTokens": 561,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.96,
    "substance_distribution": [
     0.39,
     0.39,
     0.16,
     0.05,
     0.01,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.61,
    "filler": 0.91,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.91,
    "flaw_distribution": {
     "vanity_metric": 0,
     "buzzword_stacking": 0.94,
     "none": 0,
     "hand_waving": 0.06
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "1f0c300c7eac342f879e16082930f1201ca59048a112c54de9b00494b7442ea7",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "We're thrilled to announce our Series A to supercharge our mission of democratizing intelligence. With world-class investors behind us, we're poised to redefine the category and deliver unprecedented value to customers everywhere."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.96,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.39,
       "1": 0.39,
       "2": 0.16,
       "3": 0.05,
       "4": 0.01,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.61
     },
     "filler": {
      "type": "noul",
      "noul": 0.91
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "vanity_metric": 0,
       "buzzword_stacking": 0.94,
       "none": 0,
       "hand_waving": 0.06
      },
      "confidence": 0.91
     }
    },
    "usage": {
     "input_tokens": 561,
     "output_tokens": 91,
     "cost": 0.000023562
    },
    "id": "gen-dec-1790519739-IIMdBOn3y3Anw4ohcb26",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "funding-hollow",
   "arm": "llm",
   "repeat": 1,
   "latencyMs": 800,
   "costUsd": 0.00007305,
   "inputTokens": 367,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 2,
    "filler": 0.85,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "64f587d56624bc07c8a5cd7bfb6ee309e01720f37e326702557db6c510013917",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nWe're thrilled to announce our Series A to supercharge our mission of democratizing intelligence. With world-class investors behind us, we're poised to redefine the category and deliver unprecedented value to customers everywhere."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519739-qjFLZWL6v3azFyUjfhOD",
    "object": "chat.completion",
    "created": 1790519739,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":2,\"filler\":0.85,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 367,
     "completion_tokens": 30,
     "total_tokens": 397,
     "cost": 0.00007305,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007305,
      "upstream_inference_prompt_cost": 0.00005505,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "funding-backed",
   "arm": "jev",
   "repeat": 1,
   "latencyMs": 172,
   "costUsd": 0.000024402,
   "inputTokens": 581,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 7.33,
    "substance_distribution": [
     0,
     0.01,
     0.04,
     0.07,
     0.08,
     0.07,
     0.15,
     0.24,
     0.25,
     0.09
    ],
    "substance_confidence": 0.26,
    "filler": 0.19,
    "flaw": "none",
    "flaw_confidence": 0.47,
    "flaw_distribution": {
     "vanity_metric": 0.34,
     "none": 0.6,
     "buzzword_stacking": 0.01,
     "hand_waving": 0.05
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "ef7608ea4190ed19a1a27952905f6af21f0654930c19cf1c07f4be8aee59350b",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "We raised $12M to expand our invoice-matching product from 3 to 9 ERP integrations by June. Today 210 finance teams use it; the median team reconciles 4,800 invoices a month and clears exceptions in under a day, down from four."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 6.33,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0,
       "1": 0.01,
       "2": 0.04,
       "3": 0.07,
       "4": 0.08,
       "5": 0.07,
       "6": 0.15,
       "7": 0.24,
       "8": 0.25,
       "9": 0.09
      },
      "confidence": 0.26
     },
     "filler": {
      "type": "noul",
      "noul": 0.19
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "vanity_metric": 0.34,
       "none": 0.6,
       "buzzword_stacking": 0.01,
       "hand_waving": 0.05
      },
      "confidence": 0.47
     }
    },
    "usage": {
     "input_tokens": 581,
     "output_tokens": 86,
     "cost": 0.000024402
    },
    "id": "gen-dec-1790519740-TvBrPZu6QBMtKWko8HM7",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "funding-backed",
   "arm": "llm",
   "repeat": 1,
   "latencyMs": 1325,
   "costUsd": 0.0000738,
   "inputTokens": 384,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 7,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "fde93f61e1fca7e422f242f0cf572bdb00b531825d6d1a8003cc44db2554a01d",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nWe raised $12M to expand our invoice-matching product from 3 to 9 ERP integrations by June. Today 210 finance teams use it; the median team reconciles 4,800 invoices a month and clears exceptions in under a day, down from four."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519740-LXIoNNksZ1eRpyEZxL7w",
    "object": "chat.completion",
    "created": 1790519740,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":7,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 384,
     "completion_tokens": 27,
     "total_tokens": 411,
     "cost": 0.0000738,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.0000738,
      "upstream_inference_prompt_cost": 0.0000576,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "agentic-hollow",
   "arm": "jev",
   "repeat": 2,
   "latencyMs": 221,
   "costUsd": 0.00002352,
   "inputTokens": 560,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.51,
    "substance_distribution": [
     0.66,
     0.25,
     0.08,
     0.01,
     0,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.8,
    "filler": 0.93,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.95,
    "flaw_distribution": {
     "hand_waving": 0.02,
     "vanity_metric": 0.02,
     "buzzword_stacking": 0.96,
     "none": 0
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "63da952e1cda6cf63ae6a79969502ab6aa7c96cd6286cbd59ca141dc016be198",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our AI-native platform leverages cutting-edge agentic workflows to unlock 10x productivity across the enterprise. Built for the future of work, it seamlessly transforms how teams collaborate, innovate and deliver value at scale."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.51,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.66,
       "1": 0.25,
       "2": 0.08,
       "3": 0.01,
       "4": 0,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.8
     },
     "filler": {
      "type": "noul",
      "noul": 0.93
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "hand_waving": 0.02,
       "vanity_metric": 0.02,
       "buzzword_stacking": 0.96,
       "none": 0
      },
      "confidence": 0.95
     }
    },
    "usage": {
     "input_tokens": 560,
     "output_tokens": 91,
     "cost": 0.00002352
    },
    "id": "gen-dec-1790519742-A8BlHuDwAo9G2QZlS60X",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "agentic-hollow",
   "arm": "llm",
   "repeat": 2,
   "latencyMs": 1595,
   "costUsd": 0.00007365,
   "inputTokens": 371,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 3,
    "filler": 0.8,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "64ef554d6fdc0baa1ef1daea9ac52d0b1d15945fbb1b76fff2c146abc059389b",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur AI-native platform leverages cutting-edge agentic workflows to unlock 10x productivity across the enterprise. Built for the future of work, it seamlessly transforms how teams collaborate, innovate and deliver value at scale."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519742-eWE82y93dqkYUIMpIY95",
    "object": "chat.completion",
    "created": 1790519742,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":3,\"filler\":0.8,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 371,
     "completion_tokens": 30,
     "total_tokens": 401,
     "cost": 0.00007365,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007365,
      "upstream_inference_prompt_cost": 0.00005565,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "agentic-backed",
   "arm": "jev",
   "repeat": 2,
   "latencyMs": 235,
   "costUsd": 0.000024528,
   "inputTokens": 584,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 7.37,
    "substance_distribution": [
     0,
     0.01,
     0.02,
     0.06,
     0.09,
     0.1,
     0.16,
     0.23,
     0.22,
     0.11
    ],
    "substance_confidence": 0.4,
    "filler": 0.16,
    "flaw": "none",
    "flaw_confidence": 0.85,
    "flaw_distribution": {
     "hand_waving": 0.03,
     "none": 0.89,
     "buzzword_stacking": 0,
     "vanity_metric": 0.08
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "83b552b87a442b248929456879f103598f01a46483f861b20f29a53f242c6593",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our agent drafts first-response replies for support tickets. In a six-week pilot with 14 agents at a 40-person SaaS company, median first-response time fell from 3h 10m to 41m; agents edited 62% of drafts before sending. It does not handle billing disputes."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 6.37,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0,
       "1": 0.01,
       "2": 0.02,
       "3": 0.06,
       "4": 0.09,
       "5": 0.1,
       "6": 0.16,
       "7": 0.23,
       "8": 0.22,
       "9": 0.11
      },
      "confidence": 0.4
     },
     "filler": {
      "type": "noul",
      "noul": 0.16
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "hand_waving": 0.03,
       "none": 0.89,
       "buzzword_stacking": 0,
       "vanity_metric": 0.08
      },
      "confidence": 0.85
     }
    },
    "usage": {
     "input_tokens": 584,
     "output_tokens": 86,
     "cost": 0.000024528
    },
    "id": "gen-dec-1790519744-1PbS7ALeWnKckKdsdxhV",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "agentic-backed",
   "arm": "llm",
   "repeat": 2,
   "latencyMs": 687,
   "costUsd": 0.00007485,
   "inputTokens": 391,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "e421cf5e0d015b409956f270fe540c5f0f3120a96ea0ad9e871b03f2fe35e00b",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur agent drafts first-response replies for support tickets. In a six-week pilot with 14 agents at a 40-person SaaS company, median first-response time fell from 3h 10m to 41m; agents edited 62% of drafts before sending. It does not handle billing disputes."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519744-gyhvWt29gpZmNqhSJ2C5",
    "object": "chat.completion",
    "created": 1790519744,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 391,
     "completion_tokens": 27,
     "total_tokens": 418,
     "cost": 0.00007485,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007485,
      "upstream_inference_prompt_cost": 0.00005865,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "security-hollow",
   "arm": "jev",
   "repeat": 2,
   "latencyMs": 183,
   "costUsd": 0.000023562,
   "inputTokens": 561,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.59,
    "substance_distribution": [
     0.63,
     0.25,
     0.09,
     0.02,
     0,
     0.01,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.76,
    "filler": 0.94,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.19,
    "flaw_distribution": {
     "vanity_metric": 0.34,
     "none": 0,
     "buzzword_stacking": 0.39,
     "hand_waving": 0.27
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "d0ac5827a3b25d13348ed9777a2c32a48af33cb387c80ec52c5faf77838c61c0",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Introducing the world's most advanced AI-powered security platform. Our revolutionary engine stops 99.9% of threats before they happen, giving you total peace of mind in an ever-evolving threat landscape."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.59,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.63,
       "1": 0.25,
       "2": 0.09,
       "3": 0.02,
       "4": 0,
       "5": 0.01,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.76
     },
     "filler": {
      "type": "noul",
      "noul": 0.94
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "vanity_metric": 0.34,
       "none": 0,
       "buzzword_stacking": 0.39,
       "hand_waving": 0.27
      },
      "confidence": 0.19
     }
    },
    "usage": {
     "input_tokens": 561,
     "output_tokens": 91,
     "cost": 0.000023562
    },
    "id": "gen-dec-1790519744-8fFcQRem4AMYCpjAlHJj",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "security-hollow",
   "arm": "llm",
   "repeat": 2,
   "latencyMs": 1468,
   "costUsd": 0.0000732,
   "inputTokens": 368,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 3,
    "filler": 0.85,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "6f0e1fd6c68c1ad53fb9621a785302c4c2b039e3d2ab5e1bd13c0d59a95ff4e9",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nIntroducing the world's most advanced AI-powered security platform. Our revolutionary engine stops 99.9% of threats before they happen, giving you total peace of mind in an ever-evolving threat landscape."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519745-my069gsY9I3HBiR4tMVO",
    "object": "chat.completion",
    "created": 1790519745,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":3,\"filler\":0.85,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 368,
     "completion_tokens": 30,
     "total_tokens": 398,
     "cost": 0.0000732,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.0000732,
      "upstream_inference_prompt_cost": 0.0000552,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "security-backed",
   "arm": "jev",
   "repeat": 2,
   "latencyMs": 187,
   "costUsd": 0.00002457,
   "inputTokens": 585,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 6.39,
    "substance_distribution": [
     0.01,
     0.06,
     0.06,
     0.07,
     0.14,
     0.11,
     0.18,
     0.21,
     0.13,
     0.03
    ],
    "substance_confidence": 0.21,
    "filler": 0.2,
    "flaw": "none",
    "flaw_confidence": 0.36,
    "flaw_distribution": {
     "hand_waving": 0.02,
     "buzzword_stacking": 0,
     "none": 0.53,
     "vanity_metric": 0.45
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "b6d5c5a3cb6bc7b6d2eb0e36a4763d6e34df7c59aa9d502a06a81d3484a32abc",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our scanner flags leaked API keys in pull requests before merge. On 2,300 public repositories with known leaks from 2025, it caught 1,941 (84%) with 37 false positives. It misses keys split across files, and we list the known gaps on our docs page."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 5.39,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.01,
       "1": 0.06,
       "2": 0.06,
       "3": 0.07,
       "4": 0.14,
       "5": 0.11,
       "6": 0.18,
       "7": 0.21,
       "8": 0.13,
       "9": 0.03
      },
      "confidence": 0.21
     },
     "filler": {
      "type": "noul",
      "noul": 0.2
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "hand_waving": 0.02,
       "buzzword_stacking": 0,
       "none": 0.53,
       "vanity_metric": 0.45
      },
      "confidence": 0.36
     }
    },
    "usage": {
     "input_tokens": 585,
     "output_tokens": 86,
     "cost": 0.00002457
    },
    "id": "gen-dec-1790519746-4ObhzDXSL5T0QHPPrFGQ",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "security-backed",
   "arm": "llm",
   "repeat": 2,
   "latencyMs": 1375,
   "costUsd": 0.00007455,
   "inputTokens": 389,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "e4eb040c67bee5b01347f6a7ea924fbb6a2daae61f9fb722d1e36dcc2d4e954c",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur scanner flags leaked API keys in pull requests before merge. On 2,300 public repositories with known leaks from 2025, it caught 1,941 (84%) with 37 false positives. It misses keys split across files, and we list the known gaps on our docs page."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519746-hIRMm3ggqDAKFYOhHClb",
    "object": "chat.completion",
    "created": 1790519746,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 389,
     "completion_tokens": 27,
     "total_tokens": 416,
     "cost": 0.00007455,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007455,
      "upstream_inference_prompt_cost": 0.00005835,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "funding-hollow",
   "arm": "jev",
   "repeat": 2,
   "latencyMs": 175,
   "costUsd": 0.000023562,
   "inputTokens": 561,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.89,
    "substance_distribution": [
     0.45,
     0.35,
     0.14,
     0.05,
     0.01,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.65,
    "filler": 0.91,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.93,
    "flaw_distribution": {
     "none": 0,
     "vanity_metric": 0,
     "buzzword_stacking": 0.95,
     "hand_waving": 0.05
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "1f0c300c7eac342f879e16082930f1201ca59048a112c54de9b00494b7442ea7",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "We're thrilled to announce our Series A to supercharge our mission of democratizing intelligence. With world-class investors behind us, we're poised to redefine the category and deliver unprecedented value to customers everywhere."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.89,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.45,
       "1": 0.35,
       "2": 0.14,
       "3": 0.05,
       "4": 0.01,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.65
     },
     "filler": {
      "type": "noul",
      "noul": 0.91
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "none": 0,
       "vanity_metric": 0,
       "buzzword_stacking": 0.95,
       "hand_waving": 0.05
      },
      "confidence": 0.93
     }
    },
    "usage": {
     "input_tokens": 561,
     "output_tokens": 91,
     "cost": 0.000023562
    },
    "id": "gen-dec-1790519748-QklKfCwZazLfSczNjn4E",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "funding-hollow",
   "arm": "llm",
   "repeat": 2,
   "latencyMs": 1630,
   "costUsd": 0.00007305,
   "inputTokens": 367,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 2,
    "filler": 0.9,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.8
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "64f587d56624bc07c8a5cd7bfb6ee309e01720f37e326702557db6c510013917",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nWe're thrilled to announce our Series A to supercharge our mission of democratizing intelligence. With world-class investors behind us, we're poised to redefine the category and deliver unprecedented value to customers everywhere."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519748-g5fU3fwIGJhSyTOWGM19",
    "object": "chat.completion",
    "created": 1790519748,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":2,\"filler\":0.9,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.8}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 367,
     "completion_tokens": 30,
     "total_tokens": 397,
     "cost": 0.00007305,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007305,
      "upstream_inference_prompt_cost": 0.00005505,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "funding-backed",
   "arm": "jev",
   "repeat": 2,
   "latencyMs": 189,
   "costUsd": 0.000024402,
   "inputTokens": 581,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 7.21,
    "substance_distribution": [
     0,
     0.01,
     0.05,
     0.08,
     0.09,
     0.07,
     0.15,
     0.22,
     0.25,
     0.08
    ],
    "substance_confidence": 0.22,
    "filler": 0.19,
    "flaw": "none",
    "flaw_confidence": 0.55,
    "flaw_distribution": {
     "none": 0.66,
     "buzzword_stacking": 0.01,
     "hand_waving": 0.06,
     "vanity_metric": 0.27
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "ef7608ea4190ed19a1a27952905f6af21f0654930c19cf1c07f4be8aee59350b",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "We raised $12M to expand our invoice-matching product from 3 to 9 ERP integrations by June. Today 210 finance teams use it; the median team reconciles 4,800 invoices a month and clears exceptions in under a day, down from four."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 6.21,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0,
       "1": 0.01,
       "2": 0.05,
       "3": 0.08,
       "4": 0.09,
       "5": 0.07,
       "6": 0.15,
       "7": 0.22,
       "8": 0.25,
       "9": 0.08
      },
      "confidence": 0.22
     },
     "filler": {
      "type": "noul",
      "noul": 0.19
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "none": 0.66,
       "buzzword_stacking": 0.01,
       "hand_waving": 0.06,
       "vanity_metric": 0.27
      },
      "confidence": 0.55
     }
    },
    "usage": {
     "input_tokens": 581,
     "output_tokens": 86,
     "cost": 0.000024402
    },
    "id": "gen-dec-1790519749-XjECDuVbVjQ3y7zN5U4i",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "funding-backed",
   "arm": "llm",
   "repeat": 2,
   "latencyMs": 769,
   "costUsd": 0.0000738,
   "inputTokens": 384,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "fde93f61e1fca7e422f242f0cf572bdb00b531825d6d1a8003cc44db2554a01d",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nWe raised $12M to expand our invoice-matching product from 3 to 9 ERP integrations by June. Today 210 finance teams use it; the median team reconciles 4,800 invoices a month and clears exceptions in under a day, down from four."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519750-JluhakGuvH5LKpaCRkfQ",
    "object": "chat.completion",
    "created": 1790519750,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 384,
     "completion_tokens": 27,
     "total_tokens": 411,
     "cost": 0.0000738,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.0000738,
      "upstream_inference_prompt_cost": 0.0000576,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "agentic-hollow",
   "arm": "jev",
   "repeat": 3,
   "latencyMs": 159,
   "costUsd": 0.00002352,
   "inputTokens": 560,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.5,
    "substance_distribution": [
     0.67,
     0.23,
     0.08,
     0.02,
     0,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.8,
    "filler": 0.93,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.95,
    "flaw_distribution": {
     "vanity_metric": 0.02,
     "buzzword_stacking": 0.97,
     "hand_waving": 0.01,
     "none": 0
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "63da952e1cda6cf63ae6a79969502ab6aa7c96cd6286cbd59ca141dc016be198",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our AI-native platform leverages cutting-edge agentic workflows to unlock 10x productivity across the enterprise. Built for the future of work, it seamlessly transforms how teams collaborate, innovate and deliver value at scale."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.5,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.67,
       "1": 0.23,
       "2": 0.08,
       "3": 0.02,
       "4": 0,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.8
     },
     "filler": {
      "type": "noul",
      "noul": 0.93
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "vanity_metric": 0.02,
       "buzzword_stacking": 0.97,
       "hand_waving": 0.01,
       "none": 0
      },
      "confidence": 0.95
     }
    },
    "usage": {
     "input_tokens": 560,
     "output_tokens": 91,
     "cost": 0.00002352
    },
    "id": "gen-dec-1790519750-ynRUv8vYQNRLXX6j5LgI",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "agentic-hollow",
   "arm": "llm",
   "repeat": 3,
   "latencyMs": 2557,
   "costUsd": 0.00007365,
   "inputTokens": 371,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 3,
    "filler": 0.9,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.85
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "64ef554d6fdc0baa1ef1daea9ac52d0b1d15945fbb1b76fff2c146abc059389b",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur AI-native platform leverages cutting-edge agentic workflows to unlock 10x productivity across the enterprise. Built for the future of work, it seamlessly transforms how teams collaborate, innovate and deliver value at scale."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519751-ePyCXsVlqw9ZLB4QkDt9",
    "object": "chat.completion",
    "created": 1790519751,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":3,\"filler\":0.9,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.85}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 371,
     "completion_tokens": 30,
     "total_tokens": 401,
     "cost": 0.00007365,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007365,
      "upstream_inference_prompt_cost": 0.00005565,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "agentic-backed",
   "arm": "jev",
   "repeat": 3,
   "latencyMs": 190,
   "costUsd": 0.000024528,
   "inputTokens": 584,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 7.36,
    "substance_distribution": [
     0,
     0.01,
     0.02,
     0.06,
     0.09,
     0.1,
     0.17,
     0.24,
     0.21,
     0.1
    ],
    "substance_confidence": 0.42,
    "filler": 0.16,
    "flaw": "none",
    "flaw_confidence": 0.88,
    "flaw_distribution": {
     "buzzword_stacking": 0,
     "vanity_metric": 0.06,
     "none": 0.91,
     "hand_waving": 0.03
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "83b552b87a442b248929456879f103598f01a46483f861b20f29a53f242c6593",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our agent drafts first-response replies for support tickets. In a six-week pilot with 14 agents at a 40-person SaaS company, median first-response time fell from 3h 10m to 41m; agents edited 62% of drafts before sending. It does not handle billing disputes."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 6.36,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0,
       "1": 0.01,
       "2": 0.02,
       "3": 0.06,
       "4": 0.09,
       "5": 0.1,
       "6": 0.17,
       "7": 0.24,
       "8": 0.21,
       "9": 0.1
      },
      "confidence": 0.42
     },
     "filler": {
      "type": "noul",
      "noul": 0.16
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "buzzword_stacking": 0,
       "vanity_metric": 0.06,
       "none": 0.91,
       "hand_waving": 0.03
      },
      "confidence": 0.88
     }
    },
    "usage": {
     "input_tokens": 584,
     "output_tokens": 86,
     "cost": 0.000024528
    },
    "id": "gen-dec-1790519753-6OZKLppW7BymVUzrGfUT",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "agentic-backed",
   "arm": "llm",
   "repeat": 3,
   "latencyMs": 993,
   "costUsd": 0.00007485,
   "inputTokens": 391,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "e421cf5e0d015b409956f270fe540c5f0f3120a96ea0ad9e871b03f2fe35e00b",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur agent drafts first-response replies for support tickets. In a six-week pilot with 14 agents at a 40-person SaaS company, median first-response time fell from 3h 10m to 41m; agents edited 62% of drafts before sending. It does not handle billing disputes."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519753-mAxnoCC5zK7F774l03zi",
    "object": "chat.completion",
    "created": 1790519753,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 391,
     "completion_tokens": 27,
     "total_tokens": 418,
     "cost": 0.00007485,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007485,
      "upstream_inference_prompt_cost": 0.00005865,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "security-hollow",
   "arm": "jev",
   "repeat": 3,
   "latencyMs": 159,
   "costUsd": 0.000023562,
   "inputTokens": 561,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.66,
    "substance_distribution": [
     0.58,
     0.28,
     0.11,
     0.03,
     0,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.74,
    "filler": 0.93,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.17,
    "flaw_distribution": {
     "vanity_metric": 0.34,
     "none": 0,
     "hand_waving": 0.28,
     "buzzword_stacking": 0.38
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "d0ac5827a3b25d13348ed9777a2c32a48af33cb387c80ec52c5faf77838c61c0",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Introducing the world's most advanced AI-powered security platform. Our revolutionary engine stops 99.9% of threats before they happen, giving you total peace of mind in an ever-evolving threat landscape."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.66,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.58,
       "1": 0.28,
       "2": 0.11,
       "3": 0.03,
       "4": 0,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.74
     },
     "filler": {
      "type": "noul",
      "noul": 0.93
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "vanity_metric": 0.34,
       "none": 0,
       "hand_waving": 0.28,
       "buzzword_stacking": 0.38
      },
      "confidence": 0.17
     }
    },
    "usage": {
     "input_tokens": 561,
     "output_tokens": 91,
     "cost": 0.000023562
    },
    "id": "gen-dec-1790519754-e5xWmB0sTbIlmNKbqg86",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "security-hollow",
   "arm": "llm",
   "repeat": 3,
   "latencyMs": 654,
   "costUsd": 0.0000732,
   "inputTokens": 368,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 2,
    "filler": 0.85,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "6f0e1fd6c68c1ad53fb9621a785302c4c2b039e3d2ab5e1bd13c0d59a95ff4e9",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nIntroducing the world's most advanced AI-powered security platform. Our revolutionary engine stops 99.9% of threats before they happen, giving you total peace of mind in an ever-evolving threat landscape."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519755-Ve6ZPpuQEA32hJGCR5a1",
    "object": "chat.completion",
    "created": 1790519755,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":2,\"filler\":0.85,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 368,
     "completion_tokens": 30,
     "total_tokens": 398,
     "cost": 0.0000732,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.0000732,
      "upstream_inference_prompt_cost": 0.0000552,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "security-backed",
   "arm": "jev",
   "repeat": 3,
   "latencyMs": 186,
   "costUsd": 0.00002457,
   "inputTokens": 585,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 6.52,
    "substance_distribution": [
     0.01,
     0.05,
     0.06,
     0.07,
     0.18,
     0.08,
     0.12,
     0.22,
     0.18,
     0.03
    ],
    "substance_confidence": 0.21,
    "filler": 0.19,
    "flaw": "none",
    "flaw_confidence": 0.42,
    "flaw_distribution": {
     "buzzword_stacking": 0,
     "none": 0.56,
     "vanity_metric": 0.41,
     "hand_waving": 0.03
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "b6d5c5a3cb6bc7b6d2eb0e36a4763d6e34df7c59aa9d502a06a81d3484a32abc",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our scanner flags leaked API keys in pull requests before merge. On 2,300 public repositories with known leaks from 2025, it caught 1,941 (84%) with 37 false positives. It misses keys split across files, and we list the known gaps on our docs page."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 5.52,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.01,
       "1": 0.05,
       "2": 0.06,
       "3": 0.07,
       "4": 0.18,
       "5": 0.08,
       "6": 0.12,
       "7": 0.22,
       "8": 0.18,
       "9": 0.03
      },
      "confidence": 0.21
     },
     "filler": {
      "type": "noul",
      "noul": 0.19
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "buzzword_stacking": 0,
       "none": 0.56,
       "vanity_metric": 0.41,
       "hand_waving": 0.03
      },
      "confidence": 0.42
     }
    },
    "usage": {
     "input_tokens": 585,
     "output_tokens": 86,
     "cost": 0.00002457
    },
    "id": "gen-dec-1790519755-PANKIxsk5YrpOxJeQ7cg",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "security-backed",
   "arm": "llm",
   "repeat": 3,
   "latencyMs": 1301,
   "costUsd": 0.00007455,
   "inputTokens": 389,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "e4eb040c67bee5b01347f6a7ea924fbb6a2daae61f9fb722d1e36dcc2d4e954c",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur scanner flags leaked API keys in pull requests before merge. On 2,300 public repositories with known leaks from 2025, it caught 1,941 (84%) with 37 false positives. It misses keys split across files, and we list the known gaps on our docs page."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519755-ZyMbHXYiIGOWgeVbP3nh",
    "object": "chat.completion",
    "created": 1790519755,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 389,
     "completion_tokens": 27,
     "total_tokens": 416,
     "cost": 0.00007455,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007455,
      "upstream_inference_prompt_cost": 0.00005835,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "funding-hollow",
   "arm": "jev",
   "repeat": 3,
   "latencyMs": 151,
   "costUsd": 0.000023562,
   "inputTokens": 561,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.81,
    "substance_distribution": [
     0.45,
     0.39,
     0.13,
     0.02,
     0.01,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.68,
    "filler": 0.91,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9,
    "flaw_distribution": {
     "hand_waving": 0.07,
     "vanity_metric": 0,
     "buzzword_stacking": 0.93,
     "none": 0
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "1f0c300c7eac342f879e16082930f1201ca59048a112c54de9b00494b7442ea7",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "We're thrilled to announce our Series A to supercharge our mission of democratizing intelligence. With world-class investors behind us, we're poised to redefine the category and deliver unprecedented value to customers everywhere."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.81,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.45,
       "1": 0.39,
       "2": 0.13,
       "3": 0.02,
       "4": 0.01,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.68
     },
     "filler": {
      "type": "noul",
      "noul": 0.91
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "hand_waving": 0.07,
       "vanity_metric": 0,
       "buzzword_stacking": 0.93,
       "none": 0
      },
      "confidence": 0.9
     }
    },
    "usage": {
     "input_tokens": 561,
     "output_tokens": 91,
     "cost": 0.000023562
    },
    "id": "gen-dec-1790519757-7nO9cYL6bX20anXM79qS",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "funding-hollow",
   "arm": "llm",
   "repeat": 3,
   "latencyMs": 1368,
   "costUsd": 0.00007305,
   "inputTokens": 367,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 3,
    "filler": 0.8,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "64f587d56624bc07c8a5cd7bfb6ee309e01720f37e326702557db6c510013917",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nWe're thrilled to announce our Series A to supercharge our mission of democratizing intelligence. With world-class investors behind us, we're poised to redefine the category and deliver unprecedented value to customers everywhere."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519757-JrnlI3n6kHhkGImo3m1P",
    "object": "chat.completion",
    "created": 1790519757,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":3,\"filler\":0.8,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 367,
     "completion_tokens": 30,
     "total_tokens": 397,
     "cost": 0.00007305,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007305,
      "upstream_inference_prompt_cost": 0.00005505,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "funding-backed",
   "arm": "jev",
   "repeat": 3,
   "latencyMs": 232,
   "costUsd": 0.000024402,
   "inputTokens": 581,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 7.3,
    "substance_distribution": [
     0,
     0.01,
     0.04,
     0.08,
     0.09,
     0.07,
     0.14,
     0.21,
     0.26,
     0.1
    ],
    "substance_confidence": 0.24,
    "filler": 0.16,
    "flaw": "none",
    "flaw_confidence": 0.47,
    "flaw_distribution": {
     "none": 0.6,
     "buzzword_stacking": 0.01,
     "hand_waving": 0.06,
     "vanity_metric": 0.33
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "ef7608ea4190ed19a1a27952905f6af21f0654930c19cf1c07f4be8aee59350b",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "We raised $12M to expand our invoice-matching product from 3 to 9 ERP integrations by June. Today 210 finance teams use it; the median team reconciles 4,800 invoices a month and clears exceptions in under a day, down from four."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 6.3,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0,
       "1": 0.01,
       "2": 0.04,
       "3": 0.08,
       "4": 0.09,
       "5": 0.07,
       "6": 0.14,
       "7": 0.21,
       "8": 0.26,
       "9": 0.1
      },
      "confidence": 0.24
     },
     "filler": {
      "type": "noul",
      "noul": 0.16
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "none": 0.6,
       "buzzword_stacking": 0.01,
       "hand_waving": 0.06,
       "vanity_metric": 0.33
      },
      "confidence": 0.47
     }
    },
    "usage": {
     "input_tokens": 581,
     "output_tokens": 86,
     "cost": 0.000024402
    },
    "id": "gen-dec-1790519758-YR0aXpXNICYWvUyTYRqJ",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "funding-backed",
   "arm": "llm",
   "repeat": 3,
   "latencyMs": 695,
   "costUsd": 0.0000738,
   "inputTokens": 384,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "fde93f61e1fca7e422f242f0cf572bdb00b531825d6d1a8003cc44db2554a01d",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nWe raised $12M to expand our invoice-matching product from 3 to 9 ERP integrations by June. Today 210 finance teams use it; the median team reconciles 4,800 invoices a month and clears exceptions in under a day, down from four."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519758-PkIYRdrkhB3guQnidspa",
    "object": "chat.completion",
    "created": 1790519758,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 384,
     "completion_tokens": 27,
     "total_tokens": 411,
     "cost": 0.0000738,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.0000738,
      "upstream_inference_prompt_cost": 0.0000576,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "agentic-hollow",
   "arm": "jev",
   "repeat": 4,
   "latencyMs": 230,
   "costUsd": 0.00002352,
   "inputTokens": 560,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.49,
    "substance_distribution": [
     0.6900000000000001,
     0.21,
     0.08,
     0.01,
     0,
     0.01,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.8,
    "filler": 0.94,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.96,
    "flaw_distribution": {
     "hand_waving": 0.01,
     "vanity_metric": 0.02,
     "buzzword_stacking": 0.97,
     "none": 0
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "63da952e1cda6cf63ae6a79969502ab6aa7c96cd6286cbd59ca141dc016be198",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our AI-native platform leverages cutting-edge agentic workflows to unlock 10x productivity across the enterprise. Built for the future of work, it seamlessly transforms how teams collaborate, innovate and deliver value at scale."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.49,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.6900000000000001,
       "1": 0.21,
       "2": 0.08,
       "3": 0.01,
       "4": 0,
       "5": 0.01,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.8
     },
     "filler": {
      "type": "noul",
      "noul": 0.94
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "hand_waving": 0.01,
       "vanity_metric": 0.02,
       "buzzword_stacking": 0.97,
       "none": 0
      },
      "confidence": 0.96
     }
    },
    "usage": {
     "input_tokens": 560,
     "output_tokens": 91,
     "cost": 0.00002352
    },
    "id": "gen-dec-1790519759-LGzgtM6BPY6DcLpn6PdU",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "agentic-hollow",
   "arm": "llm",
   "repeat": 4,
   "latencyMs": 785,
   "costUsd": 0.00007365,
   "inputTokens": 371,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 3,
    "filler": 0.8,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "64ef554d6fdc0baa1ef1daea9ac52d0b1d15945fbb1b76fff2c146abc059389b",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur AI-native platform leverages cutting-edge agentic workflows to unlock 10x productivity across the enterprise. Built for the future of work, it seamlessly transforms how teams collaborate, innovate and deliver value at scale."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519759-KZ9fb3NBW7gu8XmLLt4t",
    "object": "chat.completion",
    "created": 1790519759,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":3,\"filler\":0.8,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 371,
     "completion_tokens": 30,
     "total_tokens": 401,
     "cost": 0.00007365,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007365,
      "upstream_inference_prompt_cost": 0.00005565,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "agentic-backed",
   "arm": "jev",
   "repeat": 4,
   "latencyMs": 169,
   "costUsd": 0.000024528,
   "inputTokens": 584,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 7.37,
    "substance_distribution": [
     0,
     0.01,
     0.02,
     0.06,
     0.09,
     0.1,
     0.15,
     0.23,
     0.24000000000000002,
     0.1
    ],
    "substance_confidence": 0.4,
    "filler": 0.16,
    "flaw": "none",
    "flaw_confidence": 0.87,
    "flaw_distribution": {
     "vanity_metric": 0.07,
     "none": 0.9,
     "buzzword_stacking": 0,
     "hand_waving": 0.03
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "83b552b87a442b248929456879f103598f01a46483f861b20f29a53f242c6593",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our agent drafts first-response replies for support tickets. In a six-week pilot with 14 agents at a 40-person SaaS company, median first-response time fell from 3h 10m to 41m; agents edited 62% of drafts before sending. It does not handle billing disputes."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 6.37,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0,
       "1": 0.01,
       "2": 0.02,
       "3": 0.06,
       "4": 0.09,
       "5": 0.1,
       "6": 0.15,
       "7": 0.23,
       "8": 0.24000000000000002,
       "9": 0.1
      },
      "confidence": 0.4
     },
     "filler": {
      "type": "noul",
      "noul": 0.16
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "vanity_metric": 0.07,
       "none": 0.9,
       "buzzword_stacking": 0,
       "hand_waving": 0.03
      },
      "confidence": 0.87
     }
    },
    "usage": {
     "input_tokens": 584,
     "output_tokens": 86,
     "cost": 0.000024528
    },
    "id": "gen-dec-1790519760-QPMV9di6kJVPfZps2a8j",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "agentic-backed",
   "arm": "llm",
   "repeat": 4,
   "latencyMs": 718,
   "costUsd": 0.00007485,
   "inputTokens": 391,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "e421cf5e0d015b409956f270fe540c5f0f3120a96ea0ad9e871b03f2fe35e00b",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur agent drafts first-response replies for support tickets. In a six-week pilot with 14 agents at a 40-person SaaS company, median first-response time fell from 3h 10m to 41m; agents edited 62% of drafts before sending. It does not handle billing disputes."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519760-lWxn8TdRnYUJ2UyJGMag",
    "object": "chat.completion",
    "created": 1790519760,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 391,
     "completion_tokens": 27,
     "total_tokens": 418,
     "cost": 0.00007485,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007485,
      "upstream_inference_prompt_cost": 0.00005865,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "security-hollow",
   "arm": "jev",
   "repeat": 4,
   "latencyMs": 146,
   "costUsd": 0.000023562,
   "inputTokens": 561,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.62,
    "substance_distribution": [
     0.59,
     0.29000000000000004,
     0.1,
     0.02,
     0,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.75,
    "filler": 0.94,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.3,
    "flaw_distribution": {
     "hand_waving": 0.22,
     "buzzword_stacking": 0.48,
     "none": 0,
     "vanity_metric": 0.3
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "d0ac5827a3b25d13348ed9777a2c32a48af33cb387c80ec52c5faf77838c61c0",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Introducing the world's most advanced AI-powered security platform. Our revolutionary engine stops 99.9% of threats before they happen, giving you total peace of mind in an ever-evolving threat landscape."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.62,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.59,
       "1": 0.29000000000000004,
       "2": 0.1,
       "3": 0.02,
       "4": 0,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.75
     },
     "filler": {
      "type": "noul",
      "noul": 0.94
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "hand_waving": 0.22,
       "buzzword_stacking": 0.48,
       "none": 0,
       "vanity_metric": 0.3
      },
      "confidence": 0.3
     }
    },
    "usage": {
     "input_tokens": 561,
     "output_tokens": 91,
     "cost": 0.000023562
    },
    "id": "gen-dec-1790519761-zLrTIGGCrC9xahV6sEey",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "security-hollow",
   "arm": "llm",
   "repeat": 4,
   "latencyMs": 1764,
   "costUsd": 0.0000732,
   "inputTokens": 368,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 3,
    "filler": 0.85,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "6f0e1fd6c68c1ad53fb9621a785302c4c2b039e3d2ab5e1bd13c0d59a95ff4e9",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nIntroducing the world's most advanced AI-powered security platform. Our revolutionary engine stops 99.9% of threats before they happen, giving you total peace of mind in an ever-evolving threat landscape."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519761-uKcrVSiKHLWk0gtAy79U",
    "object": "chat.completion",
    "created": 1790519761,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":3,\"filler\":0.85,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 368,
     "completion_tokens": 30,
     "total_tokens": 398,
     "cost": 0.0000732,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.0000732,
      "upstream_inference_prompt_cost": 0.0000552,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "security-backed",
   "arm": "jev",
   "repeat": 4,
   "latencyMs": 219,
   "costUsd": 0.00002457,
   "inputTokens": 585,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 6.43,
    "substance_distribution": [
     0.01,
     0.06,
     0.07,
     0.07,
     0.17,
     0.08,
     0.11,
     0.23,
     0.17,
     0.03
    ],
    "substance_confidence": 0.19,
    "filler": 0.17,
    "flaw": "none",
    "flaw_confidence": 0.33,
    "flaw_distribution": {
     "hand_waving": 0.02,
     "none": 0.5,
     "buzzword_stacking": 0,
     "vanity_metric": 0.48
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "b6d5c5a3cb6bc7b6d2eb0e36a4763d6e34df7c59aa9d502a06a81d3484a32abc",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our scanner flags leaked API keys in pull requests before merge. On 2,300 public repositories with known leaks from 2025, it caught 1,941 (84%) with 37 false positives. It misses keys split across files, and we list the known gaps on our docs page."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 5.43,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.01,
       "1": 0.06,
       "2": 0.07,
       "3": 0.07,
       "4": 0.17,
       "5": 0.08,
       "6": 0.11,
       "7": 0.23,
       "8": 0.17,
       "9": 0.03
      },
      "confidence": 0.19
     },
     "filler": {
      "type": "noul",
      "noul": 0.17
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "hand_waving": 0.02,
       "none": 0.5,
       "buzzword_stacking": 0,
       "vanity_metric": 0.48
      },
      "confidence": 0.33
     }
    },
    "usage": {
     "input_tokens": 585,
     "output_tokens": 86,
     "cost": 0.00002457
    },
    "id": "gen-dec-1790519763-xYu8IyQLT8iujHOrgg6Z",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "security-backed",
   "arm": "llm",
   "repeat": 4,
   "latencyMs": 1263,
   "costUsd": 0.00007455,
   "inputTokens": 389,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "e4eb040c67bee5b01347f6a7ea924fbb6a2daae61f9fb722d1e36dcc2d4e954c",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur scanner flags leaked API keys in pull requests before merge. On 2,300 public repositories with known leaks from 2025, it caught 1,941 (84%) with 37 false positives. It misses keys split across files, and we list the known gaps on our docs page."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519763-CHV5VOSB5HrBht4xlYoI",
    "object": "chat.completion",
    "created": 1790519763,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 389,
     "completion_tokens": 27,
     "total_tokens": 416,
     "cost": 0.00007455,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007455,
      "upstream_inference_prompt_cost": 0.00005835,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "funding-hollow",
   "arm": "jev",
   "repeat": 4,
   "latencyMs": 170,
   "costUsd": 0.000023562,
   "inputTokens": 561,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.91,
    "substance_distribution": [
     0.44,
     0.36,
     0.15,
     0.05,
     0,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.63,
    "filler": 0.91,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.93,
    "flaw_distribution": {
     "none": 0,
     "buzzword_stacking": 0.95,
     "vanity_metric": 0,
     "hand_waving": 0.05
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "1f0c300c7eac342f879e16082930f1201ca59048a112c54de9b00494b7442ea7",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "We're thrilled to announce our Series A to supercharge our mission of democratizing intelligence. With world-class investors behind us, we're poised to redefine the category and deliver unprecedented value to customers everywhere."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.91,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.44,
       "1": 0.36,
       "2": 0.15,
       "3": 0.05,
       "4": 0,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.63
     },
     "filler": {
      "type": "noul",
      "noul": 0.91
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "none": 0,
       "buzzword_stacking": 0.95,
       "vanity_metric": 0,
       "hand_waving": 0.05
      },
      "confidence": 0.93
     }
    },
    "usage": {
     "input_tokens": 561,
     "output_tokens": 91,
     "cost": 0.000023562
    },
    "id": "gen-dec-1790519764-C7McDs3cG7NbGybPy0z2",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "funding-hollow",
   "arm": "llm",
   "repeat": 4,
   "latencyMs": 679,
   "costUsd": 0.00007305,
   "inputTokens": 367,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 3,
    "filler": 0.8,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "64f587d56624bc07c8a5cd7bfb6ee309e01720f37e326702557db6c510013917",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nWe're thrilled to announce our Series A to supercharge our mission of democratizing intelligence. With world-class investors behind us, we're poised to redefine the category and deliver unprecedented value to customers everywhere."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519765-bjQTgnBMxuOMJJU7QHaX",
    "object": "chat.completion",
    "created": 1790519765,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":3,\"filler\":0.8,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 367,
     "completion_tokens": 30,
     "total_tokens": 397,
     "cost": 0.00007305,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007305,
      "upstream_inference_prompt_cost": 0.00005505,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "funding-backed",
   "arm": "jev",
   "repeat": 4,
   "latencyMs": 302,
   "costUsd": 0.000024402,
   "inputTokens": 581,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 7.32,
    "substance_distribution": [
     0,
     0.01,
     0.04,
     0.07,
     0.08,
     0.08,
     0.15,
     0.22,
     0.25,
     0.1
    ],
    "substance_confidence": 0.25,
    "filler": 0.19,
    "flaw": "none",
    "flaw_confidence": 0.55,
    "flaw_distribution": {
     "hand_waving": 0.06,
     "none": 0.66,
     "buzzword_stacking": 0.01,
     "vanity_metric": 0.27
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "ef7608ea4190ed19a1a27952905f6af21f0654930c19cf1c07f4be8aee59350b",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "We raised $12M to expand our invoice-matching product from 3 to 9 ERP integrations by June. Today 210 finance teams use it; the median team reconciles 4,800 invoices a month and clears exceptions in under a day, down from four."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 6.32,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0,
       "1": 0.01,
       "2": 0.04,
       "3": 0.07,
       "4": 0.08,
       "5": 0.08,
       "6": 0.15,
       "7": 0.22,
       "8": 0.25,
       "9": 0.1
      },
      "confidence": 0.25
     },
     "filler": {
      "type": "noul",
      "noul": 0.19
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "hand_waving": 0.06,
       "none": 0.66,
       "buzzword_stacking": 0.01,
       "vanity_metric": 0.27
      },
      "confidence": 0.55
     }
    },
    "usage": {
     "input_tokens": 581,
     "output_tokens": 86,
     "cost": 0.000024402
    },
    "id": "gen-dec-1790519765-AUeijwakGSWMIPIjr46P",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "funding-backed",
   "arm": "llm",
   "repeat": 4,
   "latencyMs": 1415,
   "costUsd": 0.0000738,
   "inputTokens": 384,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "fde93f61e1fca7e422f242f0cf572bdb00b531825d6d1a8003cc44db2554a01d",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nWe raised $12M to expand our invoice-matching product from 3 to 9 ERP integrations by June. Today 210 finance teams use it; the median team reconciles 4,800 invoices a month and clears exceptions in under a day, down from four."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519766-3piKoqwaiek2bN5xhBk0",
    "object": "chat.completion",
    "created": 1790519766,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 384,
     "completion_tokens": 27,
     "total_tokens": 411,
     "cost": 0.0000738,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.0000738,
      "upstream_inference_prompt_cost": 0.0000576,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "agentic-hollow",
   "arm": "jev",
   "repeat": 5,
   "latencyMs": 175,
   "costUsd": 0.00002352,
   "inputTokens": 560,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.52,
    "substance_distribution": [
     0.66,
     0.23,
     0.09,
     0.02,
     0,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.79,
    "filler": 0.93,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.95,
    "flaw_distribution": {
     "buzzword_stacking": 0.97,
     "hand_waving": 0.01,
     "none": 0,
     "vanity_metric": 0.02
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "63da952e1cda6cf63ae6a79969502ab6aa7c96cd6286cbd59ca141dc016be198",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our AI-native platform leverages cutting-edge agentic workflows to unlock 10x productivity across the enterprise. Built for the future of work, it seamlessly transforms how teams collaborate, innovate and deliver value at scale."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.52,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.66,
       "1": 0.23,
       "2": 0.09,
       "3": 0.02,
       "4": 0,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.79
     },
     "filler": {
      "type": "noul",
      "noul": 0.93
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "buzzword_stacking": 0.97,
       "hand_waving": 0.01,
       "none": 0,
       "vanity_metric": 0.02
      },
      "confidence": 0.95
     }
    },
    "usage": {
     "input_tokens": 560,
     "output_tokens": 91,
     "cost": 0.00002352
    },
    "id": "gen-dec-1790519767-47wAsPvoJmHDVwdK4RDT",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "agentic-hollow",
   "arm": "llm",
   "repeat": 5,
   "latencyMs": 723,
   "costUsd": 0.00007365,
   "inputTokens": 371,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 3,
    "filler": 0.85,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "64ef554d6fdc0baa1ef1daea9ac52d0b1d15945fbb1b76fff2c146abc059389b",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur AI-native platform leverages cutting-edge agentic workflows to unlock 10x productivity across the enterprise. Built for the future of work, it seamlessly transforms how teams collaborate, innovate and deliver value at scale."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519767-mRJz2qic5Oj9cvQJQzrn",
    "object": "chat.completion",
    "created": 1790519767,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":3,\"filler\":0.85,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 371,
     "completion_tokens": 30,
     "total_tokens": 401,
     "cost": 0.00007365,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007365,
      "upstream_inference_prompt_cost": 0.00005565,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "agentic-backed",
   "arm": "jev",
   "repeat": 5,
   "latencyMs": 174,
   "costUsd": 0.000024528,
   "inputTokens": 584,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 7.45,
    "substance_distribution": [
     0,
     0.01,
     0.02,
     0.06,
     0.09,
     0.09,
     0.15,
     0.23,
     0.23,
     0.12
    ],
    "substance_confidence": 0.28,
    "filler": 0.15,
    "flaw": "none",
    "flaw_confidence": 0.87,
    "flaw_distribution": {
     "none": 0.9,
     "hand_waving": 0.03,
     "buzzword_stacking": 0,
     "vanity_metric": 0.07
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "83b552b87a442b248929456879f103598f01a46483f861b20f29a53f242c6593",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our agent drafts first-response replies for support tickets. In a six-week pilot with 14 agents at a 40-person SaaS company, median first-response time fell from 3h 10m to 41m; agents edited 62% of drafts before sending. It does not handle billing disputes."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 6.45,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0,
       "1": 0.01,
       "2": 0.02,
       "3": 0.06,
       "4": 0.09,
       "5": 0.09,
       "6": 0.15,
       "7": 0.23,
       "8": 0.23,
       "9": 0.12
      },
      "confidence": 0.28
     },
     "filler": {
      "type": "noul",
      "noul": 0.15
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "none": 0.9,
       "hand_waving": 0.03,
       "buzzword_stacking": 0,
       "vanity_metric": 0.07
      },
      "confidence": 0.87
     }
    },
    "usage": {
     "input_tokens": 584,
     "output_tokens": 86,
     "cost": 0.000024528
    },
    "id": "gen-dec-1790519768-Pssbxc6V8WEGnkW2omrK",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "agentic-backed",
   "arm": "llm",
   "repeat": 5,
   "latencyMs": 1360,
   "costUsd": 0.00007485,
   "inputTokens": 391,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "e421cf5e0d015b409956f270fe540c5f0f3120a96ea0ad9e871b03f2fe35e00b",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur agent drafts first-response replies for support tickets. In a six-week pilot with 14 agents at a 40-person SaaS company, median first-response time fell from 3h 10m to 41m; agents edited 62% of drafts before sending. It does not handle billing disputes."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519768-5gWODd621OHCnp8GEl4E",
    "object": "chat.completion",
    "created": 1790519768,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 391,
     "completion_tokens": 27,
     "total_tokens": 418,
     "cost": 0.00007485,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007485,
      "upstream_inference_prompt_cost": 0.00005865,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "security-hollow",
   "arm": "jev",
   "repeat": 5,
   "latencyMs": 180,
   "costUsd": 0.000023562,
   "inputTokens": 561,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.62,
    "substance_distribution": [
     0.6,
     0.27,
     0.1,
     0.02,
     0,
     0.01,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.75,
    "filler": 0.93,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.19,
    "flaw_distribution": {
     "none": 0,
     "buzzword_stacking": 0.39,
     "hand_waving": 0.27,
     "vanity_metric": 0.34
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "d0ac5827a3b25d13348ed9777a2c32a48af33cb387c80ec52c5faf77838c61c0",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Introducing the world's most advanced AI-powered security platform. Our revolutionary engine stops 99.9% of threats before they happen, giving you total peace of mind in an ever-evolving threat landscape."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.62,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.6,
       "1": 0.27,
       "2": 0.1,
       "3": 0.02,
       "4": 0,
       "5": 0.01,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.75
     },
     "filler": {
      "type": "noul",
      "noul": 0.93
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "none": 0,
       "buzzword_stacking": 0.39,
       "hand_waving": 0.27,
       "vanity_metric": 0.34
      },
      "confidence": 0.19
     }
    },
    "usage": {
     "input_tokens": 561,
     "output_tokens": 91,
     "cost": 0.000023562
    },
    "id": "gen-dec-1790519769-EnGveCpJuniCeFrsN9O4",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "security-hollow",
   "arm": "llm",
   "repeat": 5,
   "latencyMs": 1492,
   "costUsd": 0.0000732,
   "inputTokens": 368,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 2,
    "filler": 0.9,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.85
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "6f0e1fd6c68c1ad53fb9621a785302c4c2b039e3d2ab5e1bd13c0d59a95ff4e9",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nIntroducing the world's most advanced AI-powered security platform. Our revolutionary engine stops 99.9% of threats before they happen, giving you total peace of mind in an ever-evolving threat landscape."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519770-8RPPn3zyTdD3VCDuiA1S",
    "object": "chat.completion",
    "created": 1790519770,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":2,\"filler\":0.9,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.85}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 368,
     "completion_tokens": 30,
     "total_tokens": 398,
     "cost": 0.0000732,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.0000732,
      "upstream_inference_prompt_cost": 0.0000552,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "security-backed",
   "arm": "jev",
   "repeat": 5,
   "latencyMs": 188,
   "costUsd": 0.00002457,
   "inputTokens": 585,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 6.52,
    "substance_distribution": [
     0.01,
     0.05,
     0.06,
     0.06,
     0.16,
     0.08,
     0.18,
     0.22999999999999998,
     0.14,
     0.03
    ],
    "substance_confidence": 0.25,
    "filler": 0.18,
    "flaw": "none",
    "flaw_confidence": 0.34,
    "flaw_distribution": {
     "vanity_metric": 0.47,
     "none": 0.51,
     "buzzword_stacking": 0,
     "hand_waving": 0.02
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "b6d5c5a3cb6bc7b6d2eb0e36a4763d6e34df7c59aa9d502a06a81d3484a32abc",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "Our scanner flags leaked API keys in pull requests before merge. On 2,300 public repositories with known leaks from 2025, it caught 1,941 (84%) with 37 false positives. It misses keys split across files, and we list the known gaps on our docs page."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 5.52,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.01,
       "1": 0.05,
       "2": 0.06,
       "3": 0.06,
       "4": 0.16,
       "5": 0.08,
       "6": 0.18,
       "7": 0.22999999999999998,
       "8": 0.14,
       "9": 0.03
      },
      "confidence": 0.25
     },
     "filler": {
      "type": "noul",
      "noul": 0.18
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "vanity_metric": 0.47,
       "none": 0.51,
       "buzzword_stacking": 0,
       "hand_waving": 0.02
      },
      "confidence": 0.34
     }
    },
    "usage": {
     "input_tokens": 585,
     "output_tokens": 86,
     "cost": 0.00002457
    },
    "id": "gen-dec-1790519771-T09RB4tVB4jCPz8KFygX",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "security-backed",
   "arm": "llm",
   "repeat": 5,
   "latencyMs": 1268,
   "costUsd": 0.00007455,
   "inputTokens": 389,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "e4eb040c67bee5b01347f6a7ea924fbb6a2daae61f9fb722d1e36dcc2d4e954c",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nOur scanner flags leaked API keys in pull requests before merge. On 2,300 public repositories with known leaks from 2025, it caught 1,941 (84%) with 37 false positives. It misses keys split across files, and we list the known gaps on our docs page."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519771-YwaYQ7Sl3dQ0TnS8I36y",
    "object": "chat.completion",
    "created": 1790519771,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 389,
     "completion_tokens": 27,
     "total_tokens": 416,
     "cost": 0.00007455,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007455,
      "upstream_inference_prompt_cost": 0.00005835,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "funding-hollow",
   "arm": "jev",
   "repeat": 5,
   "latencyMs": 199,
   "costUsd": 0.000023562,
   "inputTokens": 561,
   "outputTokens": 91,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 1.91,
    "substance_distribution": [
     0.42,
     0.38,
     0.15,
     0.05,
     0,
     0,
     0,
     0,
     0,
     0
    ],
    "substance_confidence": 0.64,
    "filler": 0.92,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.91,
    "flaw_distribution": {
     "hand_waving": 0.07,
     "none": 0,
     "buzzword_stacking": 0.93,
     "vanity_metric": 0
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "1f0c300c7eac342f879e16082930f1201ca59048a112c54de9b00494b7442ea7",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "We're thrilled to announce our Series A to supercharge our mission of democratizing intelligence. With world-class investors behind us, we're poised to redefine the category and deliver unprecedented value to customers everywhere."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 0.91,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0.42,
       "1": 0.38,
       "2": 0.15,
       "3": 0.05,
       "4": 0,
       "5": 0,
       "6": 0,
       "7": 0,
       "8": 0,
       "9": 0
      },
      "confidence": 0.64
     },
     "filler": {
      "type": "noul",
      "noul": 0.92
     },
     "flaw": {
      "type": "choice",
      "choice": "buzzword_stacking",
      "probabilities": {
       "hand_waving": 0.07,
       "none": 0,
       "buzzword_stacking": 0.93,
       "vanity_metric": 0
      },
      "confidence": 0.91
     }
    },
    "usage": {
     "input_tokens": 561,
     "output_tokens": 91,
     "cost": 0.000023562
    },
    "id": "gen-dec-1790519773-zHypClymg7EOoXojGmN5",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "funding-hollow",
   "arm": "llm",
   "repeat": 5,
   "latencyMs": 1341,
   "costUsd": 0.00007305,
   "inputTokens": 367,
   "outputTokens": 30,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "Azure",
   "answers": {
    "substance": 2,
    "filler": 0.9,
    "flaw": "buzzword_stacking",
    "flaw_confidence": 0.8
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "64f587d56624bc07c8a5cd7bfb6ee309e01720f37e326702557db6c510013917",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nWe're thrilled to announce our Series A to supercharge our mission of democratizing intelligence. With world-class investors behind us, we're poised to redefine the category and deliver unprecedented value to customers everywhere."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519773-OumbC9Ge1tTl5V3ICEHA",
    "object": "chat.completion",
    "created": 1790519773,
    "model": "openai/gpt-4o-mini",
    "provider": "Azure",
    "system_fingerprint": "fp_99f88af1c5",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":2,\"filler\":0.9,\"flaw\":\"buzzword_stacking\",\"flaw_confidence\":0.8}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 367,
     "completion_tokens": 30,
     "total_tokens": 397,
     "cost": 0.00007305,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.00007305,
      "upstream_inference_prompt_cost": 0.00005505,
      "upstream_inference_completions_cost": 0.000018
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  },
  {
   "input": "funding-backed",
   "arm": "jev",
   "repeat": 5,
   "latencyMs": 164,
   "costUsd": 0.000024402,
   "inputTokens": 581,
   "outputTokens": 86,
   "servedModel": "typesafe/jev-1.13-20260917",
   "provider": "TypeSafe",
   "answers": {
    "substance": 7.23,
    "substance_distribution": [
     0,
     0.01,
     0.05,
     0.08,
     0.09,
     0.08,
     0.14,
     0.23,
     0.24,
     0.08
    ],
    "substance_confidence": 0.23,
    "filler": 0.19,
    "flaw": "none",
    "flaw_confidence": 0.49,
    "flaw_distribution": {
     "hand_waving": 0.05,
     "buzzword_stacking": 0.01,
     "none": 0.62,
     "vanity_metric": 0.32
    }
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "ef7608ea4190ed19a1a27952905f6af21f0654930c19cf1c07f4be8aee59350b",
   "request": {
    "model": "typesafe/jev-1.13",
    "state": {
     "copy": "We raised $12M to expand our invoice-matching product from 3 to 9 ERP integrations by June. Today 210 finance teams use it; the median team reconciles 4,800 invoices a month and clears exceptions in under a day, down from four."
    },
    "questions": {
     "substance": {
      "type": "score",
      "instructions": "How much verifiable substance does this copy contain?",
      "criteria": [
       "1",
       "2",
       "3",
       "4",
       "5",
       "6",
       "7",
       "8",
       "9",
       "10"
      ]
     },
     "filler": {
      "type": "noul",
      "instructions": "Is this copy mostly unbacked filler?",
      "criteria": {
       "true": "Claims with no evidence, numbers, mechanism or source.",
       "false": "Claims are backed by specifics a reader could check."
      }
     },
     "flaw": {
      "type": "choice",
      "instructions": "What is the main rhetorical flaw?",
      "criteria": {
       "hand_waving": "Vague claims with no mechanism.",
       "buzzword_stacking": "Strings of jargon standing in for meaning.",
       "vanity_metric": "A number with no baseline or source.",
       "none": "No major flaw; claims are specific and checkable."
      }
     }
    }
   },
   "response": {
    "model": "typesafe/jev-1.13-20260917",
    "answers": {
     "substance": {
      "type": "score",
      "score": 6.23,
      "legend": {
       "0": "1",
       "1": "2",
       "2": "3",
       "3": "4",
       "4": "5",
       "5": "6",
       "6": "7",
       "7": "8",
       "8": "9",
       "9": "10"
      },
      "probabilities": {
       "0": 0,
       "1": 0.01,
       "2": 0.05,
       "3": 0.08,
       "4": 0.09,
       "5": 0.08,
       "6": 0.14,
       "7": 0.23,
       "8": 0.24,
       "9": 0.08
      },
      "confidence": 0.23
     },
     "filler": {
      "type": "noul",
      "noul": 0.19
     },
     "flaw": {
      "type": "choice",
      "choice": "none",
      "probabilities": {
       "hand_waving": 0.05,
       "buzzword_stacking": 0.01,
       "none": 0.62,
       "vanity_metric": 0.32
      },
      "confidence": 0.49
     }
    },
    "usage": {
     "input_tokens": 581,
     "output_tokens": 86,
     "cost": 0.000024402
    },
    "id": "gen-dec-1790519774-lJH4rZSsKMJWyEOTXwQp",
    "provider": "TypeSafe"
   }
  },
  {
   "input": "funding-backed",
   "arm": "llm",
   "repeat": 5,
   "latencyMs": 685,
   "costUsd": 0.0000738,
   "inputTokens": 384,
   "outputTokens": 27,
   "servedModel": "openai/gpt-4o-mini",
   "provider": "OpenAI",
   "answers": {
    "substance": 8,
    "filler": 0.1,
    "flaw": "none",
    "flaw_confidence": 0.9
   },
   "parseError": null,
   "schemaErrors": [],
   "requestSha256": "fde93f61e1fca7e422f242f0cf572bdb00b531825d6d1a8003cc44db2554a01d",
   "request": {
    "model": "openai/gpt-4o-mini",
    "messages": [
     {
      "role": "user",
      "content": "Evaluate the marketing copy below. Reply with only a JSON object with these keys:\n- \"substance\": integer 1-10. How much verifiable substance does this copy contain?\n- \"filler\": probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler? (yes = Claims with no evidence, numbers, mechanism or source.)\n- \"flaw\": one of \"hand_waving\" (Vague claims with no mechanism.), \"buzzword_stacking\" (Strings of jargon standing in for meaning.), \"vanity_metric\" (A number with no baseline or source.), \"none\" (No major flaw; claims are specific and checkable.). What is the main rhetorical flaw?\n- \"flaw_confidence\": probability 0.0-1.0 that your choice is right.\n\nCopy:\nWe raised $12M to expand our invoice-matching product from 3 to 9 ERP integrations by June. Today 210 finance teams use it; the median team reconciles 4,800 invoices a month and clears exceptions in under a day, down from four."
     }
    ],
    "usage": {
     "include": true
    },
    "provider": {
     "require_parameters": true
    },
    "response_format": {
     "type": "json_schema",
     "json_schema": {
      "name": "bs-detector",
      "strict": true,
      "schema": {
       "type": "object",
       "properties": {
        "substance": {
         "type": "integer",
         "description": "How much verifiable substance does this copy contain? 1-10."
        },
        "filler": {
         "type": "number",
         "description": "Probability 0.0-1.0 that the answer is yes. Is this copy mostly unbacked filler?"
        },
        "flaw": {
         "type": "string",
         "enum": [
          "hand_waving",
          "buzzword_stacking",
          "vanity_metric",
          "none"
         ],
         "description": "What is the main rhetorical flaw?"
        },
        "flaw_confidence": {
         "type": "number",
         "description": "Probability 0.0-1.0 that your choice is right."
        }
       },
       "required": [
        "substance",
        "filler",
        "flaw",
        "flaw_confidence"
       ],
       "additionalProperties": false
      }
     }
    }
   },
   "response": {
    "id": "gen-1790519774-zm5gv98cCFCoFlr06cOX",
    "object": "chat.completion",
    "created": 1790519774,
    "model": "openai/gpt-4o-mini",
    "provider": "OpenAI",
    "system_fingerprint": "fp_10a9fd5a82",
    "service_tier": "default",
    "choices": [
     {
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop",
      "native_finish_reason": "stop",
      "message": {
       "role": "assistant",
       "content": "{\"substance\":8,\"filler\":0.1,\"flaw\":\"none\",\"flaw_confidence\":0.9}",
       "refusal": null,
       "reasoning": null
      }
     }
    ],
    "usage": {
     "prompt_tokens": 384,
     "completion_tokens": 27,
     "total_tokens": 411,
     "cost": 0.0000738,
     "is_byok": false,
     "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "audio_tokens": 0,
      "video_tokens": 0
     },
     "cost_details": {
      "upstream_inference_cost": 0.0000738,
      "upstream_inference_prompt_cost": 0.0000576,
      "upstream_inference_completions_cost": 0.0000162
     },
     "completion_tokens_details": {
      "reasoning_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0
     }
    }
   }
  }
 ]
}