{
  "snapshot": "2026-06-19",
  "method": "Fleiss kappa on the displayed grid: subjects = 16 categories, raters = 8 models, labels = distinct canonicalized top tools. P_i = (sum_j n_ij^2 - n)/(n(n-1)); Pbar = mean P_i (equals mean pairwise agreement); Pe (global marginal) = sum_j p_j^2 with p_j = total assignments to tool j / (16*8); kappa = (Pbar - Pe)/(1 - Pe).",
  "primary_displayedGrid": {
    "nModels": 8,
    "nCategories": 16,
    "meanPairwiseAgreement": 0.4397,
    "fleissKappa_globalMarginalPe": 0.4101,
    "Pe_globalMarginal": 0.0503,
    "fleissKappa_perCategoryMarginalPe": -0.1429,
    "Pe_perCategoryMarginal": 0.5098,
    "unanimousCategories": 0,
    "interpretation": "kappa ~0.41 is MODERATE agreement (Landis-Koch), not chaos: near-consensus on mature categories, collapse on emerging ones.",
    "matureNearConsensus": [
      "AI copywriting",
      "CRM",
      "Customer support",
      "Scheduling",
      "SEO tools"
    ],
    "emergingFractured": [
      "Email marketing",
      "GEO / AI visibility",
      "Sales engagement"
    ]
  },
  "perCategoryConsensus": [
    {
      "category": "AI copywriting",
      "consensus": 0.75,
      "distinctTop": 2,
      "topCounts": {
        "WordLift": 1,
        "Jasper": 7
      }
    },
    {
      "category": "CRM",
      "consensus": 0.75,
      "distinctTop": 2,
      "topCounts": {
        "HubSpot": 7,
        "Salesforce": 1
      }
    },
    {
      "category": "Customer support",
      "consensus": 0.75,
      "distinctTop": 2,
      "topCounts": {
        "Freshdesk": 1,
        "Zendesk": 7
      }
    },
    {
      "category": "Scheduling",
      "consensus": 0.75,
      "distinctTop": 2,
      "topCounts": {
        "Calendly": 7,
        "Celoxis": 1
      }
    },
    {
      "category": "SEO tools",
      "consensus": 0.5714,
      "distinctTop": 2,
      "topCounts": {
        "Ahrefs": 6,
        "SEMrush": 2
      }
    },
    {
      "category": "Social media management",
      "consensus": 0.4643,
      "distinctTop": 2,
      "topCounts": {
        "Hootsuite": 3,
        "Buffer": 5
      }
    },
    {
      "category": "Product analytics",
      "consensus": 0.3929,
      "distinctTop": 3,
      "topCounts": {
        "Google Analytics": 1,
        "Amplitude": 5,
        "Mixpanel": 2
      }
    },
    {
      "category": "Landing page builders",
      "consensus": 0.3929,
      "distinctTop": 3,
      "topCounts": {
        "Unbounce": 5,
        "Landingi": 1,
        "Webflow": 2
      }
    },
    {
      "category": "Project management",
      "consensus": 0.3214,
      "distinctTop": 3,
      "topCounts": {
        "Asana": 4,
        "Trello": 1,
        "ClickUp": 3
      }
    },
    {
      "category": "Lead generation",
      "consensus": 0.3214,
      "distinctTop": 3,
      "topCounts": {
        "HubSpot": 3,
        "SyncGTM": 1,
        "Apollo.io": 4
      }
    },
    {
      "category": "AI chatbots",
      "consensus": 0.3214,
      "distinctTop": 3,
      "topCounts": {
        "Dialogflow": 3,
        "ChatGPT": 4,
        "Chatbase": 1
      }
    },
    {
      "category": "Survey / forms",
      "consensus": 0.3214,
      "distinctTop": 3,
      "topCounts": {
        "SurveyMonkey": 4,
        "Typeform": 3,
        "SurveyNinja": 1
      }
    },
    {
      "category": "Marketing automation",
      "consensus": 0.2857,
      "distinctTop": 3,
      "topCounts": {
        "Marketo": 2,
        "HubSpot": 4,
        "ActiveCampaign": 2
      }
    },
    {
      "category": "Email marketing",
      "consensus": 0.25,
      "distinctTop": 4,
      "topCounts": {
        "Mailchimp": 4,
        "Klaviyo": 1,
        "HubSpot": 2,
        "ActiveCampaign": 1
      }
    },
    {
      "category": "GEO / AI visibility",
      "consensus": 0.2143,
      "distinctTop": 5,
      "topCounts": {
        "Ahrefs": 4,
        "Siftly": 1,
        "Profound": 1,
        "Surfer": 1,
        "None": 1
      }
    },
    {
      "category": "Sales engagement",
      "consensus": 0.1786,
      "distinctTop": 4,
      "topCounts": {
        "HubSpot": 2,
        "Outreach": 3,
        "Amplemarket": 2,
        "Apollo.io": 1
      }
    }
  ],
  "deterministicRule": {
    "rule": "Top pick per model per category = MODAL (majority) top tool across the model's 3 phrasings; ties broken by the first phrasing (collect.js order). Grounded models return one answer, so their modal is that single answer.",
    "note": "Governs the grid from the next snapshot forward. The Jun-19 grid retains one representative run per memory cell. Three memory models (Llama 4, GPT-OSS, Qwen) were re-captured 2026-07-08 for the next snapshot, so their Jun-19 three-phrasing raw is not recoverable; modal comparisons below for those models reflect the 2026-07-08 raw.",
    "modalWouldDifferAt": [
      {
        "category": "GEO / AI visibility",
        "engine": "GPT-OSS",
        "displayed": "Surfer",
        "modal": "BrightEdge",
        "rawSnapshot": "2026-07-08"
      },
      {
        "category": "GEO / AI visibility",
        "engine": "Qwen",
        "displayed": "None",
        "modal": "Ahrefs",
        "rawSnapshot": "2026-07-08"
      },
      {
        "category": "Email marketing",
        "engine": "Cohere",
        "displayed": "Mailchimp",
        "modal": "Klaviyo",
        "rawSnapshot": "2026-06-19"
      },
      {
        "category": "Email marketing",
        "engine": "GPT-OSS",
        "displayed": "Mailchimp",
        "modal": "Klaviyo",
        "rawSnapshot": "2026-07-08"
      },
      {
        "category": "Product analytics",
        "engine": "Llama 3.3",
        "displayed": "Google Analytics",
        "modal": "Amplitude",
        "rawSnapshot": "2026-06-19"
      },
      {
        "category": "Project management",
        "engine": "Cohere",
        "displayed": "Trello",
        "modal": "Asana",
        "rawSnapshot": "2026-06-19"
      },
      {
        "category": "Survey / forms",
        "engine": "Cohere",
        "displayed": "Typeform",
        "modal": "SurveyMonkey",
        "rawSnapshot": "2026-06-19"
      },
      {
        "category": "Survey / forms",
        "engine": "GPT-OSS",
        "displayed": "Typeform",
        "modal": "Qualtrics",
        "rawSnapshot": "2026-07-08"
      },
      {
        "category": "Survey / forms",
        "engine": "Qwen",
        "displayed": "Typeform",
        "modal": "SurveyMonkey",
        "rawSnapshot": "2026-07-08"
      },
      {
        "category": "AI copywriting",
        "engine": "Llama 3.3",
        "displayed": "WordLift",
        "modal": "Jasper",
        "rawSnapshot": "2026-06-19"
      },
      {
        "category": "CRM",
        "engine": "Llama 4",
        "displayed": "HubSpot",
        "modal": "Salesforce",
        "rawSnapshot": "2026-07-08"
      },
      {
        "category": "CRM",
        "engine": "GPT-OSS",
        "displayed": "HubSpot",
        "modal": "Salesforce",
        "rawSnapshot": "2026-07-08"
      },
      {
        "category": "CRM",
        "engine": "Qwen",
        "displayed": "HubSpot",
        "modal": "Salesforce",
        "rawSnapshot": "2026-07-08"
      },
      {
        "category": "Social media management",
        "engine": "GPT-OSS",
        "displayed": "Buffer",
        "modal": "Hootsuite",
        "rawSnapshot": "2026-07-08"
      }
    ]
  },
  "intraModelConsistency": {
    "definition": "share of (memory model, category) cells whose 3 phrasings returned the SAME canonicalized top tool",
    "rate": 0.2625,
    "same": 21,
    "of": 80,
    "perModel": {
      "Llama 3.3": {
        "same": 8,
        "of": 16,
        "snapshotDate": "2026-06-19"
      },
      "Cohere": {
        "same": 4,
        "of": 16,
        "snapshotDate": "2026-06-19"
      },
      "Llama 4": {
        "same": 4,
        "of": 16,
        "snapshotDate": "2026-07-08"
      },
      "GPT-OSS": {
        "same": 2,
        "of": 16,
        "snapshotDate": "2026-07-08"
      },
      "Qwen": {
        "same": 3,
        "of": 16,
        "snapshotDate": "2026-07-08"
      }
    },
    "reading": "models return the same #1 across their own three rephrasings only about 26% of the time; they frequently disagree with themselves."
  }
}