{
  "schema_version": 1,
  "generated_on": "2026-09-27",
  "scope": "93 canonical Epoch AI frontier-AI reference sites",
  "workflow_order": [
    "DISCOVER",
    "VERIFY_SCOPE",
    "CAPTURE",
    "CORROBORATE",
    "CLASSIFY",
    "INGEST"
  ],
  "stages": {
    "DISCOVER": "Run the task's site-specific search links and the named source families.",
    "VERIFY_SCOPE": "Confirm the source actually identifies the site, campus, parcel, project, utility case or physical observation.",
    "CAPTURE": "Record source URL, document/filing/permit/scene locator, publication or record date when exposed, access date, and the exact supported value/observation.",
    "CORROBORATE": "For material claims, seek a distinct source origin; mirrored or syndicated copies are not independent corroboration.",
    "CLASSIFY": "Use an explicit disposition: FOUND, PUBLIC_LEAD_REVIEW_REQUIRED, NO_PUBLIC_RECORD, SOURCE_UNAVAILABLE, DUPLICATE_OF_EXISTING_SOURCE, or CONFLICT_NEEDS_REVIEW.",
    "INGEST": "Submit through the research submission schema; canonical builders remain the state gate."
  },
  "publisher_workflows": {
    "project": {
      "label": "Project / program identity",
      "objective": "Identify the public project/program name corresponding to the site without collapsing a site into a broader corporate initiative.",
      "search": [
        "Official owner/developer project pages",
        "planning and permit documents",
        "utility/interconnection filings",
        "local government development records",
        "reputable local reporting"
      ],
      "queries": [
        "\"{site}\" project data center",
        "\"{site}\" project program {state}",
        "\"{site}\" site:sec.gov data center"
      ],
      "capture": [
        "exact public name",
        "site/project linkage",
        "publication date",
        "source URL",
        "source locator"
      ],
      "accept": "Direct site/campus/project linkage.",
      "reject": "Company-wide initiative with no site linkage."
    },
    "investors": {
      "label": "Investors / financing parties",
      "objective": "Identify disclosed project-specific financing or investment parties; do not infer them from ownership, tenancy, or ordinary shareholders.",
      "search": [
        "SEC 10-K/10-Q/8-K and exhibits",
        "bond and development-agreement records",
        "project financing announcements",
        "municipal incentive records",
        "official company releases"
      ],
      "queries": [
        "\"{site}\" financing investor",
        "\"{site}\" development agreement financing",
        "\"{site}\" investor data center"
      ],
      "capture": [
        "party name",
        "relationship type",
        "document/filing",
        "site/project linkage",
        "date",
        "URL/locator"
      ],
      "accept": "Source explicitly connects the party to financing/investment in the project.",
      "reject": "Market shareholders or unrelated financing with no site/project linkage."
    },
    "construction_companies": {
      "label": "Construction companies / EPC / GC",
      "objective": "Identify site-specific general contractors, EPCs, builders, civil contractors or major construction partners.",
      "search": [
        "Building permits and permit applicants",
        "planning/zoning packets",
        "county/city project records",
        "contractor project portfolios",
        "owner/developer releases",
        "procurement/bid notices"
      ],
      "queries": [
        "\"{site}\" contractor construction",
        "\"{site}\" EPC general contractor",
        "\"{site}\" permit contractor {state}"
      ],
      "capture": [
        "company",
        "role",
        "site linkage",
        "permit/contract/project reference",
        "date",
        "URL/locator"
      ],
      "accept": "The source names the contractor/EPC and ties it to the site/project.",
      "reject": "Generic contractor work elsewhere for the same owner."
    },
    "energy_companies": {
      "label": "Energy companies / utilities",
      "objective": "Identify the site-specific serving utility, power provider, energy partner or disclosed direct-supply counterparty.",
      "search": [
        "Utility service filings",
        "state/provincial regulator dockets",
        "utility large-load/interconnection records",
        "utility press releases",
        "site development filings"
      ],
      "queries": [
        "\"{site}\" utility electricity",
        "\"{site}\" power supply utility {state}",
        "\"{site}\" interconnection {state}"
      ],
      "capture": [
        "company",
        "relationship type",
        "site linkage",
        "case/docket/project ID",
        "date",
        "URL/locator"
      ],
      "accept": "Source explicitly connects the company to serving, interconnection, supply, or a site-specific agreement.",
      "reject": "Regional utility presence with no site linkage."
    },
    "owner": {
      "label": "Owner / legal project entity",
      "objective": "Resolve the current site owner/operator entity at the appropriate legal-entity level.",
      "search": [
        "County assessor/property records",
        "permit applicant/owner",
        "SEC subsidiary disclosures",
        "state corporate registries/GLEIF",
        "official owner pages"
      ],
      "queries": [
        "\"{site}\" owner property",
        "\"{site}\" LLC owner",
        "\"{site}\" parcel owner {state}"
      ],
      "capture": [
        "legal entity",
        "role",
        "parcel/address if public",
        "record date",
        "URL/locator"
      ],
      "accept": "Primary or official public record identifies the entity at the site.",
      "reject": "Parent company alone when a site-specific legal entity is disclosed."
    },
    "users": {
      "label": "Users / tenants / compute counterparties",
      "objective": "Identify disclosed site users or tenants without turning general customers or partnerships into site occupancy.",
      "search": [
        "Owner/operator releases",
        "tenant/customer announcements",
        "lease/development records",
        "company infrastructure disclosures",
        "reputable local reporting"
      ],
      "queries": [
        "\"{site}\" tenant",
        "\"{site}\" customer compute",
        "\"{site}\" Anthropic OR OpenAI OR Google OR Microsoft OR Meta OR Amazon"
      ],
      "capture": [
        "user/tenant",
        "relationship",
        "site linkage",
        "date",
        "URL/locator"
      ],
      "accept": "Source explicitly ties the user/tenant/customer relationship to the site/campus.",
      "reject": "Company-wide customer relationship with no site-specific connection."
    },
    "address": {
      "label": "Site address",
      "objective": "Resolve the most defensible public physical address or site location string.",
      "search": [
        "Official project documents",
        "permits",
        "planning packets",
        "property records",
        "utility/site filings",
        "operator/developer pages"
      ],
      "queries": [
        "\"{site}\" address data center",
        "\"{site}\" permit address {state}",
        "\"{site}\" parcel"
      ],
      "capture": [
        "address",
        "record type",
        "site linkage",
        "date",
        "URL/locator"
      ],
      "accept": "Direct official/primary address or two independent public sources.",
      "reject": "Geocoding guesses or nearby mailing addresses without site linkage."
    },
    "chip_quantities": {
      "label": "Chip / accelerator quantities",
      "objective": "Capture site-specific accelerator counts or chip-type quantities only when explicitly attributed to the site.",
      "search": [
        "Epoch site-level records",
        "owner/developer disclosures",
        "chip/supplier announcements",
        "procurement or planning documents",
        "reputable infrastructure reporting"
      ],
      "queries": [
        "\"{site}\" GPU chips",
        "\"{site}\" H100 H200 GPU",
        "\"{site}\" accelerator chips"
      ],
      "capture": [
        "chip type",
        "quantity",
        "date/period",
        "source wording",
        "URL/locator"
      ],
      "accept": "Site-specific quantity is explicitly published or preserved in a canonical site-level source.",
      "reject": "Global shipment/ownership estimates without site allocation."
    }
  },
  "domain_workflows": {
    "service_or_contract": {
      "label": "Service / energy contract",
      "objective": "Find site-specific utility service, capacity reservation, power purchase, direct supply, or contractual energy evidence.",
      "search": [
        "Serving utility tariff/service filings",
        "regulator dockets",
        "utility large-load proceedings",
        "municipal/development agreements",
        "company disclosures"
      ],
      "queries": [
        "\"{site}\" service agreement utility",
        "\"{site}\" power contract",
        "\"{site}\" load service agreement {state}"
      ],
      "capture": [
        "counterparties",
        "contract/service type",
        "capacity if stated",
        "case/docket/document ID",
        "date",
        "URL/locator"
      ],
      "accept": "Document identifies the site and contractual/service relationship.",
      "reject": "Generic tariff availability or regional service."
    },
    "regulatory": {
      "label": "Regulatory / permitting",
      "objective": "Collect site-specific planning, zoning, building, environmental, water, air, tax-incentive, or utility proceedings.",
      "search": [
        "Municipal planning/building portals",
        "county records",
        "state/provincial regulator databases",
        "environmental agency records",
        "tax-incentive/public development records"
      ],
      "queries": [
        "\"{site}\" permit planning zoning",
        "\"{site}\" environmental permit {state}",
        "\"{site}\" public hearing data center"
      ],
      "capture": [
        "agency",
        "record type",
        "case/permit number",
        "status",
        "date",
        "site linkage",
        "URL/locator"
      ],
      "accept": "Official record clearly pertains to the site/project.",
      "reject": "General policy or program pages without a site-specific action."
    },
    "compute_tenancy": {
      "label": "Compute tenancy / occupancy",
      "objective": "Establish site-specific tenant/operator/customer evidence for compute infrastructure.",
      "search": [
        "Owner/operator releases",
        "customer/tenant disclosures",
        "lease or development records",
        "infrastructure project documents",
        "reputable local reporting"
      ],
      "queries": [
        "\"{site}\" tenant data center",
        "\"{site}\" customer cloud compute",
        "\"{site}\" lease data center"
      ],
      "capture": [
        "tenant/user",
        "relationship",
        "capacity/purpose if stated",
        "date",
        "URL/locator"
      ],
      "accept": "Site-specific occupancy or tenancy directly documented.",
      "reject": "A company's general use of cloud or data centers."
    },
    "power_telemetry": {
      "label": "Power observations / telemetry",
      "objective": "Find measured or reported electrical consumption/demand attributable to the site while keeping annual aggregates separate from interval telemetry.",
      "search": [
        "Utility/regulator filings",
        "public grid market/load data",
        "company sustainability reports",
        "site energy disclosures",
        "public planning/service documents"
      ],
      "queries": [
        "\"{site}\" electricity consumption MW",
        "\"{site}\" load MW utility",
        "\"{site}\" electricity usage annual"
      ],
      "capture": [
        "measurement type",
        "value/unit",
        "interval or period",
        "meter/site scope",
        "source date",
        "URL/locator"
      ],
      "accept": "Numeric observation with clear site scope and measurement period/type.",
      "reject": "Queue MW, service capacity, generation capacity, or contracted demand presented as measured load."
    },
    "remote_sensing": {
      "label": "Remote sensing",
      "objective": "Acquire repeatable optical/TIR/SAR evidence for construction, surface/thermal change, and campus evolution.",
      "search": [
        "Copernicus Sentinel-2 / Sentinel-1",
        "USGS Landsat optical/TIRS",
        "public STAC catalogs",
        "USGS 3DEP for geometry support"
      ],
      "queries": [
        "\"{site}\" satellite construction",
        "\"{site}\" satellite thermal data center",
        "\"{site}\" SAR construction"
      ],
      "capture": [
        "scene/product ID",
        "acquisition date",
        "AOI",
        "processing method",
        "derived metric",
        "source URL/catalog identifier"
      ],
      "accept": "Reproducible scene plus derived observation with provenance.",
      "reject": "A satellite image link with no retained observation/processing record."
    },
    "cooling": {
      "label": "Cooling infrastructure",
      "objective": "Identify site-level cooling equipment, cooling plant design, water use, heat rejection, or cooling infrastructure.",
      "search": [
        "Permits and planning documents",
        "water/environmental records",
        "owner/developer disclosures",
        "Epoch cooling datasets",
        "equipment/project announcements"
      ],
      "queries": [
        "\"{site}\" cooling plant",
        "\"{site}\" cooling towers chiller",
        "\"{site}\" water use data center"
      ],
      "capture": [
        "equipment/type",
        "quantity/capacity if stated",
        "water/thermal metric if stated",
        "date",
        "URL/locator"
      ],
      "accept": "Cooling evidence is site-specific and source-backed.",
      "reject": "Generic data-center cooling design not tied to the site."
    },
    "transformer_supply_chain": {
      "label": "Transformer supply chain",
      "objective": "Find site-specific transformer specification, procurement, shipment, delivery, installation, or energization evidence.",
      "search": [
        "Utility capital/project filings",
        "public procurement/bid records",
        "owner/developer disclosures",
        "manufacturer/contractor project pages",
        "transport/logistics reporting"
      ],
      "queries": [
        "\"{site}\" transformer",
        "\"{site}\" substation transformer",
        "\"{site}\" transformer delivery procurement"
      ],
      "capture": [
        "manufacturer",
        "model/specification",
        "quantity",
        "procurement/shipment/install status",
        "date",
        "source URL/locator"
      ],
      "accept": "Event tied to the site or its electrical project.",
      "reject": "Manufacturer product announcement with no site linkage."
    },
    "independent_corroboration": {
      "label": "Independent corroboration",
      "objective": "Attach a materially independent public source corroborating a site-specific claim already supported elsewhere.",
      "search": [
        "Utility/regulator records",
        "municipal records",
        "official owner/developer documents",
        "independent reputable reporting",
        "independent remote sensing"
      ],
      "queries": [
        "\"{site}\" utility",
        "\"{site}\" permit",
        "\"{site}\" construction"
      ],
      "capture": [
        "claim being corroborated",
        "second source",
        "source type",
        "date",
        "URL/locator",
        "what exactly agrees/disagrees"
      ],
      "accept": "Second source has a distinct origin and supports the same site-specific claim.",
      "reject": "Mirror copies, syndicated duplicates, or self-referential reposts."
    },
    "grid_connection": {
      "label": "Grid connection / interconnection",
      "objective": "Resolve the site-specific public connection record or documented utility/grid relationship, while distinguishing verified queue IDs from site-level evidence and jurisdiction-only context.",
      "search": [
        "ISO/RTO/market queue records",
        "serving utility interconnection and large-load filings",
        "state/provincial utility regulator dockets",
        "municipal/development records",
        "reputable local reporting"
      ],
      "queries": [
        "\"{site}\" interconnection {state}",
        "\"{site}\" utility queue {state}",
        "\"{site}\" large load {state}"
      ],
      "capture": [
        "utility/grid system",
        "queue or project ID if exposed",
        "connection status",
        "MW/capacity type",
        "case/docket ID",
        "date",
        "URL/locator"
      ],
      "accept": "A public record explicitly identifies the site and its connection/service relationship or queue/project identifier.",
      "reject": "A jurisdiction-level queue record that cannot be defensibly matched to the site."
    },
    "chip_inventory": {
      "label": "Chip inventory / site hardware",
      "objective": "Verify site-specific accelerator inventory and chip-type attribution from canonical or primary site records; preserve uncertainty where only estimates exist.",
      "search": [
        "Epoch AI site-level chip records",
        "owner/developer infrastructure disclosures",
        "supplier/manufacturer statements with site linkage",
        "planning/permit/project documents",
        "reputable independent reporting"
      ],
      "queries": [
        "\"{site}\" GPU inventory",
        "\"{site}\" H100 H200 chips quantity",
        "\"{site}\" accelerator inventory"
      ],
      "capture": [
        "chip type",
        "quantity",
        "measurement/reporting date",
        "source wording",
        "site linkage",
        "URL/locator"
      ],
      "accept": "A site-specific chip inventory or clearly attributed estimate is preserved with its source and date.",
      "reject": "Global chip ownership/users/shipment figures without site allocation."
    }
  }
}