{
  "study_id": "tikchat-twitch-passive-2026-11-15-v1",
  "study_dates": [
    "2026-11-15",
    "2026-11-16"
  ],
  "study_type": "first_party_passive_single_broadcast_browser_observation",
  "commissioned_by": "Matt Olech, TikChat owner",
  "observer": "AI assistant operating browser tabs",
  "platform": "Twitch",
  "environment": "production TikChat, owner-started session",
  "source_identity": "Twitch-A; creator, channel and viewers anonymized",
  "scope": {
    "time_unit": "milliseconds_relative_to_formal_reference_start",
    "observation_window_ms": 6381621,
    "observation_duration": "106 minutes 21.621 seconds",
    "final_comparison_window_ms": 90000,
    "analysis_cutoff_ms": 6471621,
    "reference_first_read_start_ms": 16141,
    "reference_first_read_end_ms": 16153,
    "reference_last_read_end_ms": 6357169,
    "reference_unobserved_leading_ms": 16141,
    "reference_unobserved_trailing_ms": 24452,
    "final_operator_reads": 10,
    "final_operator_first_read_start_after_reference_end_ms": 27609,
    "final_operator_last_read_end_after_reference_end_ms": 73311,
    "final_comparison_unobserved_trailing_ms": 16689,
    "post_cutoff_samples_included": 0,
    "observation_started_mid_broadcast": true,
    "broadcast_still_reported_live_at_reference_end": true,
    "stop_reason": "owner_requested_end_while_broadcast_continued",
    "whole_broadcast_observed": false,
    "broadcast_start_end_peak_and_total_duration": "not measured",
    "scope_note": "The reference window is an observation interval, not the duration of the broadcast. The 90-second final period is a sampling window, not continuous recording. Artifact computation completed later; no observations after the cutoff were counted."
  },
  "unit": "distinct normalized rendered-author plus normalized plain-text groups; NOT unique messages or provider IDs",
  "native_provider_ids_available_in_DOM": false,
  "baseline_groups_excluded": 130,
  "results": {
    "sampled_native_groups": 6003,
    "matched_operator_groups": 5899,
    "pending_groups": 0,
    "unresolved_groups": 104,
    "group_match_percent": 98.26753290021657,
    "percentage_denominator": "6003 new author-plus-plain-text groups observed in the sampled native DOM; not all Twitch messages",
    "groups_first_matched_after_60s": 7,
    "late_match_interpretation": "first browser-observation difference; may include observer pauses or repeated content; not measured transport delay"
  },
  "content": {
    "matching_normalized_groups": 5413,
    "ambiguous_matched_groups_excluded_from_pairwise_content_claims": 486,
    "variation_selector_only_groups": 0,
    "emote_representation_review_groups_in_unambiguous_subset": 0,
    "other_differences_in_unambiguous_subset": 0,
    "unavailable_content_in_unambiguous_subset": 0,
    "reference_digest_revision_events": 29,
    "operator_digest_revision_events": 253,
    "revision_unit": "content digest changes, not necessarily distinct groups; includes baseline observations",
    "interpretation": "Pairing fixes the normalized plain text and author. Equality of those fields is partly determined by matching. The separate content check covers complete rendered text including inline-image alt labels. It does not independently validate wording that failed to match or prove message identity. Ambiguous groups can include differences and repetitions; zero other differences in the remaining subset is not a zero-defect result."
  },
  "measurement_stages": {
    "send": {
      "status": "not_measured",
      "sender_attempts": null,
      "Twitch_accepted_messages": null,
      "experimenter_messages_sent": 0
    },
    "native": {
      "status": "sampled_signature_groups",
      "groups": 6003,
      "read_count": 968
    },
    "receipt": {
      "status": "not_independently_measured",
      "received_messages": null,
      "database_rows_read": 0,
      "note": "DOM presence was not relabelled as an independent backend receipt measurement."
    },
    "render": {
      "status": "sampled_presence_in_operator_DOM",
      "reference_groups_found": 5899,
      "read_count": 978,
      "note": "DOM presence, not proof every row was within the visible viewport or seen by the creator."
    }
  },
  "join_and_reconnect": {
    "initial_connection_measured": false,
    "late_source_attach_tested": false,
    "forced_provider_reconnects": 0,
    "controlled_reconnect_tested": false,
    "isolated_product_reconnect_test_performed": false,
    "local_software_counter_tests": 12,
    "local_test_scope": "arithmetic, grouping, ambiguity, eligibility and timing logic; not Twitch transport or TikChat provider recovery",
    "future_controlled_messages_prepared": 120,
    "future_controlled_messages_sent": 0
  },
  "coverage": {
    "reference": {
      "reads": 968,
      "medianIntervalMs": 5002,
      "p95IntervalMs": 13090,
      "maxIntervalMs": 118726,
      "possibleBufferGaps": 1,
      "truncatedReads": 0,
      "unkeyedRowObservations": 1685,
      "unknownRowObservations": 271,
      "duplicateDomRows": 7256,
      "rejectedReads": 0
    },
    "operator": {
      "reads": 978,
      "medianIntervalMs": 5002,
      "p95IntervalMs": 13164,
      "maxIntervalMs": 118729,
      "possibleBufferGaps": 0,
      "truncatedReads": 966,
      "unkeyedRowObservations": 0,
      "unknownRowObservations": 618,
      "duplicateDomRows": 79874,
      "rejectedReads": 0
    }
  },
  "production": {
    "application_release": "release-sales-multichat-b5fc1dfbf0a6",
    "www_release": "youtube-study-links-79d826914425",
    "version_and_service_checks_unchanged": true,
    "checkpoints": "before observation, during observation, after completion; not a continuous trace",
    "deployments_by_observer": 0,
    "service_restarts_by_observer": 0,
    "production_fault_injections": 0,
    "additional_TikChat_collectors_by_observer": 0,
    "additional_TikChat_API_poll_loop": false,
    "production_database_access_by_observer": false,
    "server_resource_usage_measured": false
  },
  "limitations": [
    "single convenience-selected broadcast, first-party observation, no independent reviewer",
    "reference DOM has no provider message IDs; no unique-message count or exact-ID agreement",
    "same author and same normalized plain text collapse into one group throughout the study",
    "all signatures already in the initial native buffer remain excluded even if repeated later",
    "matching content forms can refer to different message occurrences; order alone does not establish identity",
    "emote-only and rows without a usable rendered author/body are excluded and counted as repeated DOM observations",
    "bounded buffers, nonuniform sampling and unknown number of missed comments",
    "a zero-overlap warning detects only some possible sampling gaps; overlap does not prove complete coverage",
    "operator used at most 800 recent matching DOM rows; this deliberate bound is not application loss",
    "unresolved is not proven backend loss, and observed time differences are not transport latency",
    "send, independent backend receipt, late join and controlled reconnect were not measured",
    "content checks apply only to matched non-ambiguous groups; do not prove all comments agreed",
    "viewer counts, audience peak/average, engagement and server CPU/RAM/I/O/network were not measured",
    "working fingerprints and salt were disposed; message-level reconstruction cannot be independently replayed"
  ],
  "privacy": "No source channel, URL, viewer name, author identifier, comments, raw message IDs, fingerprints or nonce in this publication package; result times are relative. Original start/end wall-clock times remain outside the package.",
  "publication_copy_provenance": {
    "original_package": "research-package.zip",
    "original_record_path_in_package": "data/study-record.json",
    "numerical_study_results_changed": false,
    "workflow_status_field_omitted": true,
    "reason": "Workflow publication status is not a measured outcome; the original record is preserved unchanged in the research ZIP."
  }
}
