03 · Evidence archive

The document exists in two forms: a native DOCX and a derivative PDF.

One preserves metadata and construction history. The other preserves only what the page looks like. They tell different stories.

Document inventory

FileFormatDateProvenance
1776 Thesis.docxDOCX (OOXML)Dec 30, 2020Sent by Flores to Tarrio via Telegram. No longer available in original form.
telegram-cloud-document-1-5145562784171294938.docxDOCX (OOXML)Dec 30, 2020Telegram-archived copy. SHA-256: c8522cb1.... Subject of this forensic analysis.
signal-2026-06-20-200443.pdfPDF 1.5 (rasterized)Feb 8, 2022Rasterized recreation. 9 pages, each page is a JPEG image at 200 ppi. No native text layer. Metadata stripped.
signal-2026-06-20-200414.pdfPDF 1.6Mar 9, 2022Companion document: “Superseder_March 7_FINAL” covering Tarrio private messaging. Native text layer preserved.

Chain of custody

StepActionHolderDate
1Armes creates precursor document (Google Doc)ArmesAug–Sep 2020
2Armes shares precursor with Flores (Google Drive)Armes → FloresLate 2020
3Flores assembles final DOCX, edits with GrammarlyFloresDec 29–30, 2020
4Flores sends 1776 Thesis.docx to Tarrio (Telegram)Flores → TarrioDec 30, 2020
5Telegram cloud stores archived copyTelegram serversDec 30, 2020
6House Select Committee collects document from unknown sourceCommittee2021–2022
7Rasterized PDF created (Feb 8, 2022)UnknownFeb 8, 2022
8PDF produced to Rehl defense in discoveryGovernment → Defense2022–2023
9DOCX exported from Telegram, analyzed forensicallyIndependentJun 2026
Gap: Steps 6–7 are undocumented. We do not know who created the PDF, from what source, or why. The original 1776 Thesis.docx is not in the public record.

Grammarly artifacts

The DOCX contains Forensic two Grammarly DOX plugin variables embedded in word/settings.xml. These prove Grammarly was actively editing this document and establish a server-side trail.

VariableValueMeaning
__Grammarly_42____i{"DataPartCount":1}Document submitted as single part
__Grammarly_42___1{"DocId":"132537363747608562","SessionId":4}Grammarly document ID & at least 4 editing sessions
Strategic value: Grammarly DocId 132537363747608562 is the key to recovering full edit history, session timestamps, IP addresses, and author account identity via subpoena to Grammarly, Inc. (548 Market St, San Francisco).

Document identity GUIDs

The DOCX carries two persistent identity GUIDs that survive metadata stripping and format conversion:

ID TypeValue
w14:docId (Word 2010+)39D9AE9F
w15:docId (Word 2013+){8C755BB9-038D-4E0C-982A-5C94BCB64F6B}

These GUIDs are baked into the document XML at creation. If Armes’ or Flores’ machines are ever recovered, matching these IDs would be definitive identification of the source device.

Legal analysis

Brady implications

  • The PDF’s Feb 8, 2022 creation date is exculpatory: it shows the document is a recreation, not the original
  • If the government knew this and did not disclose it, that is a colorable Brady claim
  • D.D.C. standing Brady order requires information to be produced in its original tangible format

Authentication issues

  • Rasterized PDF cannot be authenticated as the original planning document
  • No chain of custody for the PDF creation process
  • No /Producer, /Creator, /Title, or /Author metadata in the PDF
  • ModDate identical to CreationDate despite incremental save—metadata manipulation
Precedent: The government’s production of a metadata-stripped PDF as substitute for original digital evidence mirrors the pattern identified in United States v. Raniere (E.D.N.Y.), where a paralegal-created PDF concealed camera-card tampering for over four years.
Chain of custody: 9 steps from Armes to independent forensic analysis.Next: Forensics →