
«Папка №»
1. Project Concept and the Role of Artificial Intelligence
1.1 Media Description
Title: «Folder No.» Segment: Historical and cultural educational media. Niche Topic: Historical investigations based on archival documents, eyewitness accounts, and lesser-known facts of Russian and Soviet history. Primary Platform: Telegram (with adaptability for long-form posts and social media). Goal: To make historical knowledge accessible, understandable, and engaging for a broad audience. The project narrates past events not through dry dates and official formulations, but through private lives, personal letters, diaries, archival files, and rare photographs. Key Approach: Fostering critical thinking in the reader. Each publication is structured as an investigation that separates verified facts from later myths, propagandistic interpretations, and popular misconceptions.
1.2 The Role of Artificial Intelligence
As part of the AI project, it functions as an archival assistant and editorial assistant. It is designed to automate routine processes, analyze audience inquiries, and generate initial drafts.
AI Objectives:
- Analyze audience questions and comments to identify popular topics for investigations.
- Group and structure facts from digitized archival documents.
- Formulate a preliminary structure and narrative arc for a historical investigation.
- Create draft publications for Telegram according to established editorial guidelines.
- Suggest visual design ideas and generate background/illustrative elements.
- Adapt approved materials for various formats.
Important: AI does not publish materials independently. The editor is responsible for fact-checking, historical accuracy, and the final decision on publication.
2. Data Sources for Media
To prepare publications, «Folder No.» uses a strict system of verified sources:
- Archival Documents: Digitized collections, state and departmental archives, transcripts, reports, orders, and protocols.
- Personal Collections and Ego-Documents: Diaries, private letters, memoirs, and notebooks of participants or witnesses of events.
- Periodicals of the Past: Newspapers and magazines from the studied era.
- Material and Visual Evidence: Historical photographs, maps, drawings, and physical evidence.
- Historiography and Expert Works: Publications and monographs by professional historians to verify context.
Information Assessment and Filtering Criteria:
- Provenance (Authenticity): Is the author or organization that created the document known?
- Context of Creation: Under what circumstances was the document written (official report, private diary «for the shelf,» interrogation under pressure)?
- Corroboration: Are the data confirmed by other independent sources?
- Historiographical Relevance: Has the document’s interpretation been refuted by contemporary historical research?
- Completeness: Is there sufficient data to form an objective picture?
3. Data schema separating human and machine processing
The publication creation process consists of eight sequential stages.
Artificial Intelligence Executes:
Stage 01 (Collaborative / Automated Collection): Gathering anonymized reader questions from comments, polls, and submissions into a unified database. Stage 02 (AI): Analyzing and grouping questions. The AI clusters submissions, identifies recurring themes, and forms a list of potential investigations. Stage 04 (Collaborative AI + Editor): Information gathering. The AI extracts facts, names, and chronology from uploaded documents and digitized archives. Stage 05 (AI): Draft generation. The AI structures the text in an investigative genre (intriguing headline, introductory problem, step-by-step narrative, conclusion). Stage 07 (Collaborative AI + Designer): Visual preparation. The AI proposes card concepts and creates stylized graphic elements. Stage 08 (Collaborative Editor + AI Analyst): Feedback analysis. The AI groups reader questions post-publication for new topics.
Creator (Editorial):
Stage 03 (Editor): Topic Selection. The Editor evaluates the topic for conceptual relevance, historical significance, and the availability of archival material. Stage 04 (Editor): Source Verification. The Editor verifies the selected AI sources and quotes. Stage 06 (Editor): Fact-Checking and Editing. The Editor checks every date and name, removes AI «hallucinations,» and structures the style and dramatic narrative. Stage 07 (Designer): Visual Design. The Designer lays out the cards, applies the text, and checks readability. Stage 08 (Editor): Approval and Publication. Final review and deployment of the material to the channel.
4. Guidelines for Automated Services
4.1 Working with Information
- Use only the provided archival sources and documents.
- Do not speculate on facts, people’s motives, or event details.
- Cite sources: every significant fact, figure, or quote must be linked to a document.
- Consider the historical context of the era, avoiding anachronisms and modern judgmental assessments.
- For controversial or ambiguous issues, you must point out the existence of multiple versions within the sources.
4.2 Data Protection and Ethics
- Refrain from using contemporary personal data.
- Adhere to ethical standards when working with personal archives and sensitive private circumstances.
- Do not retain confidential information longer than necessary.
4.3 Text Generation Requirements (Tone of Voice)
Style: Engaging, popular-science style infused with elements of historical detective fiction, avoiding dry academic jargon and bureaucratic phrasing. Tone: Addressing the reader formally («you»), respectful, and creating a sense of a shared quest for truth. Length: Up to 2,500–3,000 characters per post or series of cards. Structure: Title with a riddle or little-known fact. A short, intriguing hook (a problem or contradiction). Step-by-step: unfolding the investigation through archival documents. An unexpected detail or plot twist. Historical conclusion: debunking a myth or summarizing findings. Wording: Strictly delineate concepts: «the document states…» versus «historians hypothesize…».
4.4 Image Generation Requirements



