We are creating a digital space where the literature and underground culture of 20th-century St. Petersburg are revealed through its places, personalities, and eras. The reader engages with quizzes, «guessing games,» and narrative stories, and can then follow the same path through real-life locations with a guided tour. The project is founded on the idea that it was precisely these places—as magnets drawing literary elites—that shaped the spirit of their time.

The project has two interconnected goals. The first is educational: to explain what the Silver Age, samizdat, and the Leningrad underground of the 1950s–1980s were; who Okhapkin, Galetsky, and Zvyagin were; and why Akhmatova’s apartment or «The Wandering Dog» is significant beyond just being a location. The second is commercial: to connect the content with the audience’s real-world experience and monetize traffic through native advertising for tours related to the same figures and locations.

We are active on VK and Zen, and we use Telegram for announcements and quick communication. The content features quizzes, «guess what,» and tests on historical knowledge. The tour isn’t a separate sidebar element; it evolves organically from the material—first, the reader learns the history of the location, and then they can see what pathways are available there.

Media material is collected from several groups of sources:1. The first group consists of archival and documentary sources: memoirs, letters, diaries, museum and archive documents, old photographs, posters, and samizdat covers.2. The second group comprises scholarly literature: academic articles, local history books, guidebooks, and publications from museums and universities.3. The third group includes living sources: interviews with experts, tour guides, and researchers.4. The fourth group consists of field notes: observations along routes, photographs of locations, and on-site clarifications.5. The fifth group includes materials provided by tour agencies and museums.6. The sixth group includes open data and old maps, if necessary to verify an address or location.

The first example is from the archive imwerden.de, featuring the publication «The Stray Dog Program.» It is based on a collection of programs and event schedules from the private archive of A. A. Golubev. This source provides specific dates for evenings, names of participants, and program titles. These are not general descriptions but discrete archival units to which one can attach a place record and a personal record.

A second example is samizdat.wiki, the Electronic Archive of the Andrei Bely Center. A search for «Archive of the journal Chasy» provides access to the complete collection of the journal «Chasy»: 80 issues from 1976 to 1990, as well as individual issues of «Krasny Shedrints» and «Archive.» This source contributes to the scheme with issue titles, years, editorial staff lists, and connections to other samizdat publications. This forms the basis for the «Genres and Movements» taxonomy and for material cards.

First: the source must have an author or organization, and it must be clear who is responsible for the information. Second: there must be a link to the original source or archive. Third: the fact must be confirmed in at least two independent places. Fourth: the date and context must be clear—that is, when and why it was written. Fifth: the data must align with what can be observed on the ground or in other verified materials. Sixth: there must be no overt promotional or political agenda that distorts the presentation.

A documentary source is superior to retelling. If a fact is debatable, it is either not presented as an assertion or is accompanied by a note indicating conflicting versions. If a source is found but not verified, it receives the status «requires verification» and will not be published until confirmed.

The project is built upon nine entity types.Location. Courtyard, embankment, apartment, cabare, communal area. Persona. Poet, prose writer, publisher, artist, tour guide. Era. Silver Age, Thaw, Stagnation. Genre or Movement. Poetry, prose, samizdat, Leningrad underground. Medium. Post, article, clip, gallery. Quiz. Question, options, correct answer, explanation. Tour. Route, locations, personas, partner. Source. Archive, book, interview, museum website, digitized collection. Media File. Photo, audio, video, scan.

Location is related to materials and tours. Person is related to materials, locations, and era. Material is related to the source, genre, era, quiz, and tour. Tour is related to locations and people. Quiz is related to the material and verification rules. Source is related to the material and the entities derived from it.

The process is structured as a sequence of steps, where each subsequent step triggers automatically, and human oversight is required at specific checkpoints.Collection. The service gathers data from open sources according to predefined queries and schedules. Cleaning and Tagging. Duplicates are removed, entries are standardized, and dates, names, addresses, titles, tags, and entity relationships are extracted. Drafting. The system compiles a draft of the material: structure, facts, quotes, source links, and suggestions for quizzes and location/tour tie-ins. Review. The draft undergoes quality control: fact-checking, source verification, image rights clearance, tone assessment, and legal risk review. Editing. The editor rewrites the text to fit the project’s voice, refines phrasing, removes extraneous material, and adds context. Publication. The material is published according to the schedule, entity relationships are recorded, and the quiz and tour integration are activated. Feedback. Audience statistics and reactions are fed back into the system to influence subsequent topics and selections.

1. Retrieves data from open sources on schedule and upon request.
2. Removes duplicates and standardizes entry formats.
3. Extracts dates, names, addresses, titles, and entity relationships.
4. Suggests tags, categories, and historical periods.
5. Builds a search index and displays related materials.
6. Drafts content based on a template: structure, facts, quotes, and source citations.
7. Prepares a quiz draft: question, options, correct answer, and explanation.
8. Suggests location and tour associations.
9. Reviews drafts for formal errors: missing source for a fact, missing answer explanation, or missing location link.
10. Verifies rights: image licensing, citation permissibility, source usage terms, and indicators of personal data.
11. Calculates statistics and sends publication reminders.

1. Sets themes, prompts, and collection guidelines.
2. Reviews the pre-check report and determines which comments are critical.
3. Verifies facts and assesses the reliability of the source.
4. Evaluates legal rights and risks based on all currently available information, and makes the final decision.
5. Edits the draft: removes unnecessary elements, clarifies wording, and adds context and project voice.
6. Checks quiz questions and answer explanations.
7. Makes the final decision regarding publication and linkage to the tour.
8. Addresses disputed cases and errors, and updates the rules for the machine.

The machine does the heavy lifting: it gathers, structures, drafts, prepares the quiz, and checks basic compliance criteria. The human reviews, refines, and makes the final decision, considering all available information at the moment. Automation speeds up the routine tasks and allows for higher output, but the semantic and legal responsibility remains with the editor.

Now, we need to establish the rules by which the automation operates within the project. This moves away from technical documentation and closer to a clear declaration of principles governing how the machine assembles, drafts, reviews, and passes material to the human user.