«TO and TO”—a Telegram media outlet focused on the philosophy and ethics of human-AI interaction in art. This media outlet does not take sides in debates, nor does it pretend to have the answers. Because it doesn’t.Its goal is to demonstrate that the questions AI poses to creativity are more complex than they appear. Readers leave not with a definitive opinion, but with an understanding of why there might not be a single answer.Formats: long-form essays, „Dual Portrait, ” „Question Without an Answer, ” „Dialogue, ” and „Voice from Within.“Audience: the questioning artist, curator, or critic; the researcher or philosopher; the curious reader.Unique Selling Propositions (USPs): — „There is no answer. And this media outlet doesn’t pretend there is.“ — „Two sides, without judgment.“ — „The reader is a participant, not just an observer.“

Media works with three types of sources.

Primary Sources (Created by the Editorial Team)Original Content—longreads, essays, dialogues. This is the core content of the media. Created by an author or the editorial team.Interviews with artists, curators, and researchers—material for «Voices from Within» and «Dialogues.» Created by an author.Reader polls on Telegram—material for «Questions Without Answers» and feedback. Created by readers in collaboration with an author.

Secondary Sources (External, Requiring Selection and Vetting)

Academic articles on AI philosophy, aesthetics, and art theory — forming the theoretical foundation for essays. Used through citation, paraphrase, and critique.

Publications by artists and curators — manifestos, essays, and interviews. They provide viewpoints for «Double Portraits.» Used through citation and comparison.

Industry news — AI art exhibitions, legal cases, and tool releases. This provides material for «The Unanswerable Question.» Used for reaction and commentary.

Open museum data (Met, Art Institute of Chicago, Cleveland Museum) — visual material and context. Used for illustrations and examples.

Visual SourcesImages generated from editorial prompts—illustrations for posts.Public domain museum collections (CC0)—historical examples and references.Screenshots of interfaces and user work (with permission)—material for analysis.

The diagram illustrates the journey of a single material from concept to publication.

Phase 1. Collection and Selection

Topic Search. Human: The editor selects the theme of the week from the thematic plan. Machine: Not involved.

Source Gathering. Human: The editor selects 3–5 relevant sources. Machine: Search services (Wordstat, Google Scholar, Telegram search) help locate materials.

Visual Selection. Human: The editor decides whether an image is needed and what kind. Machine: The image generator creates variations based on the prompt.

Stage 2. Processing

Text Writing. Human: the author writes the text, verifies facts, and formulates positions. Machine: not involved.

Fact-Checking. Human: the editor cross-references quotes and data with sources. Machine: not involved.

Visual Preparation. Human: the editor selects the final image from the generated options. Machine: the generator creates 4 variations, and the editor selects one.

Formatting. Human: the editor lays out the post for Telegram. Machine: not involved.

Phase 3. Publication and Feedback

Publication. Human: The editor publishes the post. Machine: The Telegram API delivers the post to subscribers.

Reaction Collection. Human: The editor reads and responds to comments. Machine: The bot gathers statistics (views, reactions, shares).

Analysis. Human: The editor decides which topics «landed well» and what needs to be changed. Machine: Not involved.

Next Cycle. Human: The editor adjusts the plan. Machine: Not involved.

Key principle: The machine assists with search, option generation, and statistics gathering. The human makes all substantive decisions—what to write, which positions to feature, and what to publish.

Rules by which the machine participates in the production of «TO and TO» content.

Guideline 1. The Machine Does Not Write CopyAutomated services do not generate substantive text for media. Humans write the text. A machine can assist with source research, but not with formulating core arguments.Why: Media is built on honest reflection. Generated text cannot «not know the answer”—it simulates knowledge. This contradicts the core concept.

Guideline 2. The Machine Generates Only Visual ConceptsThe image generator produces 3–5 visual variations based on the editor’s prompt. The final selection is made by the human.Why: the visual is an illustration to the text, not a standalone statement. The machine can suggest forms, but it cannot determine which form aligns with the intended meaning.

Guideline 3. The machine collects statistics, but does not interpret them.

The bot gathers views, reactions, shares, and comments. The editor reads the numbers and draws their own conclusions.

Why: Numbers do not explain why a post succeeded. Interpretation is a human task.

Guideline 4. The machine does not make publishing decisionsNo post is published automatically. Every piece of material passes through an editor.Why: media is responsible for what it publishes. Responsibility cannot be delegated to an algorithm.

Guideline 5. The machine does not interact with readers on behalf of the media.

A person responds to comments. The bot can only collect statistics.

Reason: The reader must know that they are communicating with a person, not an algorithm. Otherwise, the trust upon which the media is built will collapse.