First attempt (AI integration)

Goal: To help people who have never played games or are just starting out enter the world of video games without fear or snobbery. Objectives: · To explain where to begin in each genre. · To showcase classic games and explain their significance. · To provide clear entry points into the world of gaming. · To create a feeling that gaming is accessible and simple. · To develop a clear and understandable system covering different genres, game themes, and advice.

Core principle: Artificial intelligence collects and sorts information; the human reviews, supplements, and finalizes it. AI does not publish anything directly.

1. Telegram Comment Analysis. AI reads all comments and messages in the chat, highlights frequently asked questions using keywords, counts recurring themes, and compiles a list of the «Top 5 Questions of the Week.» A human reviews each suggested question in its entirety to grasp the context (AI does not always capture irony or underlying meaning), selects one or two of the most interesting, and decides which to address in the «Question from a Beginner» segment. • Source — only your channel and chat. Do not gather data from other channels or private messages. • Redact personal information (phone numbers, addresses, names) before presenting it to the human reviewer. • Filter toxicity: profanity, insults, and spam must be excluded from suggestions. • Do not interpret irony or sarcasm—flag such messages as «requires human review.» • Output format: question + frequency of mention + link to the message. • Do not suggest a question again if it has already appeared in the «Question from a Beginner» segment within the last three months.

2. Key Word and Topic Collection. The AI analyzes which words and topics subscribers mention most frequently in comments (e.g., «weak PC, ” „control, ” „scary, ” „difficult“) and generates a tag cloud, allowing a human to see what concerns the audience. The human reviews the tag cloud and decides which topics to focus on for the next month. • Only count words that repeat 3 or more times. Isolated mentions are less significant. • Exclude stop words (e.g., „this, ” „like, ” „well“). • Group synonyms: „weak PC“ and „old laptop“ are one discussion topic. • Do not draw conclusions—only show frequency. The decision of „what to work on“ remains with the human.

3. Forum and Discussion Analysis. AI scans DTF, StopGame, and Reddit using keywords («beginner, ” „where to start, ” „I don’t understand“) and collects recurring pain points: „I can’t figure out the controls, ” „I’m afraid to play multiplayer, ” „I don’t know which game to choose.“ A person reads the collected snippets, filters out toxic or irrelevant content, selects 2–3 pain points that align with their target audience, and converts them into post topics. • Only public platforms. Do not parse private chats or paid sections. • Only recent discussions, no older than 6 months. • Filter out toxic and irrelevant fragments (debates unrelated to beginners). • Output format: pain point + quote + link to discussion. • Do not present one person’s opinion as the „audience opinion”—show how many times the pain point recurred.

4. Critic Review Analysis. AI gathers ratings from Metacritic, OpenCritic, and reviews from Steam (only «Very Positive» and above), and highlights the most frequently mentioned pros and cons. A human reads the reviews gathered by the AI and decides whether to write about the game in simple language or if it is «for experienced players.» The AI provides the facts—the human translates them into beginner-friendly language. • Only include games with a Steam rating of «Very Positive» or higher—do not recommend questionable games to beginners. • Only include games that have Russian language support (or official/fan translations). • Highlight the pros and cons that are mentioned most often (3+ times). • Do not paraphrase reviews verbatim—only use key points. • Flag the game as «for experienced players» if the reviews are filled with jargon that is outside of your slang vocabulary.

5. Search for similar games on Steam. The AI uses the tags of the current game («puzzle», «atmospheric», «single-player») to find 5–10 similarly rated, highly-rated games, checks for Russian language support, and compares system requirements to the initial game. • Only games with «Very Positive» ratings or higher. • Must include Russian language support. • System requirements must be compared against the first game; do not suggest a game with higher requirements. • Maximum of 3 options—do not overwhelm the new user with choices. • Verify that the game is still on Steam and accessible in the user’s region.

6. Rating and Price Verification. Once a month, the AI cross-references the prices of games previously mentioned in posts with their Steam rating. If the price has changed or the game has been removed from the store, the AI creates a task: «Update Post.» A human receives this task, verifies whether the information is genuinely outdated, and manually corrects the post or curated list. • Source — Steam API only. No third-party sites with keys. • If the price has changed—create a «Update Post» task, including the link. • If the game has been removed from Steam—the task is «Remove from lists or mark as unavailable.» • If the system requirements have increased—the task is «Check if it will still run on a low-spec PC.»

7. Gathering news for the digest. AI processes 3–4 trusted gaming websites (StopGame, DTF, official developer sites) using keywords («release,» «update,» «free,» «discount»), filters out irrelevant content, and forms a list of 5–7 news items. A human reads the list and filters out news that does not pertain to beginners (esports, microtransactions, scandals). They select 3 news items and write a brief explanation for each: «why this is important for a beginner.» • Sources must be trusted: StopGame, DTF, official developer sites. • Do not collect from random public pages, forums, or Telegram channels without verification. • Filter out irrelevant topics: esports, microtransactions, and scandals are not relevant to beginners. • Output format: headline + source link + one sentence explaining «why this is important for a beginner.» • Maximum of 7 news items—the human will select 3. • Mark items as «requires verification» if the news appears on only one source.

8. Toxicity and Personal Data Filtering. The AI scans comments and chat messages for profanity, insults, phone numbers, and addresses—and removes them before displaying them to the user. The user still views the filtered content: the AI might delete an innocent joke or overlook masked toxic content. • Mask personal data (phone numbers, addresses, names, passport details) • Flag toxic messages without permanent deletion—the user must see what was filtered. • Do not filter innocuous jokes—if in doubt, flag for «manual review.» • Log all deletions—so that it can be verified whether the AI deleted anything important.

9. Content Sorting by Categories. The AI automatically sorts gathered materials into categories: «Where to Start, ” „Classics, ” „Slang, ” „Myths”—based on keywords and tags. • Sort by keywords and tags, not by meaning—a human verifies the meaning. • If material fits 2+ categories—flag it as „requires human review.“ • Do not delete materials that fit no category—place them in the „Idea Archive.“ • Output format: Category → list of materials → links.

10. Reminders and Tasks. AI generates a weekly task list: «Update the Sims 2 post — price has changed, ” „Respond to the weak PC question, ” „Check out the new puzzle game based on tags.“ • Maximum of 10 tasks per week — don’t overwhelm the user. • Prioritization: address error-related tasks first (outdated price, removed game), followed by new content. • Each task must include a link and a reason („price changed from 500 to 800 ₽“). • Do not create a task if the user has already closed it — check the status. • Once per week — a summary: „Completed / Not Completed / Overdue.“