Как искусственный интеллект создает миллионеров. ПОПРОБУЙ И ТЫ

Core technological ideas / product-style claims

• Neural networks “hallucinate” or output nonsense mainly due to bad prompting/context, not because the model is inherently broken. The speaker argues that users often query the model without providing:

  • the right device/context,
  • a relevant knowledge base,
  • clear goals/style constraints (what the user needs and how they want the response).

• Using AI as an “accelerator and multiplier,” not a replacement for expertise

  • If the user’s knowledge is zero, the AI can still produce polished-looking output that is effectively zero in substance.
  • If the user has deep domain knowledge, AI multiplies value (speed, packaging, output volume).

• AI can automate many “production” tasks without code, enabling a solo creator to operate like a small team. Examples include:

  • Budget tracking in Telegram → logs to Excel / Google Sheets via a bot
  • Document/photo-to-text transcription via a Telegram bot (e.g., business cards/memes)
  • Text-to-audio with natural voice; optionally by roles
  • Book summarization (“samaritizer”) to decide whether a text is worth spending hours on
  • Event poster aggregation: collects performances/exhibitions/matches and scrapes descriptions + images + ratings from Kinopoisk and IMDb (named “AMDB” in subtitles)
  • Content tooling: cover image analysis; podcast chaptering/clipping for distribution to networks like X/Messenger
  • A “time machine” for finding comic/rail ideas and pre-checking what may “take off”
  • Club chat automation: statuses → perks (e.g., publishing eligibility)
  • A “literary editor robot” to check logic/nonsense in ideas
  • Rapid landing page creation “in 1 hour / 30 minutes” (AI-assisted web design claim)
  • YouTube competitor cover image analysis and an idea verification workflow

Tutorials / guides / workflow patterns mentioned

How to prompt/use neural networks effectively (implied guide)

1. Work with the model properly (speaker claims neural networks must be worked on from a computer).
2. Provide relevant context: knowledge base, goals, and response style.
3. Treat output as something to verify: detect hallucinations/stupidity and extract only useful “grain.”
4. Use AI for iteration and critique, not as a mindless answer generator.

Teacher / education approach using AI (explicit advice)

• Don’t let students simply “write essays with neural networks.”
• Instead:
  • require AI to argue with the conclusions,
  • find weaknesses and logical errors,
  • ask clarifying questions,
  • connect claims to real-world objects.
• Teachers should repackage material into short formats, such as:
  • summaries,
  • mistake lists,
  • tests with traps,
  • games with points,
  • case studies.
• Teachers should become “bloggers / digital centaurs” to remain competitive.

Content creation workflow example (Reddit case study)

• Collect Reddit posts via a script.
• Send them to a Google LM “Deep Research” style workflow.
• Identify recurring templates (e.g., title structure, lengths, themes like numbers, personal confession, “spiritual crisis” tone).
• Generate a post.
• Use an AI skill to remove/adjust “Russian neuroslope” vs “English neuroslope.”
• Claimed results:
  • 3,500 likes
  • 500k+ views
  • 10 days of comment debate

Analysis of labor market / business model changes (tech-adjacent arguments)

Why layoffs happen “quietly”

• Employers stop contacting freelancers/employees for tasks that are now automated (e.g., transcription, timecodes, image tasks, assembly).
• No need for public announcements—demand simply disappears.

Compensation model collapse / efficiency penalty

• The speaker frames hourly-rate pricing as breaking due to AI reducing time dramatically.
• Example narratives:
  • “Emergency locksmith” story: clients pay for felt effort, not raw speed.
  • “SEO Polit” anecdote: added intentional “magic” animations/spinner/processing to make results feel credible.
• Core claim: pricing shifts toward perceived value and trust, not hours.

Startup / venture capital shift due to AI

• MVPs can be built in an evening without large teams → cheaper, faster tests.
• Audiences (Telegram/YouTube) enable “free hypothesis testing” vs paid advertising.
• Trust is not replicable like code:
  • “500k audience trust can’t be copied overnight.”

Perspective on “AI writing vs creativity”

• The speaker rejects the claim that AI eliminates creativity.
  • AI helps compile/research and filter ideas.
  • The speaker positions himself as a “verifier” who removes weak ideas and trash.
• Experience/expertise still matters for distinguishing “diamond vs trash.”

Critique of “AI detection” / “neuroslope detectors”

• Detectors can be fooled if people learn and adapt to evade them.
• Detectors may misclassify texts by latching onto superficial patterns.
• Core stance: even if a detector labels content “AI-generated,” that doesn’t prove lack of expertise.

Main speaker / sources

• Main speaker: Anar (Babaykin) (retired at 35; creator/author; hosts Telegram/YouTube/Max content)

• Named external references/tools:

  • Google (mentioned “Google laptop LM” and “Deep Research”)
  • Claude (text-to-workflow tool mentioned)
  • Telegram, Excel, Google Sheets
  • Kinopoisk, IMDb (event poster integrations)
  • Pavel Durov / Telegram ecosystem (mentioned as a comparison)
  • Nikita Nepryakhin (book: “I am manipulating you”)
  • Misha Tokovenin (interview mention about books/content formats)
  • Elon Musk (referenced as an example of trust via real achievements)
  • Nick Majuli (inspiration for “businessman vs creator”; speaker says he stole the idea from an American colleague)