Video summary

Как Я Нахожу НЕТРОНУТЫЕ Ютуб Ниши Используя ИИ [CLAUDE CODE]

Main summary

Key takeaways

Technology

Main idea / outcome

  • The video teaches a step-by-step workflow to find profitable YouTube “niches” using AI (Claude) and an agent-based toolchain, so creators don’t spend weeks searching manually.
  • Claimed results: one channel example reportedly reached $14,000 earnings and 2.5M views using only 10 videos, and the author attributes success to choosing the right niche.

Key technological / product workflow (tutorial-style)

  1. Install a desktop AI coding client

    • Use Google “AntiGravity” (the speaker’s preferred client), which includes an agent inside to run tasks.
    • If you don’t want to code, you can “explain in words” and the agent performs actions.

  2. Install Cloud Code (via the agent)

    • In the client, prompt the agent to install “cloud code” automatically.
    • Then use a terminal command to install the next component.

  3. Set up Claude MCP connection

    • The workflow uses MCP (Model Context Protocol) to connect tools/services to Claude.
    • Steps include:
      • Go to an “MCP Nextle” website.
      • Obtain/copy the Claude-related MCP connection data.
      • Paste it into the terminal with commands to install the MCP.

  4. Run a niche-search prompt

    • Open a new terminal/session so the MCP integration is active.
    • Provide a single prompt (in English) requesting niches/YouTube channels that match filters, including:
      • < 200,000 subscribers
      • Channel has viral evidence at least once (at least one video with ~400,000 views)
      • Started ~3 months ago and is still publishing consistently

  5. Claude returns results in structured groups + tables

    • Claude outputs 30 niche candidates and groups them by likelihood:
      • Group 1: “confirmed viral”
      • Group 2: “very high potential / near threshold”
    • The author adds a refinement request:
      • “Make everything one table and add working links.”

Claimed efficiency / scaling

  • The author says the system finds 30 channels in ~5–10 minutes.
  • Further automation is suggested:
    • Search by category
    • Run for multiple tabs to find ~300 channels at once
    • Target: “at least 10 channels every evening” suitable to start tomorrow

Evaluation / examples shown (analysis of outputs)

  • The video demonstrates checking suggested channels, including examples with strong performance despite low counts of videos (e.g., few videos but extremely high view totals).
  • Overall pitch: the returned channels are meant to match criteria for an “ideal niche,” and links should be immediately usable for review.

Mentions of community / results (not the tool itself)

  • The author references their private community “Fusion”, claiming students (and the author) achieved monetization and income using the method.
  • It also claims help with issues like YouTube application rejection (framed as “bypassing immunization and rejection,” likely meaning reconsideration/approval).

Main speakers / sources

  • Main speaker: the author of the video (creator describing their Claude + Cloud Code / MCP Nextle / AntiGravity workflow).
  • AI agent used: Claude (via MCP Nextle integration).

Original video