Зашел в ВАЙБКОДИНГ С НУЛЯ и собрал 3 стартапа – сколько заработал

Executive summary

The presenter’s main message is that building a product with AI coding tools is only part of launching a business: choose a market with momentum, validate the problem, plan distribution early, and test with a small group before scaling.

His marketplace-content product, Kartochka, had about 40 users and brought in roughly ₽5,000 during its first week. He says it earned ₽3,000 the day before filming. He also describes a separate marketplace agency with ₽1.5–2 million in monthly turnover.

Business frameworks and playbooks

• Choose a growing market, then solve a concrete problem. The presenter advises avoiding markets he believes are shrinking and looking for customer pain points that could support a product. He gives his difficulty finding inexpensive parking in St. Petersburg as an example of an unresolved problem—and a possible app idea.
• Use research to generate and assess ideas.
  • Browse SaaS products and filter by category on TrustMRR.
  • Use Twigly’s weekly idea newsletter, which he says costs ₽900 per month or has a limited free version.
  • Ask Claude to research Reddit and SaaS products, summarize ideas, and identify potential MVP features.
• Plan marketing before building. Don’t assume a website will attract customers through word of mouth. The presenter recommends defining a promotion strategy in advance, with content marketing as a core acquisition channel.
• Test small before scaling. Launch to a limited audience, collect feedback, fix operational and product issues, and only then increase traffic. He suggests moving from an initial group of around 40 users toward a larger audience—for example, 1,000 users—after testing.
• Combine product expertise with audience access. The presenter argues that founders who understand a specific industry can build more relevant features and create content for the same target audience, making product and distribution reinforce each other.
• Use influencer marketing systematically. Estimate creator rates with a service such as LabelUp, negotiate placements, test a creative concept, and reuse successful formats with additional influencers.
• Build an MVP, then polish it. Scope the first version to one or two key features rather than reproducing a large competitor product. The presenter says an AI-generated MVP can be made in a day, while refining prompts and features may take one to two weeks or even a month, depending on the project.

Products and examples

• Kartochka: An AI service for creating marketplace product-card content. Features shown include product uploads, AI-generated images and infographics, size charts, prompt suggestions, animation, and previews of how content may appear on a marketplace.
  • The presenter says he had around 40 users after a small announcement and earned about ₽5,000 in the first week, including ₽3,000 the previous day.
  • He says he delayed larger-scale promotion because providers were failing and the product still needed testing.
  • He offers support through Telegram and says he refunds credits to encourage users to report problems.
• GEO/AEO service: A business-oriented project intended to help businesses understand and improve how they are surfaced or recommended by LLMs. The presenter says it grew out of research and work with clients; he plans to discuss it in a separate video.
• Parking-app idea: A not-yet-built concept based on finding inexpensive parking in hidden courtyards in St. Petersburg.
• Content-led agency: The presenter says his marketplace agency generates leads through YouTube content. He reports ₽1.5–2 million in monthly turnover, while typical channel videos receive 300–1,000 views. He also cites 10,000 views on his VK videos, while arguing that YouTube and Instagram are stronger organic-discovery channels. These are his own assessments, not independently verified results.
• Mark Lu example: The presenter points to a founder with 356,000 Twitter followers who publicly shares project revenue. He cites roughly $4,000 per month associated with 304 startups/projects; the exact relationship between the figures is unclear in the subtitles. The broader lesson is to build a targeted audience and direct it to products.

Product development and operations

• Research → MVP scope → specification → build → test → polish: Use AI research to compare ideas, choose a small feature set, write a technical specification, and use it to guide implementation.
• AI coding workflow: The presenter uses Claude/Claude Code, VS Code, and sometimes Cursor. He recommends keeping project rules and working context in a CLAUDE.md file, tracking tasks, and asking the coding agent to create a Git branch and proceed through the build. He cautions that more intensive agent workflows and higher-end models can consume more tokens.
• Version control: Save work in GitHub commits so the project can be rolled back when a change breaks something.
• Deployment: He initially used Vercel and Neon, then moved to Timeweb. He uses temporary deployment links to test changes before putting them on the main domain. He says Timeweb’s App Platform costs more but simplifies deployment. He also advises considering where data is hosted, including Russian hosting options for projects serving Russian customers.
• Get help from the tools: He recommends keeping a dedicated chat for basic development questions, sharing screenshots, and asking the AI to explain errors or infrastructure concepts.

Metrics and targets mentioned

Metric Figure stated Kartochka users after a small launch About 40 Kartochka first-week revenue About ₽5,000 Kartochka revenue on the previous day ₽3,000 Suggested next testing scale Up to 1,000 users after initial validation Marketplace agency monthly turnover ₽1.5–2 million Typical YouTube video views 300–1,000 VK video views cited About 10,000 MVP build time with AI As little as one day MVP refinement time One to two weeks, potentially a month Twigly subscription ₽900 per month Mark Lu audience and reported project income 356,000 Twitter followers; roughly $4,000/month cited, with the exact basis unclear

Presenter and sources mentioned: Andreich (the presenter); TrustMRR; Twigly; LabelUp; Mark Lu; Reddit; Claude/Claude Code; Vercel; Timeweb; Neon; GitHub; VS Code; Cursor; Whisper Flow.