20 000 000 ₽ на ВАЙБКОДИНГЕ. Интервью с Максом Платоновым

Summary

Game developer Max Platonov describes leaving game development to build Aidentika, an AI-powered service for creating product images and listing graphics for online marketplaces. The interview also covers how he used “vibe coding” to build and operate the product, attract customers, and manage its rapid growth.

Aidentika: Features and Early Feedback

• Sellers upload a basic product photo—such as a phone picture or an image on a white background—and the service generates polished, studio-style product images and infographics with text and selling points.
• The aim is to make the process simple: upload, generate, and receive results without configuring AI tools or writing prompts. More detailed settings are available for advanced users.
• Platonov says a newer premium generation mode produced results impressive enough to change the minds of designer friends who had initially dismissed the idea.
• Early feedback on his first, game-focused image-generation bot was that it was useful but too expensive, and that users could generate images themselves. That pushed him toward a simpler product for marketplace sellers, including his mother, who wanted professional-looking photos of her cosmetics.
• A recurring limitation is that image models can distort products or misrepresent newer items. Aidentika compensates users with internal credits (“sparks”) when generations fail, but cannot fully control the underlying models.

Development and Business Results

• Platonov says he built an initial prototype of the service in one night, then spent several weeks adding the landing page, admin tools, analytics, moderation, and other features.
• He initially handled support himself and aimed to respond within minutes. As users arrived, he hired a support assistant, paid a collaborator to make short-form videos, and later brought in another vibe coder to expand functionality and help resolve technical issues.
• Short videos were central to customer acquisition. A casual video showing the product reportedly reached about 300,000 views; Platonov recalls seeing 5,000 rubles in sales the next morning. He says one video format later became widely copied and began to wear out, making new marketing ideas important.
• He reports 20,635,000 rubles in payments over roughly 90 days. This is a gross payment figure, not profit: he had not completed a full cost calculation and said he had spent more than $100,000 on AI generation. At one point, he recalls daily revenue reaching around 300,000 rubles.
• He also cites roughly 610,000 image generations on the service, excluding video generations. The operation requires ongoing work on storage, processing, queues, server capacity, and reliability.
• A server incident caused data loss because backups were on the same server. Platonov says he gave affected users a free business-tier allocation to help restore trust.
• The marketing dashboard tracks video views by day, platform, creator, and post, and includes moderation and payout records. Creator compensation depends not only on views but also on engagement such as shares, likes, and comments. He says rates can range from about 6 kopecks per view to as much as 20 kopecks. One creator reportedly earned 200,000 rubles for a video.
• He sees a crowded market of easily copied AI services and argues that a product needs a specific audience, useful features, and reliable marketing—not just a quick build. He is exploring international expansion, including Spanish-speaking marketplaces, while noting the complexity of payments, company setup, and local compliance.
• Launching a service in Russia also brought obligations beyond those of publishing a game, including data handling, fiscal receipts, terms and privacy documents, and regulatory registration.

Vibe-Coding Guide and Tools

Platonov’s practical advice is to start with a small feature that directly addresses a user problem, show it to potential users, and validate the idea before investing heavily in polish.

1. Use AI as a collaborator. Describe the problem and desired outcome plainly; ask follow-up questions and request alternatives, trade-offs, and best-practice solutions.
2. Break large tasks into smaller steps. Discuss architectural options first, then give manageable tasks to AI agents and review the proposed trade-offs.
3. Plan for the real workload. Provide context about the current product and expected scale; avoid over-engineering for hypothetical volumes.
4. Test and iterate. He describes rewriting major parts of Aidentika’s storage, processing, generation, and queue systems as actual usage exposed weaknesses.
5. Validate models with experiments. He compared image models using an A/B test that let users choose between two results. Although one model was faster and cheaper, he switched back after seeing more complaints about product distortion.

Tools he mentions include:

• VS Code as his main working environment.
• Claude Code (rendered in the subtitles as “Cloud Code”) for most coding and discussion tasks; he says the leading tool can change quickly.
• Cursor as an alternative environment for trying different models.
• Grok for some frontend work, especially visual interfaces and animations.
• ChatGPT Codex for an independent review of generated code, including checks for optimization, best practices, and security issues.

He also uses agents for site monitoring and reports, and describes connecting an assistant to Google Calendar and TickTick. His view is that agents can automate repeatable tasks, but setup, access, testing, and oversight remain significant work.

Reviews, Tutorials, and Analysis

• Product feedback: Initial game-development users considered the first image-generation offer expensive; later, designer acquaintances reportedly praised the premium mode. Users also report image distortion, which remains a product limitation.
• Guide: The interview includes a hands-on discussion of vibe coding: tool selection, task decomposition, architectural planning, A/B testing, and starting with a minimal feature.
• Business analysis: The speakers discuss the importance of distribution and audience-building as AI products become easy to copy, along with the operational and compliance burden of scaling a service.

Main Speakers and Sources

• Evgeny Grishakov — interviewer and host.
• Max Platonov — guest; game developer and creator of Aidentika.