What is Shtruzel?
Shtruzel.ru is a Russian-language AI media portal I built solo. No team. No outsourced dev. The entire codebase was written in collaboration with Claude Opus 4.6 — from architecture decisions to every component and feature.
The project covers everything around AI: news, in-depth articles, guides, model reviews, tutorials, and a community feed where users can post their own content.
Why build this?
The Russian-speaking AI audience is huge and underserved. Most quality content is in English. The goal was simple: create a go-to hub for Russian developers and enthusiasts who want to follow AI without a language barrier.
The SEO strategy was built-in from day one — targeting high-volume Russian queries like how use chat gpt, "review claude code" — topics that get tens of thousands of searches monthly but have weak, low-quality competitors.
Tech Stack
Everything visible in the codebase and deployment:
Frontend: React + TypeScript
Hosting/Deploy: Vercel
AI integrations: Anthropic API (Claude), OpenAI API
Analytics: Yandex Metrica (primary Russian analytics tool)
Content: custom CMS with editorial + community moderation layers
What the site includes
Content sections:
News — fast-turnaround coverage of model releases and AI updates
Articles — analytical deep-dives
Guides — step-by-step how-tos
Reviews — model comparisons with benchmarks
Tutorials — hands-on practical lessons
Model catalog — structured database of AI models across categories: text, image, code, video, music. Each model card includes pricing, specs, ratings, and use cases.
Free tools:
Prompt generator (for ChatGPT, Claude, Midjourney)
Error translator — paste a terminal error, get a plain Russian explanation
Text rewriter — humanize AI-generated content
GPT chat — free, no VPN required (important for Russian users)
Community feed — registered users write posts, share experiments, ask questions. Everything goes through moderation before publishing.
Model comparison engine — side-by-side benchmarks, pricing, and verdicts.
How Claude Opus 4.6 built this
I didn't use Claude as a "code autocomplete." The workflow was closer to pair programming with a senior engineer who never gets tired.
Opus handled:
Full component architecture from scratch
Complex state management decisions
SEO metadata structure and sitemap logic
API integration patterns
Edge cases I didn't think to ask about — it flagged them proactively
The key insight: Opus 4.6 is good enough to hold the full context of a growing project across sessions if you structure your prompts as engineering briefs rather than one-off requests.
Early results
The site is young but the SEO foundation is already working. Articles targeting specific Russian-language AI queries are indexing and pulling organic traffic from Yandex (dominant search engine in Russia). The model catalog pages are particularly strong — structured data + thin competition = fast rankings.
Community is growing organically through Telegram (@shtruzel).
What's next
Expand the model catalog (targeting 200+ entries)
Add user ratings and reviews to model pages
Build out the comparison engine with more filtering options
More free tools (the error translator gets the most usage by far)
If you're curious about the Russian AI media space or want to dig into any part of the stack — happy to answer in the comments.