Why this and why like this 🧭
This was a for-fun build to test vibe-coding: fast iterations guided by feel, where validating ideas quickly matters more than pristine architecture. I kept it tidy though: FSD (Feature-Sliced Design) to structure the codebase and i18n to mirror RU/EN content. The result is a tiny, shippable MVP that can grow without tears.
What’s inside 🧩
- Astro + Tailwind — static rendering, tokenized theming, tiny bundles.
- Content — Markdown + Decap CMS (GitHub backend, OAuth): edits are commits.
- i18n — mirrored collections:
posts/ruandposts/en. - FSD — layers/features/slices reduce coupling and duplication.
- CI/CD — GitHub Actions → Docker Buildx → GHCR → VPS.
- Caddy — TLS and reverse proxy.
- A bit of agents 🤖 — for scaffolding, config stubs, and boilerplate drafts.
Key takeaways ⚡
- You can ship an MVP fast: Astro + Markdown + Decap cover blogging and pages out of the box.
- Agents smooth the learning curve for a new framework: scaffolds, migrations, draft tests.
- Caveat: agents can over-engineer — extra scripts/helpers, noisy pipelines, DRY violations. That hurts scalability and clarity. Counter it with reviews, linters, guardrails on auto-gen paths, and periodic cleanup.
- FSD + i18n enforce discipline: mirrored locales and isolated features make refactoring safer.
- Tech debt will appear anyway — schedule architectural pit stops every few iterations.
What’s next 🚧
I’m keen to explore:
- Auto-generating profile pages via a LinkedIn-style RAG pipeline 🧠📄;
- More automation: post templates, PR preview envs, auto-tagging;
- A public release with a lightweight monetization layer (pay-what-you-want / micro-subscription).
If that sounds interesting, I can polish docs, add features, and open up parts of the pipeline. ✨
Bottom line 🎯
A small project makes a great lab: I validated the stack, paired FSD+i18n, and tried vibe-coding with agents. Stay critical of auto-gen and keep architecture hygienic, and “fast & frugal” becomes “fast & resilient.” 💪