This is the first post. So instead of pretending it’s a polished essay, here’s exactly what this site is and what I’ll send you in the newsletter.

What this site is

I’m Slava. I build AI systems for a living — not in the “AI consultant” sense, but in the “1 person, 52 projects, 2 servers, 188 secrets in a vault” sense. This site is the public face of that work.

Three things you’ll find here:

  1. Recent activity on the home page — the terminal block in the hero is a real feed from my production AI systems. Hermes processes a memory recall, OpenClaw answers a Telegram message, the RAG retriever embeds a query — you see it.
  2. Writing like this — case studies on agents, memory, RAG, prompt caching, and what it actually takes to ship AI features in production.
  3. Projects — the AI work that’s running right now. Open standards (WARD, getward.org), open source (claude-code-security-hooks, Slavka memory pattern), and tooling I built because I needed it (Hermes, OpenClaw, WebPrüfer).

What the newsletter will contain

One email per week. No fluff, no AI-hype recap, no listicles.

The format I’m trying:

  • One case study — an AI system I shipped or tried to ship last week. Numbers when I have them.
  • One build-in-public note — what I’m working on, what’s stuck, what’s surprising.
  • One observation — about agents, memory, RAG, prompt caching, model routing, or the bits of AI infra nobody talks about.

If a week is too quiet, the email is shorter. Better short than padded.

Why I think this matters

Most “AI experts” on LinkedIn never ship. Most agencies selling “AI implementation” don’t run anything in production — they sell slides. I run AI agents, RAG pipelines, prompt-cached LLM chains, and self-hosted vector search on machines I pay for. The live feed at the top of this page is real.

Subscribe if you want notes from someone actually building AI in production, not someone selling the idea of it.