Hello, this is Sergei.
Looks like everyone from my LinkedIn feed is in Perugia this week, so instead of trying to compete with the conference stream, I thought I’d share something more practical: my personal AI setup. Maybe someone find it interesting.
I’m not a big fan of off-the-shelf tools, and even less of no-code solutions. Both usually come with compromises — missing features, awkward workflows, or limits that eventually get in the way. So over time I’ve built a local stack that actually fits how I work.
Here’s what it looks like:
Cursor
Still the best AI coding assistant I’ve used. Nothing else really comes close.
I use OpenCode when starting a project from scratch, and then switch to Composer 2 or Sonnet 4.6 for feature work, refactoring, and fine-tuning.
Under the hood, the .cursor folder has become a system of its own — rules, templates, reusable prompts, and project structures that keep development consistent.
Claude (free tier + Projects)
My go-to for asking questions, pressure-testing ideas, and writing PRDs or technical specs.
AI for Newsroom
Beyond being a product, it’s also part of my workflow. Under the hood it has content aggregation, analytics, and monitoring features that help me stay on top of AI and media developments every day.
Room Service
This is my Telegram-based operating system for work (and no, not OpenClaw).
It handles almost everything operational: updating AI for Newsroom, searching for newsroom AI initiatives across articles, emails, and BlueSky, sending weekly Spotify stats and recommendations, task and meeting tracking, invoice issuing, subscription monitoring etc.
It runs on several AI layers, distributing models by task and skill to stay both output-effective and cost-effective.
The part I like most: it’s self-learning. Every Friday it collects approved outputs and updates its own prompts and rules.
Now it's time to read more of this week's updates from 'AI For Newsroom'.