Detailed program

Starting April 13 → 7 weeks · Mondays & Wednesdays · live on Zoom

How a week works

Each week has two live Zoom sessions — Mondays (theory + demos) and Wednesdays (live coding). Between sessions, expect 3–5 hours of independent practice on your own project, plus an always-on WhatsApp group for questions.

I run all sessions personally — Gleb Kalinin. 20+ years in tech, ex-CPO of Ozon.travel, the past two years focused on teaching agentic workflows and coaching product managers and consultants.

Two formats, one rhythm

The lab deliberately avoids the lecture format. Most learning happens through demos and working on real participant cases together, in real time.

Monday · ~1 hour

Theory + demos

Theme of the week, concept walkthroughs, and live demonstrations. Not a lecture — closer to a guided show-and-tell where I solve a real problem on screen and you can interrupt with questions any time. Recordings always available.

Wednesday · ~2 hours

Live coding clinic

The heart of the lab. Participants bring their real work — a stuck task, a half-built workflow, a "could Claude do this?" question — and we solve it together in real time. Group pair-programming where the whole cohort learns from each person's case.

WhatsApp · 24/7

Always-on chat

I'm in the WhatsApp group daily. Screenshots, error messages, "is this the right approach?" — answered the same day. The group stays open after the cohort ends.

What we build on Wednesdays

Real tasks from past cohorts. No synthetic exercises — participants bring their own consulting, product, and operational problems.

YouTube → research pipeline

Pull subtitles via yt-dlp, extract topics, dispatch parallel subagents to research each one. Final report in Markdown, ready to share with a client.

KPI calculator for a sales team

Full Next.js app: API sync, bonus calculation, role-based dashboards. Built end-to-end across two Wednesday sessions by a non-developer.

WhatsApp / Telegram bot on Claude Agent SDK

A bot that takes voice notes, turns them into structured tasks, and writes them to Notion. Full cycle: SDK → deployment → live in your messenger.

Competitor analysis via Firecrawl

Scrape competitor sites, extract pricing and features, generate a comparison table. Runs on a weekly schedule afterwards.

Building your own skill

Package a recurring workflow (e.g. generating a client report) into a reusable Claude Code skill, with tests so it doesn't break next month.

Presentation from notes

Markdown notes → Marp slides → PDF → ElevenLabs voiceover → deploy to Vercel. One prompt, finished artifact.

Meeting transcription & analysis

Hook up Fathom: auto-extract decisions, action items, and client quotes into your CRM after every call.

Personal memory system

A file-based memory layer so your agent remembers your profile, preferences, and project context across sessions.

Custom MCP server

Wrap an internal API or database in an MCP server so any agent — yours, a teammate's — can call it. Zero-to-working in one session.

Hooks for automation

Pre/post-tool hooks: auto-commit, validation, WhatsApp notifications. Turn Claude Code from "a tool you launch" into "an environment that runs itself."

Week-by-week plan

Week 1

Mon 13.04 / Wed 15.04

Intro to Agents

Week 2

Mon 20.04 / Wed 22.04

Prompting & Context

Break Week

27.04 — 01.05

No sessions

Catch up on materials, work on your project, rest.

Week 3

Mon 04.05 / Wed 06.05

Subagents & Research

Week 4

Mon 11.05 / Wed 13.05

MCP, Skills, Hooks

Week 5

Mon 18.05 / Wed 20.05

Agent SDK & Deployment

Week 6

Mon 25.05 / Wed 27.05

Practice & Presentations

Ready?

Questions about whether the lab fits your situation? Email me or message me on LinkedIn — happy to do a 15-minute sanity-check call before you commit.

Pricing & signup