AI agent trainingBuild your
Agent OS
for real work
Practical AI agent training—from framing the right task to building a system that researches, creates, checks its work, and helps you make decisions.
Materials, news, and enrollment for the next cohort are in the AGENT OS Telegram bot · the bot is in Russian
GOALBuild a management brief and prepare decision options
Research
Build
Analysis
Review
Materials
Control
Orchestratorplans and delegates
OUTPUTBrief · risks · options · questions for the team
STARTSeptember 7, 2026Telegram onboarding
LIVE SESSIONSWed + Fri, 10:00–12:00Moscow time (UTC+3)
FORMAT4 weeks · 8 sessionspractice and support
TUITION€400₽40,000 · $500
01
An AI agent is more than another chat
AI agent training is the practical process of designing systems that receive a goal, make a plan, use tools, check intermediate results, and return a useful work product.
REGULAR CHATAnswers a questionYou do everything else
AGENT OSRuns a workflowYou control key decisions
02
One technology.
One technology.
Six ways to apply it.
You will build a working agent workflow for your role, processes, and quality standards—not a disposable course demo.
For owners and executives
Stop being the bottleneck in every process
Turn recurring management work into controlled agent workflows.
03
From a goal to an outcome
From a goal to an outcome
you can actually verify
Every workflow follows the same clear architecture.
- 01GoalDefine the result and constraints
- 02ContextProvide rules, knowledge, examples
- 03ActionsConnect files, browser, services
- 04ReviewBuild in quality criteria
- 05SystemSave a repeatable workflow
04
Four weeks to build
Four weeks to build
your own Agent OS
Two live sessions each week, hands-on work between meetings, and active support in the private group.
8live sessions
4 weeksof group support
up to 15participants
lifetimerecordings and materials
05
Do not just watch.
Do not just watch.
Build and apply.
Every part of the program produces an artifact you can use in your work.
Research agent
Collects sources, compares positions, and produces a verifiable brief.
Document and data operator
Turns files, tables, and routine documents into a structured output.
Role-specific agent
Works to your standards while leaving professional judgment with you.
Connected workflow
Uses multiple tools and asks for confirmation at critical points.
Your Agent OS
A complete agent system implemented in a real task.
06
A practical AI agent
A practical AI agent
training program
Across eight sessions, you will build your own Agent OS step by step: choose a task, define the rules, connect data and tools, launch a working product, and prepare it for adoption.
PART 1 · 2 WEEKS
System foundations
September 9–18 · sessions 1–4
Session outcome You will choose a capstone task and define exactly what the agent must deliver, where it may act independently, and where judgment stays with you.
YOU WILL LEARN
- how agents differ from chatbots and conventional automation
- the goal → plan → action → review → correction loop
- autonomy levels and human decision points
- which tasks are worth delegating—and which are not
YOU WILL BUILD
- an audit of recurring work
- a value, frequency, complexity, and risk assessment
- clear acceptance criteria for the outcome
Between sessions: collect real inputs and benchmark how long the task takes today.
Session outcome You will create a workspace where the agent understands the project, remembers the rules, and consistently works to your quality standards.
YOU WILL LEARN
- the structure of an agent workspace
- standing instructions and project memory
- how to provide context without overload
- completion criteria and review templates
YOU WILL BUILD
- the structure of your own project
- rules, constraints, and the human role
- a repeatability test across several runs
Between sessions: identify drift across three runs and strengthen the instructions.
Session outcome Your agent will conduct complex research, show the evidence behind every important conclusion, and state uncertainty honestly.
YOU WILL LEARN
- research planning and search strategies
- primary and secondary sources
- facts, assumptions, and conflicting evidence
- citations and project knowledge bases
YOU WILL BUILD
- a source set with Perplexity, Comet, and NotebookLM
- an evidence table
- a peer fact-check of the conclusions
Between sessions: verify the key claims manually and expand the knowledge base.
Session outcome You will build a repeatable pipeline that processes documents and spreadsheets, validates data, and produces the report or presentation you need.
YOU WILL LEARN
- data extraction from documents
- cleaning, normalization, and transformation
- batch file operations
- completeness and accuracy checks
YOU WILL BUILD
- a workflow using your real materials
- a defined output structure
- tests with incomplete and problematic files
Between sessions: process a real batch and create an error log.
PART 2 · 2 WEEKS
Build, test, and adopt
September 23 – October 2 · sessions 5–8
Session outcome Your agent will complete a useful action beyond chat—in a browser, database, or business service—with controlled access and approval for risky steps.
YOU WILL LEARN
- the difference between tools, APIs, and MCP
- connecting browsers, databases, and services
- least-privilege permissions
- secrets, action logs, and rollback
YOU WILL BUILD
- one integration for your project
- mandatory approval checkpoints
- logging, error handling, and a fallback path
Between sessions: run the workflow three times and confirm that it is repeatable.
Session outcome You will turn a real need into a testable website, application, bot, or internal utility.
YOU WILL LEARN
- moving from a problem to a clear specification
- defining the boundary of version one
- choosing the right product format
- testing, iteration, and release
YOU WILL BUILD
- a first version in Bolt, Lovable, Codex, Claude Code, Cursor, or Antigravity
- tests for the critical user journeys
- a published product or launch-ready build
Between sessions: run three user tests and prioritize the fixes.
Session outcome You will know when multiple agents genuinely improve speed and quality, and build a team without wasteful loops or unnecessary cost.
YOU WILL LEARN
- one agent versus several specialists
- the orchestrator, builder, researcher, and critic roles
- sequential and parallel execution
- context handoffs and stopping conditions
YOU WILL BUILD
- a role architecture for your project
- a live team run with context handoffs
- a time, cost, and quality comparison
Between sessions: compare the same scenario in single-agent and multi-agent modes.
Session outcome You will test the system against difficult scenarios, fix critical weaknesses, and leave with a clear plan for regular use after the program.
YOU WILL LEARN
- test suites and quality criteria
- common failures and false confidence
- privacy, safeguards, and fallback scenarios
- impact metrics: time, cost, and quality
YOU WILL BUILD
- a stress test for the capstone project
- a practical operating guide
- a final demo and 30-day adoption plan
After the program: adopt the system in real work and improve it using the metrics you collected.
TELEGRAM BOT · FREE
Start with the materials.
Start with the materials.
Join the course when you’re ready.
The AGENT OS bot contains useful AI agent materials, course news, and announcements. You can also enroll in the next cohort there. The bot is in Russian.
IN THE BOT
- Useful materials
- News and announcements
- Enrollment for the next cohort
07
Not theory about agents.
Not theory about agents.
Systems already at work.
Three projects demonstrate product creation, infrastructure operations, and automated content production.
01 / OPEN SOURCE
02 / INFRASTRUCTURE
03 / SEO PRODUCTION
TowerForge
A game engine by agents, for agents. AI agents took the project from idea and product design through testing and a GitHub release. Anyone can now connect an agent and build a game with it.
Open GitHubZyvero
A corporate networking service where AI agents orchestrate servers, provision new infrastructure, and continuously check service availability.
Private corporate projectVetvi
A partner marketing service where AI copywriters research topics, write search-optimized articles, and publish them to the website autonomously.
Open the service08
Learn in the same mode
Learn in the same mode
you will use at work
Short explanations lead directly to action. The value comes from a repeatable learning loop.
your
task
task
09
A practitioner at the intersection
A practitioner at the intersection
of technology, products, and business
“My job is to teach you to see the architecture of work: what to delegate, which tools to provide, where to review, and how to turn a good result into a system.”
Ivan Lindberg is a serial entrepreneur, startup mentor, and digital transformation expert. He builds technology products and applies an AI-first approach to product, research, and operations.
Skolkovo tracker and business mentorNTI trusted expertDigital transformation expertInternational speaker
10
One cohort.
One cohort.
Full immersion.
Your tuition includes all eight live sessions, practical work, feedback, Telegram support, recordings, and materials.
- four weeks of training and support
- your own capstone project
- lifetime recordings and materials
- up to 15 participants
PROGRAM TUITION
€400or ₽40,000 / $500
PAYMENT IN TWO PARTS€200 × 2or ₽20,000 × 2 / $250 × 2
No extra cost: the first payment is due before the start and the second by September 21.
Payment is handled through Tribute. Once payment is confirmed, Ivan will manually add you to the private Telegram group.
Join the cohort Not sure yet? Get a free consultationAsk a question on TelegramThird-party AI subscriptions are paid separately.
IN PARTNERSHIP WITH
Learn and launch projects
Learn and launch projects
with Yandex Cloud.
AGENT OS is delivered in partnership with Yandex Cloud. After the program, you can keep developing your project in cloud infrastructure.
FOR PROGRAM GRADUATES
up to ₽50,000
in cloud grants for your project
Participants who complete the program and meet the partner’s requirements may receive a grant. The decision and terms are determined by the partner. The live program and Telegram bot are in Russian.
FREE 30-MINUTE CONSULTATION
Still unsure?
Still unsure?
Let’s look at your task.
In a free 30-minute consultation with Ivan, we will discuss your task, decide whether AGENT OS is a good fit, and identify the best first step. The consultation is held in Russian.
- your workflow and the point of delegation
- whether the program fits your task
- where to begin without extra tools
No. You can start without a software development background. The terminal, files, permissions, APIs, and MCP are explained through concrete tasks. Technical participants can build more advanced workflows.
Live online, every Wednesday and Friday from 10:00 to 12:00 Moscow time (UTC+3). We explain the principle, build on screen, and adapt the solution to participant projects.
The live cohort, learning materials, and Telegram support are delivered in Russian. This English page is a localized description of the program.
You will receive the recording and materials. You can ask questions about the practice in the private group throughout the four-week program.
Bolt, Lovable, Codex, Claude Code, Cursor, Perplexity, Comet, NotebookLM, Antigravity, and OpenRouter. The stack may be updated when a better tool fits a particular job.
No. We use free plans wherever possible, but some advanced capabilities may require you to purchase access from a third-party provider.
Yes. Split payment is available in two equal installments: ₽20,000, €200, or $250 each. The first payment is due before the start; the second is due by September 21, 2026. The total tuition does not change.
Participants who complete the program and meet the partner’s requirements may receive up to ₽50,000 in cloud grants for their project. This is not guaranteed for every participant: Yandex Cloud determines the terms and decision.
Once payment is confirmed, Ivan will manually add you to the private Telegram group with live-session links, materials, and support.
Yes. The public offer explains the refund process and the treatment of services already provided, in line with applicable law.
Yes. Message Ivan on Telegram to discuss the number of participants and a shared workflow your team wants to implement.
Do not start with a tool.
Start with your task.
The next cohort starts on September 7, 2026. Enrollment is limited to 15 participants.
Continue in the AGENT OS botMATERIALS AND NEWS · FREE
Not ready to join yet?
Not ready to join yet?
Stay in the loop.
Leave your email for useful AI agent materials, invitations to free webinars, and news about upcoming AGENT OS cohorts.
NO COMMITMENTUseful updates,
when you need themUpdates and events are delivered in Russian. You can unsubscribe at any time.
when you need themUpdates and events are delivered in Russian. You can unsubscribe at any time.