Your Own AI Agent
From a one-evening prototype to a full platform. Each participant builds their own agent and turns it into a working tool.
The problem
AI agents make headlines every week, yet how to build your own is still far from clear.
- You know the buzzwords, but it feels like “this isn’t for me — I’m not a programmer”
- Tools change every three months, so learning one specific product makes no sense
- What you need is not a market overview but a working environment of your own
Course goal
A tangible result, not a general overview
By the end, your own agent is running in Telegram: topic-based projects, voice, memory across sessions, a fallback model in case of failure, separate personal and work spaces — plus one automation for a real task of yours.
Optionally, you can also build a standalone artifact with the same method: a website, a desktop utility or an infrastructure script.
The method
The course follows a real journey, not a made-up textbook one
- Our own platform grew from an evening prototype in six months
- Each step solved a specific pain point, not an abstract textbook topic
- It was built by a non-programmer, whose role was architecture, defining tasks and accepting results
- You follow the same path and solve the same problem yourself in every session
How it actually happened
The platform’s journey: 5 steps in six months
- AprilA Telegram bot, a subscription-based model and voice — a prototype in one evening
- MayA single chat became a mess — topic-based projects replaced the one stream
- June–JulyMemory across sessions, a fallback model, a task coordinator
- AugustIsolated environments for personal use, work and other users
- SeptemberA single engine, a shared skill set, a complete platform
Session format · 3 hours
Hands-on practice with theory along the way — not the other way around
Lecture
20–25 minA principle, not a specific product.
Lab
about 2 hoursShort bursts of theory before each new step.
Failure post-mortem
10 minA real failure from our practice and the rule we drew from it.
Wrap-up
10 minEveryone demonstrates their result.
Program
9 core sessions (0–8) + advanced session 9. A single agent grows from session to session.
- 0How an agent works. A map of key terms
- 1A prototype in one evening
- 2Rules and the human’s role
- 3Projects instead of a single chat
- 4Memory and knowledge
- 5Skills for your own tasks
- 6Reliability and multiple models
- 7Separating personal and work
- 8Task management. Project defense
- 9Sandbox: any project from a spec dictated by voice · advanced
Session 9 · advanced module
The same method builds more than a personal assistant
- Telegram bots for any purpose, not just your own
- Web applications: updating existing websites and building new ones, frontend and backend
- Infrastructure projects: networking, smart home
- Standalone desktop applications — all the way to builds for the target platform that run independently of the original system
The spec is dictated by voice through the same interface — from then on only the task changes, not the method.
Results as you go
Progress is visible from the very first session
- After session 1
The bot responds and runs your first command
- After session 4
The agent remembers agreements across sessions
- After session 5
It handles a real task of yours: emails, documents, record-keeping
- After session 8
A full-fledged, fault-tolerant agent with separate environments
- After session 9
A standalone artifact outside the platform: a website, a utility or a script for your home or network
One-day corporate workshop
- Morning — a 30-minute introductory lecture and session 1
- Afternoon — sessions 2 and 3
- Wrap-up — an overview of sessions 4–8 and a demo of session 9
By the end of the day, every participant has a working prototype.
What you need
A minimal barrier to entry
- A laptop or access to a server
- An active subscription to an LLM service — most people already have one
- Willingness to work hands-on in every session, not just listen
- No coding required
The bottom line
Built by a non-programmer. Started in a single evening. The course takes you down the same path — faster, and without the bruises.
Request a program for your team
Tell us about your team and its tasks — we will send you the program and suggest a format: the full course or a one-day workshop.
- +998 99 802 18 71
+998 90 092 18 28 - info@acropolis.uz
support@acropolis.uz — technical support - TUT Workspace, 7, 1st Darkhan passage, Asaka mahalla, Mirzo-Ulugbek district, Tashkent
- Office: Mon–Fri, 10:00–18:30
Technical support: 24/7/365
Or message us on Telegram — the bot will pass your question to an engineer.