Minimum stack to start development with AI

XX

When a person first starts learning AI development, there is a feeling that dozens of tools need to be installed. Articles and videos often mention a huge number of services: databases, cloud platforms, frameworks, task queues, and so on.

In practice, things are much simpler for the first projects.

To start doing real things — Telegram bots, small web services, or automation — four basic components are enough:

• AI-IDE is the place where you write code and communicate with the model
• runtime – the environment where code is executed
• Git - version control system
• deployment platform – the place where your application works

These four elements form the minimum development infrastructure.

If you think of the development process as a chain of action, it looks like this:

Шаг	Что происходит
идея	появляется задача
AI-IDE	генерируется код
runtime	код запускается
Git	сохраняется история изменений
деплой	приложение становится доступным

Each tool is responsible for its part of the process.

---

## AI-IDE: Where the main work takes place

The first and most important tool is the AI-IDE.

An IDE (Integrated Development Environment) is a program in which a developer works with code. In normal development, it’s just an editor and build tools. In AI development, IDE has a language model built in that can:

• code
• edit
• architecture
• mistake
• factor

That is, IDE becomes not just an editor, but an interactive environment for working with AI.

### Examples of AI-IDE

Инструмент	Особенность
Cursor	полноценная AI-ориентированная IDE
VS Code + AI-плагины	классический редактор с AI-интеграцией

In such an environment, the developer usually works as follows:

1. open-ended
2. formulates the model problem
3. receiver
4. edit
5. launcher

So IDE becomes the *centre of all development..

For a detailed analysis of AI editors, see the section:

→ Рабочее местоXX

---

## Runtime: Where the code runs

The next stack element is runtime.

Runtime is a program that runs your code.

When AI generates code, it’s just text. In order for a program to work, it must be executed in the right environment.

In modern AI projects, two runtimes are most often used.

### The main runtime to start

Runtime	Где используется
Node.js	веб-приложения, API, Telegram-боты
Python	автоматизация, скрипты, AI-инструменты

Both languages have a huge library ecosystem and are well supported by AI models.

For example:

Тип проекта	Часто используется
Telegram-бот	Node.js
API	Node.js
автоматизация	Python
скрипты	Python

For the first projects, just select one runtime and use it for all experiments.

For details, see the article:

Что нужно знать о runtime

---

## Git: Why You Need Version Control

The third element of the minimum stack is Git.

When development happens with AI, the code changes very quickly. Sometimes a model can:

• rewrite
• remove
• propose an experimental solution

Without a version control system, such changes can easily break a project.

Git solves this problem.

It saves the history of changes and allows you to go back to the previous version of the project.

### The main features of Git

Функция	Что делает
commit	сохраняет изменения
history	показывает историю
revert	откатывает изменения

For example, a typical workflow looks like this:

1. AI generates a new function
2. developer checks the code
3. changes are saved in Git
4. if there is a mistake, you can go back.

Git also allows you to store code on remote platforms.

### Where projects are usually stored

Платформа	Назначение
GitHub	самый популярный хостинг кода
GitLab	управление проектами
другие сервисы	альтернативные репозитории

---

## Platform for Deploy: where the application works

The last element of the minimum stack is deploy.

Deployment is the process of publishing an application on a server.

When the program only works on your computer, only you can use it. Deploy makes the application available to other users.

For example:

• Telegram bot must work on the server to receive messages
• the website should be accessible by URL
• The API must accept requests from the Internet.

For this purpose, cloud platforms are used.

### Popular Deploy platforms

Платформа	Для чего подходит
Vercel	веб-приложения
Railway	backend и API
Render	серверные приложения
VPS	полный контроль над сервером

Modern platforms allow you to deploy the application in a few minutes.

The process usually looks like this:

1. code is sent to Git
2. platform connects to the repository
3. the application is automatically collected
4. the project is run on the server.

---

## What does the minimum development infrastructure look like

If you combine all the elements, you get a simple design architecture.

Компонент	Что он делает
AI-IDE	создание и редактирование кода
runtime	выполнение программы
Git	управление изменениями
деплой	публикация проекта

All these elements are related.

The development process looks something like this:

1. the idea is formulated in AI-IDE
2. AI generates code
3. code is run through runtime
4. changes are saved in Git
5. the project unfolds on the platform of deploy.

Beginners often think that to develop you need to immediately study complex systems:

• Docker
• Kubernetes
• microservices
• task-line
• complex databases.

But for the first projects, this is redundant.

---

## Outcome

Starting with AI doesn’t require a complex infrastructure.

The minimum stack consists of four components:

Компонент	Назначение
AI-IDE	работа с кодом через AI
runtime	запуск программ
Git	контроль версий
платформа деплоя	публикация приложения

This is enough to create real projects: Telegram bots, web applications and automation.

As the project grows, you can add new tools, but to start this set already covers the entire development cycle.