Free · English & Kazakh · A QairuHub project

Everything aboutvibecodingin one place

The tools, the models, the workflow and the failures — taught hands-on, in English and Kazakh, by people building from Kazakhstan.

Try it in 10 minutes Or start the full path →

16: modules
9: hands-on labs
18+: tools compared
12: interactive drills

01 Start here

You cannot read your way to this

Everything on this site is downstream of one thing: opening a terminal and asking an agent to build something. People spend weeks reading about AI coding and learn less than they would in one honest hour of trying it and watching it fail.

So start there. Ten minutes, no card, no experience. Then come back and the rest of this will actually mean something.

Your first 10 minutes →

02 What is inside

Four things, not one course

01The path16 modules from what vibe coding actually means to running several agents at once. Story, evidence, and a drill in every one.Start the path →02The labsNine guided sessions on your own machine. Exact commands, exact prompts, and a check so you know it worked.Open the labs →03The tools and modelsEvery serious AI coding environment and model side by side, with free tiers, real weaknesses and what to pick first.Compare them →04The libraryThe documentation, research, incidents and people worth your time, sorted by what you are trying to do.Browse materials →

03 Not one product

Claude Code is where we start. It is not where this ends.

The workflow matters more than the brand. We teach it in Claude Code because it shows you the most, then show you the same loop in Cursor, Codex, Gemini CLI and Copilot — including the options that cost nothing, which matters when a card does not always work from here.

Claude Code
Cursor
Codex
Gemini CLI
Copilot
Windsurf
Cline
Aider
Kiro
Lovable
v0
Bolt

See all the tools →

04 The workflow

One loop, nine steps

Idea, context, agent, code, repo, test, deploy, logs, back to idea. Act II walks it once, slowly, building a real sign-up app you finish by the end. It is one part of the course, not the whole of it.

05 The path

Three acts. Why it matters, how the work is actually done, and what happens when one agent becomes several.

Why

The story, the evidence, the mindset

How

The 9-step cycle, one step at a time

III

Scale

From one agent to a team of them

06 One promise

We tell you what does not work, too

AI coding is genuinely useful and genuinely oversold. Both sides are in here with sources — including the controlled trial where experienced developers were 19% slower while believing they were 20% faster, and the production database an agent deleted during a code freeze. You will leave able to tell the difference.

07 How this was made

Agents built this site overnight

The research, the writing, the Kazakh, the fact-checking, the deploy — done in one night by a team of AI agents directed by one person. The final module opens the whole run: the prompts, the guardrails, the parts that went wrong. The most honest way to teach this is to show you a real one.

Read the run →