A second free course: Python for data and AI

What we launched

The site’s second free course: “Python: from data to your own digest” — 55 lessons that end with a program which fetches fresh data itself, files it in a database, counts, draws charts, assembles a report page and runs on a schedule without you.

The first lessons are published. The whole text is open always, in Kazakh, Russian and English, with no account and no payment — like the Go course.

Why Python is the next course

Three reasons, and every one of them can be checked.

The first: it is the language data is analysed in today. In the TIOBE index for September 2026 Python has 17.76%, first place with twice the share of the next language; TIOBE itself warns that the index measures the popularity of searches, not the quality of a language or the amount of code written in it.

GitHub gives a sharper and more interesting picture. Between September 2024 and August 2025, 9.26 million repositories were created in Python — second place, just behind JavaScript (9.35 million) — and by contributor count Python gave up first place to TypeScript for the first time. But where AI is concerned there is nothing to argue about: nearly half of all new AI repositories are written in Python — 582 thousand in a year, up 50.7%.

The second: there is local demand for it. Check it yourself rather than taking our word: open a job search for Kazakhstan and compare two queries, “Python” and “Go”. When this announcement was written the gap was more than tenfold in Python’s favour. The numbers change daily; the order of magnitude does not. The Go course teaches building a service whole; Python adds what is asked for most often.

The third, and the main one: in data and AI, Python is the default language. Models, training, table processing, reports — it comes first here, and almost any library you meet will get on with it. It is not the only one: SQL, R, Julia and Java all exist, and each is better in its own place. But Python is the cheapest place to start, and whoever can analyse data does not depend on which tool is in fashion this year.

Why data and AI rather than the web

The web is covered: the Go course takes a reader to a blog of their own, live on the internet at their own domain, with a database, sessions, HTTPS and a service that survives a reboot. Repeating that in another language would add no skill.

Data is different. It does not go stale with a framework: a table, an average, the link between two series and an honest chart will still be needed in ten years.

And there is a third reason people say out loud now: AI writes code. It has already taken the routine, and that changes not the demand for programmers but what is asked of them. The value moves to whoever can set the task and check the result — and you can only check what you understand. A course that teaches you to count for yourself prepares exactly for that. In the last lessons a language model appears beside the numbers, and with it the code that checks what it wrote.

What the reader will count

Not abstract examples about students and their marks. Prices, the tenge and money — what touches everybody in Kazakhstan.

Here is what the reader’s own program gives them in the first lesson: the price index of 2010 is 100, and in 2025 it is 348.1. Prices grew 3.48 times, and that reads both ways: today’s thousand buys what 287 tenge bought in 2010, while what cost a thousand back then now costs 3481.

Then comes the table that makes learning to count worth it. Inflation, % per year:

The countries in the table were chosen not by their borders but by the likeness of their situation: a small open economy with a currency of its own, dependent on imports and on one or two large markets. Georgia and Armenia are neighbours within the region; Poland is the same logic in Europe.

Everybody got the shock of 2022. The world came back to 3.0%, the compared countries to 3–4%, and Kazakhstan re-accelerated to 11.4%.

Along the way the reader counts their own personal inflation from their own receipts and compares it with the official figure; finds out how much money there is per tenge of GDP and who created it; reads what the law says about issuing money and what it says about credit; trains a first model and sees where it lies.

What the course will not do

It will not name a culprit in advance. A conclusion announced before you have opened the data is somebody else’s conclusion. Here the counting is yours.

We show where both of the usual stories fail. “The outside world is to blame” does not explain the table above: the compared countries had the same shock. But “they simply printed money” does not explain it either: broad money grows 13–18% a year in Armenia, 11–17% in Georgia and 12–21% in Kazakhstan — comparable growth with inflation three times apart. So several series have to be looked at together: money, the exchange rate, tariffs, and where the new money goes.

It will not ask you to take our word. This site has already published pieces on money, banks and inflation. In the exercises the reader takes a claim from one of our articles and checks it against open data. If it does not hold, write to us and the article gets corrected.

How the lessons are built

The same way as in the Go course: on the method of Viktor Fyodorovich Shatalov, the Soviet teacher whose system let schoolchildren cover the syllabus several times faster. We consider it the best thing twentieth-century teaching produced.

The whole first, the details after. Every lesson starts with a working program: you run it understanding nothing yet, and only then take it apart. The first lesson needs nothing installed at all — fifty lines run on bare Python.

The supporting signal. At the end of the walk-through comes the map of the lesson: one picture on one screen, which you can redraw by hand.

A picture instead of a definition. A price index is not “a basket deflator” but a ruler that had a hundred divisions in 2010 and has three hundred and forty-eight today.

The right not to understand the first time. No marks and no failed tests. One required exercise per lesson — small and always doable — and two more if you want them.

The language whole, with nothing left for later

The syllabus was drawn up before the first lesson was written: a list of 78 elements of the language, each with a lesson of its own. Slices and dictionaries, exceptions and files, sets, generators, a package of your own — all of it is on the list rather than in plans for the future.

Before publication every lesson is checked against that list by machine: if a topic assigned to a lesson is not explained in its text, the lesson does not go out.

How this differs from what already exists

We are neither a state body nor a private school.

We spend no public money. The project takes no funding from the budget and takes part in no state programmes.

We take on no obligations towards you and demand none from you. No enrolment, no deadlines, no expulsion. Started, abandoned, came back six months later — that is your business.

We sell nothing at the end. The course is not a funnel to a paid tier: there is no paid tier. And no required exercise needs a paid key: the data is open, and the language model in the last lessons runs on your own machine.

Where to start

Go in order; the lessons are linked.

1. Counting for yourself: prices, the tenge and one question — a program that works with nothing installed.
2. A workplace: Python, the project’s environment and the first run — installing the language and making an environment.

Or open the course map in full — the whole path to your own digest is visible there.

What you need to start: a computer, the internet and half an hour. Nothing else.

Sources

• TIOBE: the index of programming languages
• GitHub Octoverse 2025: the languages of the year
• World Bank: the consumer price index
• Course map: “Python: from data to your own digest”