# AI Without Magic — How Machines Think and How to Make Friends with Them

> A popular science book by Dmytro Olefirenko: how artificial intelligence really works — no formulas and no “magic.” 59 chapters, comics, hands-on practice with Claude. Printed to order in English, Ukrainian and Russian.

- Dmytro Olefirenko · for everyone aged 12 and up
- just over 300 pages
- large format, 205 × 290 mm
- full-color diagrams and comics
- about 1 kg — a proper book, not an e-book 🙂
- Printed to order · in English, Ukrainian and Russian
- This page as a website: https://ai-book.dmytro-odesa.click/en/

## About the book

Twelve-year-old Veronica is preparing a project for a school contest — a survey: “How much time do my classmates spend on their phones?” Her rival Max promises to do it all “in one evening” with AI alone. Halfway through the book Veronica slips up: a source that doesn’t exist sneaks into her draft — and she learns to check. In the end she takes second place, but she is the only one who can answer the jury’s unexpected question.

Beside her is her dad, an IT engineer, whom she catches out more than once. Together they travel from the mechanical automata of the 18th century to today’s AI agents, and then learn to work with Claude: writing strong prompts, checking answers, protecting personal data, and using AI as a tutor rather than a way to cheat.

## What readers get

- **Understanding, not tricks.** How a machine learns from examples, what a neural network is, why a chatbot “guesses the next word,” and where hallucinations come from — through games and everyday examples.
- **The habit of checking.** How to catch invented facts, check sources and numbers, and not trust a “confident tone” — a skill that matters far beyond AI.
- **Practice from the first chapters.** Ready-made “Try it in Claude” prompts (all work on the free plan), “Without a computer” games, and 8 projects in the appendix.
- **Protection from scammers.** A cloned voice saying “Mom, I’m in trouble,” fake payment pages, stolen accounts, deepfakes — and a simple family code-word rule.
- **Honest learning.** How to turn AI into a tutor, not a cheating machine: academic integrity and the “would I be ashamed?” test.
- **A look at the future.** Which jobs will change, which skills AI won’t replace, and where to start learning at 12–17.

## Who it’s for

- **Teens aged 12–17.** Students who already use AI or soon will. The chapters are short, and each has a scene, a game or a comic.
- **Parents and grandparents.** The book was written for adults too. The Introduction has a “How to read together” section: a chapter an evening with roles, games instead of reading.
- **Teachers.** The “Practice” appendix has projects and lessons with Claude. Its materials may be copied freely for lessons and clubs — non-commercially, crediting the book.

## From the author

> “If you can’t explain it to a six-year-old, you don’t understand it yourself.”
> — a line often credited to Albert Einstein

To be honest, no one has found it in Einstein’s writings or letters. Checking who really said what is one of the skills this book teaches.

Explaining to a six-year-old how artificial intelligence works is genuinely hard: there are tokens, probabilities and billions of numbers. So I tried to explain it to a twelve-year-old instead — no formulas, just shells on a beach, a “guess the next word” game, comics and real conversations between Veronica and her dad.

And I wrote it so that parents and grandparents would enjoy it too: read together, argue, check — and be surprised that the “magic” actually makes sense.

— Dmytro Olefirenko

## Contents

### Part 1. What AI Is and Where It Already Is

From a morning full of “invisible AI” to a first conversation with Claude and data centers.

1. The AI You Don’t Notice
2. What Is “Intelligence,” Anyway?
3. The Turing Test and the Chinese Room
4. Weak and Strong AI
5. First Conversation with Claude
6. The Cloud: Where AI Really Lives

### Part 2. A Short History of a Big Dream

From the Mechanical Turk to AlphaGo — a collection of stand-alone stories.

7. Mechanical Marvels
8. Ada Lovelace and Charles Babbage
9. Alan Turing
10. The Summer of 1956
11. The AI Winters
12. Cybernetics, Kyiv-Style
13. Machine Against Champion
14. The Big Bang, 2012–2022

### Part 3. How a Machine Learns: From Examples to Neural Networks

Correlations, neighbors, neurons, weights and learning from mistakes.

15. Correlation and Prediction
16. Tell Me Who Your Neighbors Are
17. The Living Neuron and the Artificial One
18. The Perceptron
19. Layers and Weights
20. Learning from Mistakes
21. Data Is Food for AI
22. How a Computer Sees a Cat
23. Memorizing Versus Understanding

### Part 4. How Modern Language Models Work

Tokens, attention, three stages of training — and why AI makes things up.

24. The “Guess the Next Word” Game
25. Tokens and the Map of Meaning
26. The Transformer and Attention
27. Three Stages of Training
28. Why AI Makes Things Up
29. Sees, Hears, Draws
30. How Off-the-Shelf Services Work
31. The Open Book: RAG
32. Agents

### Part 5. The Claude Workshop

Hands-on: dialogue, the prompt recipe, files, search, projects, artifacts.

33. Getting to Know Claude
34. Dialogue
35. Three Personalities
36. The Desktop App
37. The Prompt Recipe
38. Files and Photos
39. Search and Research
40. Projects
41. Artifacts
42. Memory, Incognito, Settings
43. Claude as a Tutor

### Part 6. Other Platforms: Who Is Who

An honest look at other assistants and how to choose a tool.

44. Who Is Who: Chat Assistants and Open Models
45. Creative Tools
46. AI for Programmers
47. How to Choose a Tool

### Part 7. Ethics and Safety

Privacy, deepfakes, scammers, bias, authorship.

48. Privacy
49. Deepfakes and Disinformation
50. Beware of Scammers!
51. Bias
52. Who Is the Author?
53. The Price of Intelligence
54. Friend or Program?
55. AI Safety

### Part 8. The Future and Your Place in It

Jobs, skills, where to start — and the final project.

56. Professions That Will Change — and Ones That Will Appear
57. Skills AI Won’t Replace
58. How to Start Learning AI
59. The Final Project

Plus: the Introduction “How to Read This Book,” “About the Author,” and appendices — a 150-term glossary, answers to “Check yourself,” practice (projects and lessons with Claude) and what to read and watch.

## Excerpt: Chapter 16. Tell Me Who Your Neighbors Are

“Dad, look! Is this a mussel or a whelk? It’s half buried in the sand, and I can’t tell.”

“Then let’s ask its neighbors.”

“Shells have neighbors?”

“Everything has neighbors, if you draw it the right way.”

Dad took a stick and drew two axes in the wet sand. As always, there was a tape measure in his backpack.

“Measure everything we collected today: length and width. I’ll do the drawing.”

### Shells in the Sand

Half an hour later, a picture appeared in the sand, very much like the one you saw in Chapter 15. Across — the length of the shell; up — its width. Each shell is one dot. Dad marked the mussels with little circles and the rapa whelks with crosses.

And once again, the dots aren’t scattered at random. The mussels — long and narrow — gathered at the bottom right. The whelks — wider — at the top left. Two camps, each in its own corner.

The numbers we use to describe an object (here, length and width) are called **features**. And the name “mussel” or “whelk” is a **label**, or a class. The task of putting a new object into one of the classes — “mussel or whelk,” “spam or not,” “cat or dog” — is called **classification**.

Now for the new shell. Veronica measures it — and a question mark appears in the sand. What next? Dad draws a circle around it so that it takes in the three nearest dots. Two of them are mussels, one is a whelk. A vote: two against one. So it’s most likely a mussel.

That’s the whole method. It’s called **k-nearest neighbors (kNN)**: find the k most similar familiar examples and see which class there are more of. Here k = 3.

> **Did you know?** One of the first people to describe the neighbors method was the statistician Evelyn Fix. As a young woman, she worked as an ordinary schoolteacher.

### How Many Neighbors Should You Ask?

We choose the number k ourselves, and a lot depends on it. If k = 1, a single nearest neighbor decides everything. But what if that one was a mistake: a whelk measured wrong, or put in the wrong camp? Then the answer will be wrong too.

If k is too big — say, all forty shells — the whole beach gets a vote, and whichever class simply has more members wins. So k is chosen somewhere in the middle. And when there are two classes, people pick an odd k, so the votes can’t split evenly.

“Dad, does this method take long to learn?”

“It doesn’t learn at all. It just remembers all the shells. And it only thinks when you bring a new one.”

That’s true. In Chapter 15, we “squeezed” all the dots into a single line and then used only the line. kNN doesn’t squeeze anything: it keeps all the examples in memory and compares each new one with every one of them. For forty shells, that’s nothing; for a billion photos, it’s very slow.

The method was described in 1951 by the American statisticians Evelyn Fix and Joseph Hodges — in a report for the U.S. Air Force School of Aviation Medicine. Today it’s one of the very first things almost everyone learns when they start machine learning.

**Under the hood.** If you’ve already learned the **Pythagorean theorem**, here’s how the distance between dots is calculated: it’s the hypotenuse of a right triangle whose legs are the difference in lengths and the difference in widths. Here’s the whole of kNN in Python — nine lines:

```python
import math
shells = [(6, 3, 'mussel'), (7, 3.5, 'mussel'), (5, 2.5, 'mussel'),
          (4, 4.5, 'whelk'), (5, 5, 'whelk'), (3.5, 3.5, 'whelk')]
new = (5, 3.5)
k = 3
dist = lambda s: math.dist(new, s[:2])
nearest = sorted(shells, key=dist)[:k]
votes = [s[2] for s in nearest]
print(max(set(votes), key=votes.count))
```

The program calculates the distance from the new shell to each of the six, takes the three nearest, and prints the winning class: “mussel.”

**Later in the book:**

- “The Doctor and Similar Cases” — why AI is a helper here, not the one who makes the diagnosis
- “Where Else Neighbors Live” — recommendations, face recognition and searching documents
- “Neighbors on Paper” — a game for two or more players: graph paper, two pencils and a piece of string
- “Try it in Claude” — check kNN and check Claude at the same time

“So it’s a mussel,” said Veronica, and finally pulled the shell out of the sand. It was long, narrow, and dark.

“The neighbors didn’t lie,” Dad smiled.

“This time. Because I measured them carefully.”

*End of Chapter 16. Another 58 are in the book.*

## For parents: common questions

**Are children even allowed to use Claude?**

A personal Claude account is for ages 18 and up — those are Anthropic’s rules. The book says so plainly and suggests working from an adult’s account, side by side. So practice becomes a reason to sit together. No paid subscription is needed.

**Won’t my child start cheating with AI?**

Quite the opposite: a whole chapter teaches how to turn AI into a tutor that asks questions instead of handing over answers. The plot itself is built on the idea that “doing it all in one evening with AI” is a losing strategy.

**My child doesn’t read books at all**

The chapters are short, and each has a scene, a game or a comic. For those who don’t like long texts, the Introduction offers a short route of 14 chapters that can be read on their own. The comics can simply be flipped through.

**I don’t understand any of this myself**

The book was written for adults too. The Introduction has a “How to read together” section: a chapter an evening with roles, games instead of reading, questions for each other.

**Why is all the practice in Claude?**

So that a child learns on one tool instead of jumping between ten. Other services (ChatGPT, Gemini, Copilot and others) are reviewed honestly in a separate part. The book is independent and is not an Anthropic publication.

**AI changes every month. Won’t the book go out of date?**

Most of the book is about how AI works and about habits of thinking; that doesn’t age. Services are described as of fall 2026, and the facts and figures have been checked.

## About the author: Dmytro Olefirenko

IT engineer, independent consultant

Hi! My name is Dmytro Olefirenko, and I’m an IT engineer. My job is to make sure that the sites and apps thousands of people use every day keep working and don’t “crash.”

I started as a programmer, building sites in PHP. Then I was responsible for making sure new versions of programs came out on time and without nasty surprises. Eventually I moved to where I work now: building the “invisible kitchen” that large services run on. Today I am an independent consultant and work with companies from different countries.

I use AI every day — and very actively. For me Claude is like a programmer partner who never gets tired. But there is one rule I never break: I read and check code written by AI as carefully as my own.

> The wheel never decided where to go. That was always decided by the person. It’s the same with AI.

**Certifications:**

- AWS — Amazon’s cloud services
- CKA — Certified Kubernetes Administrator
- Claude Certified Associate — Foundations
- Claude Certified Developer — Foundations

## How to order

The book is printed to order. Choose the edition’s language and send me an email: how many copies and where to ship them. I’ll reply with the price.

- **English** (EN) — adapted translation for international readers
- **Українська** (UA) — the original
- **Русский** (RU) — edition for the Moldovan market

- Shipping via Nova Poshta: Shipped via Nova Poshta to different countries.
- Price on request: It depends on the language, the number of copies and the destination country.
- Contact for orders — on the website, “Order” section: https://ai-book.dmytro-odesa.click/en/#order

## Other languages

- [Українська](https://ai-book.dmytro-odesa.click/ua/index.md)
- [Русский](https://ai-book.dmytro-odesa.click/ru/index.md)

This book is independent: Anthropic did not commission, review, or endorse it. Claude and Anthropic are trademarks of Anthropic, PBC.

© Dmytro Olefirenko, 2026
