Why I Am Writing This Now
I think the time has finally come for me to write about artificial intelligence.
I want to talk about AI itself, how I use it, how my understanding of it has changed and how I see what is happening in the world today.
I also want to explain why I decided to write about it now.
At first, I thought I could fit all my thoughts into a single article. But there was simply too much to say, so I decided to divide the story into several parts.
I want to describe where we were, where we are now, where we are going and, finally, where we may eventually arrive.
Until recently, I did not feel confident enough to make predictions. Too many things were changing, and I felt it was simply too early to understand the final shape or even the general direction AI would take.
Then, in 2026, something changed in my own practical experience.
To Draw a Line, You Need Two Points
To build any forecast, whether it is technological, economic or something else, you need at least two points.
The first point represents where you were.
The second point represents where you are today.
Once you have both points, you can connect them and draw a line. That line gives you a chance to estimate where the third and fourth points may appear.
With only one point, it is extremely difficult to understand where you are heading.
That was my problem for almost four years.
I had one point in my head, but I kept asking myself:
Where is the second one?
I needed something I could connect to the first point. Only then could I begin to understand what was realistic, what was unrealistic and which direction AI was actually taking.
This article therefore begins with the first point: where we were.
Then it moves to the second point: where we are now.
From there, I will try to describe a possible third point in a dedicated article, mainly focused on the IT industry, because that is the field I know best.
Finally, I want to move even further and discuss the fourth point: not only where the IT industry may go, but where humanity as a whole may eventually arrive.
That final point will be much less precise. It may describe a period that neither you nor I will live to see. Some events may happen earlier, some later, and some may take a completely different form.
But even when the exact timing is uncertain, the direction is still worth discussing.
So, I invite you to join me on this journey.
A journey into the world of curiosity.
A world where wrong is right, where the impossible becomes possible, without any limits or boundaries.
Let's fly.
The First Dot
For me, the first point appeared in 2022.
That was when I first encountered ChatGPT.
My very first reaction was something like:
"This must be a prepared answer."
Almost immediately, another thought followed:
"This cannot really work."
To be honest, I was slightly disappointed.
Not because ChatGPT existed, but because I had completely missed it.
If someone had come to me a few years earlier and said,
"Hej, we will give you unlimited resources. Your have is to build a system that can communicate naturally with people."
I would probably have answered:
"Let's look for another project))."
I simply would not have believed this one could work.
But they did it.
I understood the general idea behind predicting the next word in a sentence. The concept itself was relatively straightforward.
You take a sequence of words and try to predict which word is most likely to come next.
For example, if the sentence is:
"What is the capital of France?"
then predicting the word "Paris" makes perfect sense.
But what happens if the sentence ends with only:
"What is the capital..."
How can a machine possibly know what comes next?
It could be the capital of France, Germany, Poland or any other country.
As it turned out, it didn't need to know.
Instead of guessing the country, it could simply continue with:
"Could you clarify which country you mean?"
That was the moment that surprised me.
It turned out that the amount of human knowledge and written language available on the Internet was already large enough to train a model that could understand context, ask clarifying questions and maintain a meaningful conversation instead of generating random text.
Machine learning itself was nothing new to me.
I had first encountered it around 2013, when companies such as Google and Yandex were already using neural networks for image recognition and similar tasks.
Even back then, a system could analyse an image and recognise a car, a shirt, a particular pattern or another object.
That was already impressive.
One of the main ways humans perceive the world—vision—was already being partially automated.
A camera could look at a room and determine that it was looking at a kitchen. Based on that information, another system could decide what action to take next.
Because of that, I never considered AI itself to be something magical or completely new.
What I found difficult to believe was that language could be processed well enough to create a truly natural conversation.
Looking back, perhaps I should not have been so surprised....how I could not see that.
Life had suggested to me several times that it might work. Machine translation was improving, grammar correction tools were becoming increasingly accurate, and voice assistants already understood basic commands
Human language is incredibly complicated. Every language is different, and even the same word can have completely different meanings depending on the context.
If I simply say,
"Hello."
the next likely word may also be "Hello."
But what comes after that?
The model has to wait, understand additional context and then continue predicting one word after another.
What the team behind ChatGPT achieved was something that many people—including me—did not believe to.
Why do I say that?
Because neither Google, Microsoft nor any other major technology company introduced a conversational AI product like ChatGPT before OpenAI did.
Didn't they think about it?
I think they did.
Didn't they believe it would work?
I don't think so.
Large organizations often miss the very opportunities that are right in front of them.
Google, despite its deep expertise in machine learning, was not the first to introduce what became a completely new era of search.
Today, that is already history. The company that opened a new era of search was not Google.
My First Experience with ChatGPT
Once I started using ChatGPT, I immediately tried to turn it into a good junior developer. I will skip the experiments outside software engineering)) — such as asking it to help me redesign my kitchen...how to prepare soup etc....will stay within the professional area that I know best.
It didn't work.
Over the next couple of years, I repeated the same experiment several times, usually with intervals of about six months.
Each time, I wanted to answer the same question:
Can it take a task, understand it and deliver a reliable result?
Each time, the answer was almost the same.
Not yet. The problem was - code must be further essentially adopted. And sometimes it takes more time that you can develop with your hands.
At that stage, my conclusion was that ChatGPT was little more than a very well-indexed version of the Internet.
It understood questions much better than a traditional search engine. It could combine information from different sources, explain difficult topics and present answers in exactly the format I requested.
That alone made it extremely useful.
But it was still not a junior developer.
It helped me write code, explain libraries, generate small snippets and occasionally solve individual problems. However, I could not trust it with an entire task.
No matter how I approached it, I estimated that it saved me no more than ten percent of my programming time.
The biggest improvement appeared somewhere else.
It became an excellent learning tool.
Whenever I entered an area that was new to me, I could ask questions naturally instead of searching through dozens of websites. Instead of spending an hour collecting information from different sources, I often received a structured explanation within minutes.
That alone was already changing the way I worked.
A New Form of Search
Within roughly six months, ChatGPT had almost completely replaced Google for my everyday searches.
Today, I mainly use Google when I already know the website I want to visit or when I need to find one very specific page. In most cases, that is simply because Google is still the default search engine in my browser.
Years ago, Google used to display the approximate number of search results. You could search for something like "Titanic" and immediately see that it had found hundreds of thousands or even millions of pages. I once read that these numbers were not exact. They often looked almost random to me. Maybe they were only estimates, or maybe Google counted only part of the results.
What mattered was the psychological effect. You looked at those numbers and felt that Google had access to almost unlimited information.
After switching to ChatGPT, however, I noticed something interesting.
Whenever I occasionally returned to Google, I rarely opened anything beyond the first page of results. In many cases, even the first page was only partially relevant. It felt as though traditional search had stopped evolving years ago.
Personally, I believe search in its traditional form is slowly disappearing.
The reason is simple.
When using a search engine, you open one website and find part of the answer. Then you open another website and find another part. Every site has a different layout, different navigation, different advertising and a different writing style.
You spend time searching instead of learning.
You look for the relevant paragraph. You close pop-ups. You ignore advertisements.
You adapt to a completely different interface every few minutes.
All of that consumes mental energy. Most people simply do not want to spend that energy anymore.
A language model works differently.
It uses aggregated knowledge, tries to understand what you actually mean, gives you a direct answer and, when necessary, asks additional questions to clarify your request.
The interface is always the same.
You do not need to learn a new website every time you search for information.
That was the moment I stopped thinking about ChatGPT as a chatbot.
I started thinking about it as an entirely new way of searching, organising and accessing information.
That was my conclusion during the first stage.
I still could not build a reliable junior developer from it.
It still saved me only a relatively small amount of programming time.
But as a tool for learning, understanding and searching for information, it had already become part of my everyday work.
That was where we were.
What Happened to Stack Overflow
Something else happened almost immediately.
When I first started using ChatGPT seriously, I had the feeling that I was witnessing a revolution.
Not the kind of revolution we saw with Angular, React or Bootstrap, where a new framework changes the way developers build applications.
This felt much bigger.
I had the impression that it could influence not only software development over the next five or ten years, but society as a whole.
To be honest, there were moments when that thought was genuinely frightening.
To understand why I think this change is so significant, let's look at Stack Overflow.
Only a few weeks after I began using ChatGPT regularly, I realised that I was no longer visiting Stack Overflow for most of the questions I had asked there for years.
At least for me, its value dropped dramatically.
I suspect many other developers experienced exactly the same thing. This is only my personal observation, not a formal analysis, but the change felt almost immediate.
Stack Overflow had existed for many years and had built one of the most valuable technical knowledge bases ever created.
Its moderation was strict.
Very strict.
Personally, I was never a fan of that approach.
I tried to ask questions there several times, but the experience usually ended with disappointment. The moderation often felt rigid and almost military. Eventually, I simply stopped trying.
At the same time, I have to admit that this strict approach produced something remarkable.
Questions became clearer.
Answers became better.
The community voted on those answers, gradually moving the most useful ones to the top.
One question could receive five, ten or even twenty different solutions, each ranked by thousands of developers.
Think about how much human knowledge, effort and time went into creating that structure.
The result was almost an ideal dataset.
Clear technical questions.
Multiple answers.
Community evaluation.
Quantitative signals showing which solutions developers considered the best.
For machine learning, it is difficult to imagine a better source of structured programming knowledge.
I cannot say exactly how much of that data was used to train modern language models, and I do not want to speculate.
What I can say is that resources such as Stack Overflow clearly helped create a world in which AI can answer programming questions with remarkable accuracy.
In a strange way, Stack Overflow may have completed one of its historical missions.
It organised programming knowledge so well that artificial intelligence could eventually learn how to present that knowledge in a much more convenient form.
For me, that is one of the most fascinating examples of technological evolution.
A platform spent more than a decade building an extraordinary knowledge base.
Then a completely new technology arrived and changed the way people accessed that same knowledge.
That demonstrates the scale of the force we are talking about.
Stack Overflow was acquired by Prosus in 2021 for approximately 1.8 billion US dollars.
Even a platform of that size and value can be seriously weakened when a fundamentally new way of solving the same problem appears.
And if it happened to Stack Overflow...
It can happen to many other companies as well.
That was another important point that shaped the way I started thinking about artificial intelligence.
The Second Dot
The year is 2026.For me, the year started quite quietly. There were no major changes, and nothing suggested that my opinion about AI was about to change.
Then I came across a news headline claiming that ChatGPT had improved by fifty percent.
I remember thinking:
"That's just marketing."
How could it possibly be fifty percent better?
If I ask:
"What is the capital of France?"
the answer is still Paris. Nothing has changed.
From my perspective, it simply answered a little better. It certainly did not feel like a fifty percent improvement.
At least, that was what I thought.
Until one particular project.
I needed to create a landing page, so I decided to try one of the AI website builders again. I had tested similar tools a couple of years earlier and had not been particularly impressed, but I thought it was worth giving them another chance.
The wizard produced a spectacular landing page from just a few sentences.
The problem was that it was far from what I actually needed.
Then I tried to adjust it.
That was where everything started falling apart.
Instead of improving the existing design, it kept generating content that gradually moved further and further away from my original idea.
My conclusion was the same as before.
These tools behaved more like very well-indexed collections of templates.
They understood approximately what you wanted and selected something similar from an existing collection, adapting it to your request.
But they did not really understand the design itself well enough to continue developing it together with you.
After struggling with it for a while, I decided not to waste any more time.
Instead, I opened ChatGPT.
We had quite a long discussion about the landing page. We talked about the structure, the sections, the layout and what I was actually trying to achieve.
Eventually, ChatGPT proposed a design that I genuinely liked.
So I asked:
"Can you generate the HTML?"
It did.
I remember thinking:
"Wow."
I realised that I could almost copy and paste the generated HTML directly into my Everest platform. Only a few adjustments were needed to connect it to dynamic content.
But then another thought came to my mind.
To be honest, I am quite lazy when it comes to repetitive work. If something can be automated, I would rather automate it and move on to more interesting problems.
So I asked another question.
"Can you generate the CSS using my own naming convention and organise it the way I usually write it?"
Again...
It did.
At that point, I could almost copy and paste the entire template into my project. I only had to replace the icons, prepare the images and connect the dynamic content.
That was the moment when I finally understood what that "50% improvement" actually meant.
It was no longer just answering questions better.
It had become genuinely useful.
The Second "Wow"
About a month later, almost by accident, I decided to try the Codex plugin in WebStorm.
That was my second "wow" moment.
For the first time, AI was no longer just answering questions or generating small pieces of code.
It was working directly inside my project.
It could modify existing files.
It could create new ones.
It understood the structure of the project.
But what impressed me the most was something else.
It started understanding the way I write code.
It followed my conventions, my project structure and, in many cases, my way of thinking.
I was genuinely impressed.
That was my second point.
Four years earlier, I had tried to achieve something that seemed impossible.
Once again...
They did it.
For the first time, I felt that I had something very close to a junior software engineer sitting next to me.
One that never gets tired.
One that never argues.
One that doesn't bring emotions into the discussion.
One that simply listens, understands the task and starts working.
At that moment, I realised that AI was no longer saving me ten or fifteen percent of my time.
In my own work, the improvement already felt closer to thirty or even fifty percent.
That was the moment when I finally understood what had changed.
Now I had my second point.
Four years earlier, I had seen only the beginning.
Now I could finally see where we were.
For the first time, I could connect those two points and draw a line.
Once you have the line, you can begin estimating its direction.
You know where it started.
You know where it is today.
You know its approximate speed.
You know what has already been achieved.
And only then can you begin making predictions.
That is exactly what the next article is about.