With Krasimir Kostadinov you can talk about geopolitics, astrophysics and history. And all of it in the context of technology, of the march of time, of the trends that affect the lives of every one of us.
Today's guest is the chief technology officer of Scalefocus: an engineer, architect and programmer with wide-ranging interests and deep knowledge, which he often shows off in his tech blog. In his current role, one of his direct duties is… to peer into the future. To try, thoughtfully, sensibly and from every angle, to predict where technology is taking us.
He will speak on exactly this topic on August 28, 2025, at this year's edition of DEV.BG All in One, the biggest event run by DEV.BG, Bulgaria's leading IT community platform, which traditionally brings the IT community together for important conversations (tickets are still available, you'll find them here). Digital Stories will be there too, with a stand of its own, and while we wait for the event, Krasimir and I will… peer into the future. For a sober and very different look at AI and the other big themes of the new age.
Join us for a stroll through what lies ahead!
– “The future is already here—literally.” That's how the invitation to your talk at DEV.BG All in One begins, and it's the very same sentence I use to sum up my own book, which I'll be happy to present to visitors at the stand I'll have at the event. So… how exactly is the future here?
– If you ask most people which sounds more absurd and unbelievable, quantum mechanics or the theory of relativity, they'll say the former: something both exists and doesn't exist. Yet the theory of relativity has the Andromeda paradox—
if one person is sitting and another is running past him, what is happening in the Andromeda galaxy from the perspective of the two will differ by several days.
That's the metaphor I'm talking about. According to physics, this is just as indisputable a fact as the fact that we have to calibrate the clocks on satellites to run faster than those on Earth. Only then, viewed from here while they're in orbit, do they appear to run at the same speed as earthly clocks, so that they can match the ones on Earth.
– So we mustn't miss the changes, if I understand the metaphor. Just as physics has the twin paradox, where twins age differently depending on spacetime. And just as artificial intelligence is arriving in software development to change everything…
– Yes, most people somehow intuitively remember the idea of the twins, but when you talk about relativity, there's also the present, and it too is relative for different observers. There is no absolute present, and the future is literally here; it depends on the observer.
In that sense, we can't avoid talking about the elephant in the room: AI. But the future and development of the IT sector have many more aspects, and some of them are, in my view, even more important.
The IT industry has gone through a lot of turbulence in recent years. I don't know if you remember the big fad in the US—“learn to code”—when even the president came out to explain that learning to program was the future.
\– And now Jensen Huang is saying: “Come on, there's no need to learn programming anymore—learn physics instead.”
– There are two types of people who, when they talk to you about AI,
are worth listening to, but more to hear what the topic is than to believe them that this is the future.
Figures like Jensen Huang have a personal interest in selling you AI, because they make money from it. It's only natural that a watermelon seller won't speak badly of watermelons. The same goes for Mark Zuckerberg and Satya Nadella.
The others whom I think we shouldn't trust too much on how great and important AI is are those currently trying to make money from the technology. But even if you're not one of them, when a tech leader steps out against the backdrop of everyone who has invested enormous sums and keeps explaining how AI is a transformative technology in everything,
you're obliged to sing along with the chorus if you don't want your shares to fall.
If you don't, others will think: “That's not an AI company. Look at what the competition is doing, and these people are still telling the same old stories.” And that would be fatal for the share price.
– As far as I understand, you think we're talking about unjustifiably high enthusiasm, pumped up by tech leaders. But then whom should we listen to if we want an objective assessment of AI? Most of us are certainly impressed in our daily lives by what this kind of technology can already do.
– I'm more moderate, because I've seen quite a few things and I try to analyze what AI actually is.
What's my reading of AI? There are all sorts of angles, but I look at it in terms of cause and effect. For us to have AI, a certain accumulation had to happen. Ray Kurzweil uses this approach to predict the future. Many people explain how he predicted the fall of the USSR, because his theory was that the Soviets were number one in telephones per capita, that this meant free communication, and you can't keep a dictatorship going under those conditions. In other words, he turned out to be right, but he was very far from the truth as far as the reasons were concerned.
In recent years we've seen great progress in cloud technologies, which gave us the ability to process huge volumes of data and ultimately led to AI. Without the cloud and big data, we'd have neither the computing power nor anything to train it on—nothing. Before the technologies that accumulated over the years, including Kubernetes—being able to orchestrate such huge computing workloads centrally—AI would have been impossible.
In my opinion, the problem with people who predict the future is that
99% of them say: “It will be like what we have now, only more so.”
“Now we have the AI trend, and from here on we'll just have more and more AI.”
– The thing is, over the last two years that's exactly what's been happening. We can see how big the progress is; every week a new model appears, capable of things we thought were impossible.
– But if you look at the development, it's improving by 3 percent, by 5 percent. Real development comes when you have a new architecture, for example, like transformers—a new way of doing AI. Then you can make a leap forward. Incremental improvements, more tokens for instance—to me that's just squeezing out the last drops.
Right now we don't have the conditions necessary for AI to develop dramatically. We have neither more data to train it on—everything was annotated long ago—nor any new architectures.
They're trying various amalgams of existing solutions, but the progress isn't convincing. For me, there are no preconditions for anything dramatic in the development of AI. And let's not forget something else.
All of AI is really still just a brute-force search for the right pattern.
– But each of my modest experiments shows that AI keeps crossing new boundaries. Not long ago we laughed at models for being unable to add two-digit numbers, and recently two of them performed at gold-medal level at the International Mathematical Olympiad without any special preparation.
– I'll give you a very unexpected example. Everyone is trying to cram AI in everywhere. My reading is… that it irritates me. I don't need Copilot everywhere! I open a terminal, I open Word, Teams… every product wants to use AI. Why? Just wait and see. Look, I've got a new laptop, and on the keyboard
there's a dedicated hardware Copilot button! What kind of madness is that?
As someone who has lived through several similar tech cycles, that's why I'm more moderate. Of course, it could also be the arrogance of age.
At home I have a keyboard from two or three years ago. The button in the same spot was for Cortana. A year later they discontinued that assistant, which was supposedly going to be everywhere.
To me, it's like your grandmother giving you a thousand jars of strawberries when you run a bakery and have no idea how to sell strawberries. You start putting them in everything—strawberry tarts, cakes, muffins. To me that looks like a sign that
Microsoft really doesn't know how to sell AI and is just trying to push it into everything. For me, it's a symbol of desperation, not success.
Judging by experience so far, whenever something gets its own button on Windows keyboards, Microsoft kills it within two or three years. Of course, it may not happen with Copilot—the context isn't one-to-one—but this behavior by corporations tells me clearly what's going on.
– Where do you see the applications of AI, though? Even if we're as skeptical as you, surely there are areas where it's undeniably useful. It helps me in each of my three professions…
– There's certainly AI in the future, but for me
it won't be what all the hype associates it with today.
Yes, in some respects AI is revolutionary, I wouldn't dispute that. If we look at the development of human knowledge, how it's produced and “stored”—one of the most ancient writing systems belongs to the Sumerian civilization. We invent writing, we preserve knowledge, and you no longer need to start everything from scratch or have someone tell it to you in person. You separate knowledge from the person and the moment.
The next big leap is Gutenberg's Bible—with it you can mass-produce knowledge and distribute it cheaply. Then comes the internet, and knowledge becomes entirely digital. You no longer have linear production costs—2,000 books, roughly speaking, cost twice as much as 1,000, whereas here costs hardly depend on scale. Suddenly knowledge spreads widely, and that even has social aspects: where the world is heading because of this free exchange of information.
For me, AI is the next step forward. And at the same time, it's a risk to all of that.
When you start producing content with AI, that's the end of the “internet of links.” Back in the day, when you had a website, you made it accessible and Google sent you traffic. Google is a search engine; it wants to show the best, most substantive, most cited sites, and you benefit because your site becomes accessible and you get more visits.
And now we've arrived at the “zero-click” internet, where you don't click on links. From the moment Google put an AI assistant at the very top, you almost always take your information from there without going any further. Yes, for now it's far from perfect, a hell of a lot of hacks have been demonstrated, some of which have already been blocked—wrong years, wrong calculations—but it works more and more…
– What's more, these are steps ChatGPT took long ago, cleaning up its answers along the way, and it's only a matter of time before it happens here too. And the really big question is where we're heading, when suddenly it's no longer so much social networks as the AI giants that will determine, even more invisibly than before, content and the flow of information.
– That's right. The problem being created right now is that, because of this generated content, you have no incentive to create your own. That will have a very serious impact on content online and will eventually lead to more unreliability.
There are many risks in terms of information—how people access it, produce it, consume it. For me it's literally a pivotal moment, just like Gutenberg's Bible.
Many people claim AI will take our jobs.
– And in the IT sector, we could say it's already happening…
– It's not just that. Right now there's a perfect storm of multiple factors in the IT industry—trade wars, geopolitics. There's a trend toward technological nationalism and an awareness in the world that wasn't so pronounced a while ago. States and blocs are competing fiercely in geopolitical terms on three levels.
Two of them have existed since the dawn of civilization—the economy and the army—but the digital world has become the third arena of confrontation.
Today you can't succeed as a state or bloc unless you have all three at once; they only work in combination. At the moment, the only country that has all of them is the US. Europe has an economy, but it lags seriously behind militarily, and how competitive it is digitally is highly debatable.
Everyone now realizes that being digitally weak is a risk.
Europe has the economic and technical potential to do whatever it decides to do technologically. Until now it relied mainly on American technology, simply because it was practical and because it saw the US as a partner it could count on. That's no longer quite the case. For that reason, a lot of things will be rethought technologically. The most obvious example is cloud technology—there's more and more talk of AI sovereignty, of independence. That will give rise to
a trend of re-creating services that have already been built around the world and even work successfully,
but they simply have to be European. This is an old topic for China; they too are working hard toward sovereignty in digital technologies. And it can't help but affect Bulgaria as well. Little by little, once the money for this starts pouring in, a lot of jobs will start to appear.
– I can already sense where this train of thought is going. That reports of the IT industry's death are greatly exaggerated, and it will simply change, because there will be qualitatively new, yet unavoidable and mandatory challenges…
– Yes.
AI can't take our jobs, for purely economic reasons.
Under any economic theory, if AI does it—okay, it happens on a mass scale—consumption falls. At some point there's little consumption, and that matters even to the companies offering AI, because AI doesn't consume. It can optimize things for you or open up new opportunities, but it doesn't consume.
If we have twice as much AI, we don't have twice as much consumption. Only people can consume.
If AI takes jobs on a mass scale, we go into recession, huge companies can't pay the electricity bill for their AI, and that's the end of it.
AI is tied to us. We don't realize that it functions, and its electricity gets paid for, thanks to our consumption, directly or indirectly.
Mass job loss, no matter in which sphere, is impossible, because the economy works in such a way that when consumption falls, there's no money for anyone. Including for AI, regardless of why it happened.
Wars are a precise example. Even the Hundred Years' War between England and France wasn't continuous—it was a series of wars that historians grouped together as one. You can't wage an immediate, serious war for a long time. In the same way, you can't keep cutting in one industry for long—which is what's happening in the US, and what everyone keeps waving around.
– So what is the reason, if not AI? What are the main elements of this “perfect storm” behind the layoffs in the West and the visible crisis in our industry in Bulgaria?
– On a global scale, I recommend JetBrains, which collects extremely varied data from 700,000 programmers on all sorts of topics. Over the past five or six years, including during the pandemic, the number of programmers worldwide has grown by between 700,000 and 1,100,000 a year.
Right now, between 19 and 20 million people worldwide work in software development.
Despite AI, despite everything, that number keeps growing. The US also has very detailed statistics. Over the past two years, 1.6 million programmers have been working there. In the last year the population grew by about 3 million, while their number stayed the same, meaning proportionally slightly fewer people practice the profession.
It's true, there were a lot of layoffs, but there's no decline. Those 1.6 million have simply been redistributed.
I look at the US not only because it has usable data, but because in our field almost every trend from there later arrives in Europe, and we can anticipate it.
To me, the trends look like something else.
Not like “AI is eating our jobs,” but like “we are a maturing industry.”
In other words, we're starting to become like normal industries. Which is a shock for us, because we're used to the bubble inflating with endless growth.
– And what can we expect, now that we're a maturing industry?
– Here's some very interesting statistics, again from the US. They have data going back to the 19th century on how many people one person working in farming feeds. At the start, in 1820, it was one to 2.5–3.2, and now they've reached about 180 people! Heavy mechanization, combine harvesters. And yet the number of people working in farming is growing—after 2014 there was a slight decline because of the organic food fad, but now it's rising again.
One person per 180 is enough to feed the rest! In the IT sector, there's been clean data since 2001: the US population was 282 million, with about 1 million IT professionals. The ratio to the population gradually falls and falls—from 1:289, it became 1:259 in 2010, and ten years later 1:229. Naturally, it's not as if every other person will become an IT specialist at some point. Someone has to feed us…
Over the past two years, for the first time, there's been a reverse trend. From a peak of 1:210, last year it was 1:212. Which means that in the US there's a plateau we've reached as an industry. To me, that's a normal trend. If we compare it with farming, which has existed for as long as civilization has—farming is the sign that we're no longer wandering hunters—the trend is for the number of people employed to decline.
For us, it's still growing, just not as fast as we're used to.
We haven't yet become like farmers, where every year there are proportionally fewer people and yet there's so much food. We're simply an industry that is maturing right now.
– And what lies ahead? What do you see in your crystal ball as a chief technology officer… I'll be endlessly curious to hear it at DEV.BG All in One. Besides being happy to talk with anyone interested in these topics at the conference itself, I'll take this opportunity to invite you to continue our conversation after it. Because I'm also curious about your view of Bulgaria specifically, of where the sector here is heading… Thank you for this conversation—to be continued!
