I, to be honest, do see that risk. I'll confess it openly: as a software engineer, I fear for my job and my profession in the age of advanced AI.
What does the future of software engineering look like from its Mecca, Silicon Valley? Will the programmer's profession even exist? What is worth learning today? Is it possible to prepare for an exciting but also challenging future?
Minko Gechev holds a black belt in karate… and in programming too, as a seasoned senior developer at Google. At our previous meeting, the spotlight was on his fascinating path from Troyan, a small town at the foot of the Balkan Mountains, to one of the software giants of our time; now we'll peek into some of the most exciting questions of the day. We continue and expand on our conversation on the DEV.BG podcast.
More about the guest
Minko Gechev is part of Google's team in Silicon Valley. He is one of the developers of Angular, the open-source framework that made today's hugely complex websites possible.
He completed his master's degree at Sofia University in 2015, and in 2017 he co-founded the startup Rhyme. The company offered world-class online courses, and our guest led its team of software engineers. Just a year later the project was bought by the giant Coursera—its first acquisition of the kind.
Minko has received the President's John Atanasoff Award, named after the Bulgarian-American inventor of the electronic digital computer, in two different categories. For many years he has worked on open-source software. His project Codelyzer, for static code analysis of Angular applications, has been downloaded more than 100 million times.
– Do you think we could reach the point where, instead of a software team, we have AI agents doing everything? And we just give instructions and check the results?
– I'm a bit more pessimistic on that point. I see that the CEO of Shopify, for example, wants a justification every quarter for why a role can't be taken over by AI and you need to hire a person. Jensen Huang, the CEO of NVIDIA, said that within a few years there will be no need for software engineers.
I can't imagine it being quite that extreme.
In my view, if we project what's happening right now and assume it will keep developing at the same pace, at some point we might hit a slight plateau.
It might happen, but most likely… not in the next 10 years, maybe further down the line.
I could turn out to be wrong. For now that's how I see things—as a raising of the level of abstraction, as you mentioned.
– Even that, though, is an immediate threat to many people. Hardly anyone would argue anymore that AI makes us much more efficient. I was more moderate at first, but I see how every next iteration takes over a large part of my work and my time and makes me many times more efficient. We're no longer talking about 5 to 10 percent, which is still colossal for companies like Google; we're talking about efficiency multiplied several times over.
– Yes, I felt it too. I built an app in Go, a language I hadn't written in for about ten years. And I was 10 times more productive with Cursor!
There are different theories. I know people who believe that
the web platform has lived through its greatest days. To some extent I think the same,
especially given chatbots and the way they extract, take information from websites, which also reduces people's motivation to create so much content.
Maybe we'll have different ways of sharing that information. Maybe it will become a bit more centralized, to my regret, since I like the decentralized model. The possibilities for creating software will also grow. But
I don't see a risk of programmers disappearing.
The industry may correct itself a little, but not that dramatically. The big companies keep hiring an enormous number of people.
– Yet news keeps coming in the opposite direction.
– Yes, Microsoft laid off 3 percent; they reprioritized their projects. For now they're putting even more emphasis on AI. There are investments like that at Google too.
At the same time, we don't have enough people on the team!
We need more to get more work done.
– How do you think the software engineer's profession will change? Will it exist at all? Every new stage of technological progress raises the abstraction; we're already the people who give instructions and then check, rather than write computer code. Will it stay that way for long?
– Right now I see three main ways things could develop. One is exactly what you said: you write a prompt, and from there your code is generated. There will probably be some need for human intervention at a lower level. If today we write in Java, only 1 in 100, or even 1 in 1,000 programmers will ever need to look at the actual code the compiler produced. In the same way, one day only a few may look at what AI has written.
Perhaps over time there will be more need for review, so that share won't be 1 percent of developers but maybe 15 to 20 percent. From there on, as the quality of the models improves, the share will shrink.
Another option I see is abstraction. But we already have the highest level of abstraction we can imagine. A new web application could be a composition of prompts. The individual components would be instructions, layered one onto another. That's a kind of intermediate option.
The third is to keep developing things as we do now, with a slightly smarter assistant that completes our code.
I think absolutely all the solutions along this spectrum will exist, depending on what you're building. But all of this is speculation.
– That's exactly what interests me, especially from your point of view as an engineer with experience at Google. Isn't it logical that we front-end developers will be the first to shrink in number, if not disappear? For example, I hardly write HTML and CSS anymore, and that used to be a sizable part of my work.
– We really won't be writing that much HTML and CSS; you may have to optimize them. With JavaScript, where there's logic you need to implement…
– That's often repetitive too.
– Yes, but right now AI generates invalid authentication code, for example. I've also talked with colleagues who work on one of the popular AI development environments—it often produces code that doesn't work.
At some point these things may get cleaned up. The technology may evolve, and that's an idea I'm “playing” with too. On my blog I've been working on writing a framework optimized for AI. Things may well develop that way. If you write code by hand, it has to be “fun” to write. Like Ruby, say, or the new syntaxes in JavaScript—more expressive. If AI generates the code, that's not necessary—
it can generate repetitive code that isn't so pretty, but certain constraints are placed on it so that it can avoid making mistakes.
That's something I've thought about a lot. I don't know yet what that code will look like. But I'd like to be the one to come up with it.
– Here in Bulgaria, there's a noticeable drop in demand for people in IT. Of course, there may be various reasons, but in my view one of the factors is undeniably that we're much more efficient. You're saying you don't feel anything similar in Silicon Valley.
– Part of the reason we're not hiring is that we want to invest in certain products, in certain places. We need more people on our team; we'd be even more productive if we had twice as many engineers. Even with AI! It doesn't help us that much in developing frameworks and technologies.
I talk with product teams, and they're definitely more productive with AI. I can't really share metrics here. But we would be more productive with more engineers.
– In Bulgaria it's definitely harder to land your first job, too, and at some point that could become a problem. I need to know the project to give AI its tasks, and then know Angular to see where it goes wrong. But if new people don't enter the profession, if in 5 years things develop along one of the less extreme scenarios you listed, that could turn into a problem. Because there won't be people who know the technology.
– It's true that we'll use the new tools to build applications. But I'm not worried that the software industry won't have plenty of opportunities. Technology is developing extremely fast; we'll have new ideas, new development paradigms.
But we really are at a stage where everything is changing so fast that we don't know exactly where we are or where things will go. And
that leads to some unease, to anxiety, for everyone.
I don't know either…
– How do you use generative AI in your daily life?
– Almost constantly lately! Even now, as I'm writing the blog post for Angular 20, there was a section I was missing. A colleague had written an alternative version, so I gave Gemini the blog post with the section headings, gave it my colleague's document and told it: “Fill in this part of the article based on the other document.” It did a wonderful job.
I use it for writing code too; I experiment. At Google we don't use Cursor, because you don't know where the code goes, but I experiment with it, with various development tools; I use Bolt, I build my own tools.
There are at least two ways we can benefit from generative AI as developers—by using applications that use AI, and by generating code faster to be more productive. I try to find where we can integrate web technologies more, so as to make it easier to build AI applications.
– What gives me peace of mind and balances out the rest is that I've already managed once to change my profession drastically. So… I could learn to lay tiles, although that requires more dexterity. If you had to stop programming and leave the software industry altogether, what would you do? Is it too late to become a police officer, as you told me you once dreamed of?
– It's too late; I've checked. Karate coach is an option.
– What does karate give you in programming? Do they have anything in common at all?
– Definitely! There's something I want to connect more with everything else. It's Japanese culture, where you take up your craft and keep perfecting it every day. For example, there are 12 or 13 sequences of movements—kata—that we practice in karate. You can learn them in a month if you push yourself. But people practice them their whole lives, and every day you can do them a little better, be a little better.
There's beauty in that. For some reason it reminds me of mathematics,
you have basic concepts and you keep building on them. In karate you start to feel… “If I bring gravity in here and relax my legs, use my hips, I can get more power.” Small changes, but you feel them happening; you connect with physics.
It's similar with programming. There are very few concepts you can use. You need a basic loop, basic constructs, variables, and you already have a language ready for anything.
But you can combine these things in different ways, and unexpected connections form in your brain.
You can actually model the whole world with them!
Consistent iteration and improving efficiency. That's what I find they have in common.
In Eastern systems you can take up a craft and practice it all your life. In Western ones we're more driven to increase our influence. You change your position and you're never good, because you move on to the next one.
– I find it fascinating that you of all people say this, having spent recent years working at the heart of Western thinking, Silicon Valley.
– That's true,
there's a conflict between the two value systems. I've started thinking about it lately.
You can take the view that the craft is building the capacity to increase your influence over time. There are people who write code all their lives, perfect themselves in the craft, and that's wonderful. But part of the reason I decided to return to purely software engineering work at the expense of more product-oriented work was that I missed that aspect.
There's a lot of creativity in it, and in a way you're even dealing with more familiar things. With the fundamentals.
– What makes sense to learn and develop now, so that we can be more or less confident it will still matter?
– Again, this is speculation. It seems to me that
even if AI writes the application, this engineering thinking—about composing different elements that build a more complex system—will be useful to us.
Creating abstractions properly and so on. Whatever language we use, whether we're designing buildings or writing programs…, it's a similar way of thinking.
I'm not going to design a building; it would fall down. But ideas come to me intuitively about how things could be done so that they're extensible, reusable.
– We're meeting today in Sofia. What brought you back here, and how did you find the city?
– Sofia is great!
It was pretty good when I left for Silicon Valley about 10 years ago, and now it's even better.
I'm here with my wife, and we're having a small celebration in Bulgaria with relatives for our wedding. We'll spend a month here. She's from the US and is learning Bulgarian, and I want to see friends, spend time with my family, and I hope it becomes a tradition for us.
– I really hope so! I'd be glad to continue this conversation on home soil soon!
And for anyone interested in your remarkable path, in working at Google and specifically on Angular, in advice on how to succeed if your dream is to work in Silicon Valley, and in what sets working there apart, I recommend our conversation on the DEV.BG podcast:
Remarkable stories await you—set aside an hour, you won't regret it…
