The shift in technology

Mikael: You're leading very important education institution in this field and you're a software developer. How do you see AI progress and it’s impact on the software development field? Do you agree that there is a significant shift happening?

Yabebal: Absolutely. In my lifetime - I'm 42 now - this is the biggest shift I've recognized. When I think about what it means for humans and look at history, I can compare it to major transitions like the scientific revolution. There was a book about how the establishment of the British Royal Society was influenced by surgeons who opened the human heart - something that was previously considered untouchable in Greek and philosophical thinking. I feel we're experiencing a very similar shift.

This shift is more cognitive, not just material. Unlike previous technological advances like computers, which were still about the relationship between materials, tools, and us, this is getting closer to the internal element - to the very sense of what makes us human. It's progress in technology, but it's also fundamentally different.

What does it mean when humans have a companion that they've created, one that can understand their unstructured language so well? I don't know what that implies. It's very radical. I expect the world to really settle after an explosion, but when it settles, how will it look? It's very hard to imagine.

Mikael: That's very clear. You've basically said what a former Microsoft system architect have told me: generative AI is a cultural thing, not just a technological change. It's about cognition, about intellect, not only about technical capacities.

This leads to my main question: do you think people will now forever be divided into "priests of AI" - those who are very deep in machine learning and neural networks - and everyone else? Recently, DeepSeek published an extremely technical paper on model improvements. They wrote code in assembler for Nvidia chips to bypass export limitations to China. This feels like only maybe 10,000 people worldwide can do this level of work. Do you think this priesthood is getting smaller?

Yabebal: I've had moments where this goes up and down. Sometimes it feels almost impossible to penetrate because there's so much fog - so many possibilities, some will come through, some won't. Some are exaggerated, some are grounded. It's very hard to distinguish what's valuable from what's not.

This fog creates an emotional cycle. Depending on what you hear and how you connect the dots, you see patterns. Some patterns seem stable and make you happy - that's when humans buy. Some make you worried - that's when humans sell.

The economy is still driven by people. Google still wants me to use their products, still wants me to subscribe. Businesses need me and compete for my attention. So it's still a cooperative process.

The ultra-specialized knowledge that makes these systems work - like DeepSeek's optimizations - will become concentrated in fewer hands, just like when we went from assembly language to higher-level programming. When you have more abstraction, fewer people need to understand the lower levels.

This trend will probably continue geopolitically. The usual players will maintain control, but they still want others to pay, so they play cooperatively to maintain the global economy. There will still be a market for those who aren't ultra-specialists, but the ultra-specialists will become richer and more powerful. This inequality in power and economy has been increasing with every technological advance.

Impact on software developers

Mikael: Traditionally, computer science people were seen as super-humans who command a magical language of code. That made it a good career path - take your child, make them a computer science major, they'll get a decent salary with zero unemployment.

But today it's different. Within developers, there are sorts. Those who can do what DeepSeek is doing are very rare. Others are very generic. And because of the technology, someone who knows how to structure things well - maybe a lawyer who can structure prompts and context - might be able to use tools like Lovable to build something functional without needing to code.

This puts extreme pressure on freshly educated developers. How do you see this?

Yabebal: Everyone is evolving, including software developers who are not sitting while others are trying to catch up in writing codes using AI. It's like a marathon where everybody is taking their advantage and running with it.

Computer scientists now have advantages they can leverage to move into other fields too. The question about how many new people will choose computer science as a career - that will definitely shift.

The junior role becomes very difficult because many won't make it out. But the senior role - once you've been playing in it and survived - gives you a certain advantage because you're very close to the technology and learning at a speed probably better than a lawyer.

For current developers, if they're lazy and not taking advantage of what they have, they'll struggle. But if you're adopting new tools, you can do a lot using your existing advantage. Developers who are awake now have a better advantage in making it in this race than a lawyer, in my opinion, because they have a much better understanding of where technology is going, how to apply it, and how to put together the complexities.

10 Academy's approach

Mikael: Tell me about 10 Academy. Are you creating developers from people who weren't developers, or upskilling existing developers?

Yabebal: We're upskilling. For the 10 Academy technical training, we identify people who already have coding experience and are highly motivated. We create a launching pad in terms of opportunity. These are people who have already done their homework. Now the question is providing them a platform where they can become globally competent, able to compete and win at a global stage, not just locally.

We've always chosen cutting-edge, emerging fields - quantum computing, data engineering, AI engineering. We know that if talented people see a clear pathway from where they are to a liberated state where they don't have to worry about basic needs, they'll do whatever it takes.

There are billions of people, at least 1.5 billion here in Africa, with a lot of young talent that needs to be crafted. Even if the education system doesn't work perfectly, the sheer numbers will make some people great on their own, against all odds. We identify those people and don't allow them to get wasted.

We work with 19 countries in Africa, though most participants come from Ethiopia. We don't necessarily require computer science degrees - we just need people who know how to code substantially.

Changes in education and curriculum

Mikael: Do you think computer science curriculum needs to be completely redone for the generative AI era?

Yabebal: Because we work with graduates and students, I'm probably not qualified to speak about reforming basic education systems. The people who come to us are hungry, young, and already know the world.

The curriculum is one issue, but I don't think it's the biggest. The main problem is education that has no reward built into it. People move when there's a linearized reward system - you work hard, do stuff, then get rewarded. In the US, it's much more linearized. Here, the reward system isn't linearized - your talent doesn't pay off as much as other things.

Even without AI, just stopping at where we were before ChatGPT, computer science graduates often weren't really developing much competence. It's not just the curriculum, but a combined effect of insufficient motivation, introduction, depth, and resources.

Training approach with AI

Mikael: How does this translate into your teaching and training process?

Yabebal: It's very experimental, similar to biology. The mantra is: if AI is not working for you while you're sleeping, you're losing money. Make it work even when you're not active, because it's all about time and how much AI works for you.

Every challenge we give now has an AI component. Our tutors are now AIs. We have a section called "tutorials" that are just prompts - series of prompts that say "search this and get to that point" through GPT, DeepSeek, Gemini, or Claude. The knowledge is organized in one place.

How you leverage AI as a junior is now much more important. You can't be productive at the old pace. The speed of productivity must increase. The new junior is an AI-experienced person coupled with proficient AI use.

A good junior is proficient with AI; a not-so-good junior is one who's wrong about this. Those who aren't proficient with AI probably won't get jobs, but there's still hope because people have to start somewhere - people have to be junior anyway, so there has to be a market.

Future of software development roles

Mikael: I have a theory that junior developers or AI-savvy people could work directly within organizations - like a doctor's office hiring someone to create custom software for that specific practice, or a legal firm getting someone to build tools for their specific clients and tasks. Could this become a new profession?

Yabebal: You probably understand organizations better than me - their needs, requirements, repeatability, uncertainties, aversion to uncertainty. Current organizations are complex beasts with many elements beyond simple software. They spend a lot of money not just for immediate value, but for protection against unseen risks.

A new field will emerge because the need for purely writing software diminishes, but it becomes more coupled across departments. You need management, business knowledge, and technical skills combined. Like at Anthropic, you have philosophy majors, and at DeepSeek, many people were history majors.

There will be new components that make certain things cheap - knowledge retrieval, quick thinking - while stable, formal processes become more expensive. Software development won't disappear because it's fundamentally mathematical, but it will be augmented with other things we don't currently see.

New roles will emerge. The economy is still within humans, but there's a new player that's very fluid - it fills the entire space from thinking to product to service. The entire space has to integrate with this new component.

For newcomers, the education system must be augmented. The timescale of education is shortening because software evolution has become so fast. Ten-year programs will be suboptimal. We need short waves and long waves coupling together to form richer patterns of education.

Mikael: Thank you very much for this discussion.

Yabebal: Thank you.