Junior programmers will disappear. A Kazakhstani from Silicon Valley explained why

Zhanibek Datbaev is known as the “legend of Kazakh computer science and programming.” While still in school, he became the first Kazakhstani to win a silver medal at the International Olympiad in Informatics (IOI). Zhanibek is a graduate of KazNU (KBTU) and Purdue University in the United States through the Bolashak program. Later, he worked as a software engineer in Silicon Valley at companies such as ipsy and Celonis, and was responsible for training infrastructure, ML infrastructure, and neural networks at Meta AI.

Zhanibek Datbaev, Chief Technology Officer at Nace.AI

On the future of the programming profession

In my opinion, programmers will not disappear, but the profession itself will undergo radical changes in the coming years. We are already seeing a shift in the approach to programming: more and more tasks are being handled by AI agents. Many large companies—OpenAI, Anthropic, Google—are already using their own code agents trained on internal codebases. These systems deeply understand the company’s infrastructure and can write vast amounts of code.

For example, Google stated that in 2024, AI wrote about 25% of new code at Google, in 2025 it reached 50%, and now it is around 75%. The pace of change is extremely high.

However, humans are still needed for AI. People must provide good instructions, guide the process correctly, and control the results. Therefore, programmers will remain—simply working at a higher level of abstraction.

The changes occurring can be compared to the automation of production, where machines perform many processes while humans manage the system. In the past, manufacturing involved a lot of manual labor. Today, many processes are automated: robots assemble parts, while humans oversee the system. Another fitting analogy is with the military. Today, thanks to technology, one soldier can be comparable to a thousand soldiers from the 18th century. Something similar is happening now with programming and AI.

Looking back historically, in the 1970s and 80s, people wrote code almost at the level of processor instructions. Then higher-level languages appeared, and programmers no longer had to think about low-level details. Now we are moving to yet another level higher. AI allows humans to avoid writing a significant portion of code manually.

Despite the development of artificial intelligence, the market still needs specialists who can learn quickly and adapt to new tools. It is no longer enough just to write code; it is important to know how to work effectively with AI.

Therefore, it is necessary to work as much as possible with AI agents and understand their limitations. If AI is used superficially, one can fall into a trap: it seems like everything is working, but you do not understand what errors are inside. For example, a Veracode study showed that about 45% of AI-generated code contains vulnerabilities if not checked by a human. Therefore, it is important not to blindly trust AI but to build a verification system. Future developers will become more like architects and operators of AI. You sit at the helm but write code by hand only minimally.

At the same time, human expertise will remain critically important in complex systems: integrating services, designing architecture, understanding business logic, context, the consequences of business decisions, and handling critical errors.

What is happening in the industry now

I believe we are currently experiencing the third major wave of computer systems. The first wave was classic deterministic systems, where code executed exactly as the human had written it. The second wave was recommendation systems: social networks, content recommendations, ranking. The third wave is generative AI. Now systems are becoming much more dynamic. Humans may not write code directly but instead give instructions to AI, which generates most of the system.

AI is already being used not only in programming. For example, it is now actively applied in pharmaceuticals and the development of new drugs. In the past, the drug creation cycle could take about 20 years. Companies had to analyze hundreds of thousands of molecular variants, conduct massive laboratory research, and then gradually narrow down the list.

Today, technology allows this process to be accelerated dramatically. Instead of hundreds of thousands of variants, it is possible to quickly select a few thousand most promising ones and then test them further. Some companies claim to have reduced the development time for certain drugs from decades to just a few years. This is a good example of how AI is beginning to influence not only IT but also fundamental industries—medicine, science, and manufacturing.

But, of course, AI is not used only for good purposes. There have already been cases where it was used to hack systems, automatically search for vulnerabilities, and create phishing attacks. Therefore, another problem is growing in parallel—security. Although I think this is inevitable in a way. Any technology is used differently. Perhaps this will even lead to stronger security systems, as companies will be forced to adapt faster and defend themselves better.

The high pace of technological development is largely due to global competition among the largest technology companies. Everyone is trying to create the best model and capture a larger market share. Because of this, progress is happening very quickly.

For example, over the last two years, major IT companies have invested tens of billions of dollars in AI infrastructure. Microsoft continues to scale its partnership with OpenAI, Google is actively developing Gemini, and Meta is betting on open-source models.

But there is an important point: large corporations have huge amounts of internal data that will never be made public. For example, consulting companies have accumulated internal methodologies and analytics over decades. And this cannot simply be downloaded from the internet.

About the Nace.AI startup

We are building an AI-based financial audit system for the corporate sector. Our goal is to automate financial audit from the planning stage to the final report. Our agents analyze large volumes of documents, identify risks, test transactions, and generate investor reports. This allows us to speed up processes up to 20 times. We are also developing technology that allows companies to launch their own AI models within their own infrastructure.

Many companies are afraid to use large public AI services because data passes through external servers. Even if companies promise not to use data for model training, no one can give a 100% guarantee. Therefore, it is important for corporations that data remains within the organization and does not leak out. This is especially critical for auditing, finance, and working with customer information.

About hiring principles at Nace.AI

When selecting employees, we pay attention not only to technical skills but also to the personal qualities of candidates. Honesty, energy, the desire to create something new, and the ability to adapt quickly are important to us.

We have an unwritten rule: we want to hire people who raise the overall bar of the team. Over the past few months, our team has doubled in size and continues to expand.

Main forecast for the next two years in the industry

If we look at the current development trajectory, there is a high chance that in two years, almost all code will be written by AI. Today, programmers still write part of the code manually. But gradually, people will write more like specifications and tasks for AI.

About the vibe coding phenomenon

Initially, vibe coding was perceived as a tool for beginners: a person simply asks AI to write something and gets a working prototype, understanding almost nothing inside. But now it has already become part of everyday work. The main thing is to properly set up the development environment and the code verification system.