The Future of Programmers in the AGI Era: Changing Roles and Career Strategy
Introduction
I wrote my first line of code in 1989.
My parents bought me a used NEC PC-88, and I taught myself to write and run programs in N88-BASIC by imitating what I saw. How many years have passed since that day?
Back then there was a magazine called Micom BASIC Magazine (affectionately known as "BASIC Maga") from Dempa Shimbunsha, which published program listings for all kinds of personal computers.
So I have now been programming, as a hobby and as a profession, for more than 35 years, and over that time the development environment and the way information circulates have changed and evolved to a degree that can only be called revolutionary.
And yet, through all of it, the one writing the code was always "me"—a flesh-and-blood human being.
That good old era is now, at last, truly coming to an end.
With the dramatic advances in LLM technology since around 2023, the programming industry has reached a major turning point.
In particular, the evolution of large language models (LLMs) such as OpenAI o3 and o1, Claude 3.5, and Gemini 2.0—and the eventual emergence of artificial general intelligence (AGI) beyond them—is predicted to fundamentally transform the roles of programmers and AI engineers.
The World Will Change Dramatically Before and After AGI
Today, we consider the future of programmers by dividing it into two eras: before and after the emergence of AGI.
This is not merely a chronological division; it should be understood as a turning point at which the relationship between technology and humans changes fundamentally.
By AGI (Artificial General Intelligence), we mean artificial intelligence that possesses flexible, human-like thinking and learning abilities, with broad problem-solving capabilities not limited to any specific domain.
Whereas today's AI is "narrow AI" specialized for particular tasks, AGI can understand context the way humans do, draw on experience to handle new situations, and solve problems creatively.
The currently booming LLM can be seen as a key milestone in AI's evolution from narrow to general intelligence.
The World Before AGI
In the world before AGI, let us first remind ourselves that AI functions as an amplifier of individual ability for specific tasks.
This is not unique to AI, of course. Engineering in general can be viewed as an "amplifier" in the sense that it produces a larger, or more valuable, output from a given input.
- In electrical engineering, we literally amplify small signals to drive speakers or enable communication.
- In software engineering, programs amplify human capability, enabling computation and processing at a scale impossible by hand.
- In mechanical engineering, levers, gears, and other mechanical elements convert small forces into large ones.
Now let's think about the output you get when working with AI tools—especially high-performance LLMs.
As you will notice once you use them, getting good work out of an LLM ultimately depends on the user's knowledge, skills, and experience; the quality of the user's input strongly influences the quality of the AI's output.
When a programmer with advanced expertise and a deep understanding of the subject uses a high-performance LLM, the output can be of extremely high quality.
On the other hand, if you use an LLM without a basic understanding of the subject, you are likely to get only superficial results of limited practical value.
Occasionally, when you type a careless, half-hearted prompt, the LLM comes back with an equally dull answer—an oddly human behavior that makes you suspect it might be testing your ability.
This suggests that the fundamental skills and expertise of programmers are becoming more important in the AI era, not less.
So far we have considered the world before AGI.
So what will the world after AGI look like?
The World After AGI
In the world after AGI, AI will be able to perform intellectual work at or above the human level autonomously.
This will dramatically transform every phase of the traditional V-model of software development. From business requirements analysis through detailed design, implementation, testing, and operations, AGI will take the lead in carrying out the development process.
Concretely, AGI will handle most of the technical work: proposing optimal approaches at the requirements stage, automatically designing system architectures, generating high-quality code, creating and running comprehensive test cases, and even monitoring and optimizing systems in operation.
What's more, in every phase it will produce output on par with seasoned professionals, while also absorbing the latest knowledge in a timely fashion and learning autonomously—so it will become difficult for humans to keep up with AI even within narrow specialties.
In this way, many aspects of programming and its surrounding work (from upstream to downstream) will be automated, and the very concept of software development may be forced to change.
However, this does not mean the human role will disappear; rather, it suggests that it will evolve toward higher-order judgment and creative problem solving.
1. Before AGI: A New Development Landscape Shaped by Collaboration with AI Tools
1.1 The Present and Future of Automated Coding
In today's software development, the arrival of coding-savvy LLMs and AI pair programmers such as GitHub Copilot and Amazon CodeWhisperer is transforming the development process.
These tools have evolved beyond simple code completion (and for someone who started with N88-BASIC, even IntelliSense in Turbo C and Visual Studio was quite a shock) to context-aware suggestions and code generation from natural language.
In fact, at our company, natural-language instructions have become the norm, and our library of prompt templates keeps growing.
They have also gained more advanced assistance features, such as automatically suggesting security patterns and recommending performance optimizations.
Moreover, because recent LLMs have very large context windows, you can feed them long stretches of code in one go and have them improve the overall structure or find bugs—taking over a substantial share of the work that used to require veteran engineers.
For these tools to deliver their full value, however, it is essential that the programmer has solid knowledge of architecture design, a security mindset, and experience with performance tuning.
LLMs make suggestions by combining existing knowledge and best practices, but the ability to judge whether a suggestion is appropriate—and to correct it when necessary—still rests with humans.
In test automation as well, LLM assistance is steadily evolving. Capabilities that streamline the entire testing process are emerging, from automatic detection of edge cases to automatic creation of integration tests and even automatic generation of performance test scenarios. These capabilities are used most effectively in the hands of programmers who understand the fundamentals of test design and bring a quality assurance perspective.
1.2 What the Shift Upstream Actually Looks Like
As LLMs take over routine coding work, the center of gravity of a programmer's job is shifting toward more creative and strategic territory. Particularly noteworthy is the deepening involvement in upstream work such as system design and architecture. Programmers today are taking on the role of understanding the essential challenges of the business and turning them into concrete technical solutions.
Some would argue that this is no longer really a "programmer" at all.
A slight digression: even before LLMs appeared, we had all heard the pitch "let it handle the tedious parts so you can focus on the high-value work" tens of thousands of times.
"With Framework X, the framework takes care of the tedious parts, so programmers can focus on the business logic that matters!"
or
"Library Y lets programmers concentrate on writing high-value code!"
If you think of AI and AI-powered code generation tools as a generalized version of those sales pitches, perhaps the concept is not so new after all.
The difference is that the impact is hundreds of times greater than any "Framework X" or "Library Y," and a world in which there is almost nothing left to program by hand is becoming a reality.
Of course, that world has not fully arrived yet. Right now we are arguably at the stage of "vibe coding"—cutting and pasting the fragments of programs an LLM produces, by hand or with the help of convenient tools like Cursor and Windsurf.
Returning to the main thread: with the arrival of AI, the people who held the role of "programmer" are now expected, for example, to make decisions about overall system design that take into account not just the technical implementation but also the business's growth strategy and future user needs. This process demands the ability to grasp requirements precisely through deep dialogue with stakeholders and elevate them into feasible technical solutions.
2. After AGI: A New Model of Human-AI Symbiosis
2.1 The Transformation Brought by Fully Automated Development
The emergence of AGI has the potential to change the very nature of software development.
As mentioned at the outset, an era may come in which the entire development process—from requirements definition through design, implementation, testing, and deployment—is carried out autonomously by AI. But this does not necessarily mean that programmers will lose their reason for being.
Rather, in the AGI era, the programmer's role is expected to be elevated to higher-order judgment and creative problem solving. More complex and nuanced decisions will be required: evaluating the validity of AI-proposed solutions, making ethical judgments, and assessing the impact on human society.
2.2 The Programmer's New Reason for Being
The most important capability required of programmers in the AGI era may well be the ability to understand the intersection of technology, business, and human society—and to bring them into harmony.
The work will shift toward evaluating the social and ethical implications of solutions proposed by AGI, verifying alignment with the company's technology strategy, judging validity from security and privacy perspectives, and above all, making strategic decisions aimed at maximizing business value.
Fundamental decisions—such as setting a system's purpose and determining the direction of development—will also remain important responsibilities that humans should bear. These decisions require not only technical knowledge but also deep insight into business and society.
3. A Career Strategy for the Future
Going forward, programmers will be expected to grow into "technology strategists" who combine technical skill with business sense.
To get there, you need to keep sharpening your fundamentals in programming and design while also cultivating the ability to understand the essence of the business, with a customer-oriented and creative perspective.
Improving human qualities such as ethical judgment and communication skills is equally essential. These soft skills play an important role in collaborating effectively with AI and in exercising leadership within a team.
3.1 What Programmers Should Do Before AGI Arrives
Build experience
Programming is, above all, about experience. Read lots of code and write lots of your own. A while back I used to recommend reading the source of famous OSS projects on GitHub; today, sparring with an LLM is also valuable. Once you gain a certain amount of experience, you tend to cling to your own "style"—but sparring with an LLM has made it easier than ever to pick up the latest knowledge and new ways of expressing things.
Develop logical thinking
Gaining programming experience builds more than knowledge—it trains logical and mathematical thinking. I believe that studying programming is a faster route to logical thinking than studying "logical thinking" itself. Once you start writing programs of a certain scale, you have no choice but to do structural design, iterating between analysis and design. In fact, the time spent on analysis and design often exceeds the time spent programming. Through this you learn to factor problems into their components, which sharpens your logical thinking. Logical thinking will remain a powerful weapon for you even after AGI.
Build finished products, release them, put them before the world, and gain experience engaging with users
If you program, you must see things through to a finished product. A single isolated feature is not enough—always build something complete. And then release it. When I started programming, there were few venues for publishing your work. My first releases went out through magazine submissions—small pieces. I have never forgotten the thrill, as a junior high school student, of submitting my work to a magazine and receiving a 10,000-yen fee for it. Today there are countless ways to release and distribute your work. Always release it. What do users want? How do they react? Having a point of contact with the world is extremely important.
3.2 Programmers Who Remain Valuable After AGI
Even in the AGI era, while the role will change, the importance of programmers will not be lost.
In particular, the value of programmers with the experience and skills described above will only increase.
Traditionally, programmers have served as "technical translators," converting business requirements and human intent into languages computers can understand (programming languages).
Their main responsibility was to understand system requirements and express them as precise code. In the AGI era, however, most of this technical translation work will be handled by AGI.
Instead, programmers will evolve into "translators of value creation" who understand users' essential needs and possibilities and elevate them into real-world solutions. Concretely, they will identify latent needs that users cannot yet clearly articulate, and work with AGI to shape them into feasible solutions.
This process requires an integrated understanding—and translation—of diverse factors: business context, social impact, technical feasibility, and human sensibility and user experience.
In other words, the role of understanding the strengths of both AI and humans and combining them optimally will unquestionably become more important than ever—and the people best suited to that role are programmers with real-world experience.
To that end, it is essential that individual programmers start now: sharpen your fundamentals, build experience, acquire skills, and learn how to amplify them with AI. Continuing to raise your own value while adapting flexibly to technological progress will be the key to building your future career.
Finally, this article presents one scenario based on current technology trends and predictions. It is important to keep revisiting your career strategy flexibly as technology advances and markets change.
Qualiteg is looking for talented people to join us!
We are actively recruiting talented engineers and researchers to work on our in-house services and AI/ML R&D. We are still a young company, and together we hope to build the most comfortable engineering culture in the world.
