Who Will Become the Next Senior? How Generative AI Erodes the Development Pathway in Software Engineering
Through interviews with software engineers in South Korea, this study argues that Generative AI is eroding the career development pathway for juniors by absorbing entry-level tasks into senior-AI workflows, thereby depriving novices of the essential "productive struggle" needed to gain expertise and creating a perceptual asymmetry that hinders organizational correction.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the world of software engineering as a massive, bustling kitchen where the most complex dishes are cooked by master chefs. For decades, the way a young cook became a master wasn't by reading a recipe book in a library; it was by chopping onions, peeling potatoes, and scrubbing pots. These "boring" tasks were the secret training ground. They forced the young cooks to make mistakes, taste the sauce, realize it was too salty, and fix it. This process of struggling, failing, and fixing is how they learned to understand the flavors of the food.
Recently, a magical robot chef appeared. This robot, powered by Generative AI, can chop onions and peel potatoes faster than anyone else, and it never makes a mistake. It seems like a huge win for the kitchen. But here is the tricky part: if the robot does all the chopping, the young cooks never get to hold the knife. They never learn what a bad onion feels like, or how to fix a burnt sauce. They might end up with a perfect-looking dish, but they won't know how to cook it themselves. This paper explores a scary question: If the robot does all the beginner work, how will we ever train the next generation of master chefs? The researchers are worried that the "training ladder" is being pulled away, leaving young engineers stranded without a way to climb up.
The Robot That Ate the Training Wheels
This study dives into a quiet crisis happening in the software world, specifically looking at how new tools like ChatGPT and GitHub Copilot are changing the path from "junior" (newbie) to "senior" (expert) software engineer. The researchers, based in South Korea, interviewed 14 people: eight students or recent grads just about to enter the workforce, and six experienced engineers who have been working for at least six years.
They found a pattern they call "Absorption." It's like a vacuum cleaner sucking up all the small, messy, beginner-level tasks. In the past, a senior engineer would say, "Hey, you're new, go fix these tiny bugs and write this simple documentation." This was the "productive struggle"—the hard work where juniors actually learned how to think. But now, the senior engineers can just ask the AI to do it. The AI is so good at the basics that the seniors don't need to delegate the work anymore. They can do it themselves in seconds.
The result? The entry-level work that used to be the training ground for new engineers has vanished. The seniors are doing the work, but the juniors aren't getting the practice.
The "Black Box" Problem: Knowing the Answer, Not the Why
The researchers discovered that when juniors use AI to skip the struggle, they lose something crucial: the ability to recognize what doesn't work. One student explained that before AI, they would struggle with a problem, fail, and then learn from that failure. Now, the AI gives them the answer immediately. They get a perfect grade, but they feel hollow.
Imagine you are learning to ride a bike. If a magical force field holds you up so you never fall, you might look like you're riding perfectly. But the moment the force field disappears, you crash because you never learned how to balance. The paper suggests that AI is acting like that force field. Juniors are getting "fluency" in using the tool, but they aren't building the deep understanding of the code itself. They don't know what a "bad example" looks like because they've only ever seen the AI's perfect output.
One senior engineer put it bluntly: "The reason someone is called a senior is that they know how work should not be done." But if juniors never get to make the mistakes, they never learn what "not to do" looks like.
The Peer Pressure Trap
You might think, "Well, couldn't the students just choose to stop using AI and learn the hard way?" The paper says no, and this is where it gets really interesting. The researchers found that in university classrooms, using AI has become the norm. It's like a race where everyone else is on a jetpack. If you try to walk, you fall behind.
One student described the pressure: "Everyone except me was using GPT, so everyone except me was getting nearly perfect scores. I no longer had the freedom to work through assignments on my own... so in the end, I had no choice but to use it too." Even if a student wants to struggle to learn, the system (grades, deadlines, peer competition) forces them to take the shortcut. The "desirable difficulty" of learning—the struggle that makes you smarter—is being erased by the collective pressure to be fast and efficient.
The Blind Spot: Why No One Fixes It
So, why doesn't anyone stop this? Why don't the seniors say, "Wait, we need to let the juniors do the hard work"? The paper found a strange disconnect, or "perceptual asymmetry."
The seniors, who have already learned their craft the old-fashioned way, think the situation is manageable. They believe that if a junior has real-world experience, they can fill in the gaps using AI. They think, "We can wait for them to learn on the job." But they don't realize that the "on-the-job" learning opportunities have been eaten by the AI. They are looking at the problem from the top of the mountain, thinking the path is clear, while the juniors are at the bottom, realizing the path has disappeared.
Conversely, the juniors feel the pressure but can't change the system. They are stuck in a loop where they need experience to get a job, but they can't get experience because the jobs are being done by AI. The seniors don't see the trap because they've already escaped it, and the juniors are too busy trying to survive to fix the trap.
The Solution: Designing a New Playground
The paper concludes that we can't just hope individuals will figure this out. The system is broken, and it needs a fix from the top down. The researchers suggest we need to deliberately design "training zones" where AI is either banned or used in specific ways that force humans to struggle.
They point to other fields for inspiration. In aviation, even though planes are highly automated, pilots are still required to practice manual flying so they don't lose their skills. In nuclear power, operators must train on simulators to handle emergencies that automated systems usually prevent.
Similarly, software companies and universities need to create "safe spaces" for failure. This might mean:
- In Schools: Designing assignments where AI can't do the work, or grading based on the process of solving a problem, not just the final answer.
- In Workplaces: Creating specific, small tasks that are only for juniors to do, ensuring they get the chance to make mistakes and learn from them, even if it's slower than using AI.
The bottom line is that the path to becoming a senior engineer is being eroded. If we don't intentionally rebuild that path, we might end up with a generation of engineers who can talk to robots but don't know how to build anything without them. The question "Who will become the next senior?" doesn't have an easy answer, but the paper suggests that if we don't design a new way to learn, the answer might be "no one."
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.