Some Thoughts After Talking With Younger Students About AI, Work, and Choice

An Interview

Recently, several younger students from my master’s alma mater contacted me for a course assignment. They wanted to interview an alumnus who had already graduated.

We talked about many things: studying the major, finding a job, applying for graduate school, taking civil service exams, and the fast-changing AI landscape. As they described their thoughts, I could clearly feel their anxiety about the future.

Most of them are in their third year of university, right before one of the most important turning points before graduation. Some are preparing for graduate school, some are considering civil service exams, and some want to enter the job market directly. For software engineering students, this choice is already difficult. The negative stories around the IT industry in recent years have further weakened their confidence in future employment.

The 996 culture, endless overtime, and the so-called midlife crisis at 35 have indeed hurt many people in this industry for a long time. Everyone wants a better balance between work and life, but real pressure often forces people to give up their rest time.

You can choose to leave this competition, of course. But as long as the industry still offers opportunities, there will always be people who are willing to spend more time and endure more pressure. More young people keep flowing in, competition keeps rising, and eventually it becomes the involution everyone knows today.

More importantly, they now have to face another variable: AI.

If I Went Back to That Age

During the conversation, they asked me a question:

If I could go back to their age, what would I most want to do, and what would I most want not to do?

I thought about it for a while.

When we were learning programming, setting up the environment meant following tutorials step by step. Algorithm code had to be typed character by character. Bugs had to be debugged one step at a time. Very often, simply getting a project to run already took a lot of time.

Back then, it was hard to imagine that today’s developers could use AI agents, code completion, and natural language prompts to finish many repetitive tasks. Things that used to require reading documentation, searching error messages, and trying tutorials one by one can now often start from a decent answer produced by a single sentence.

So if I were really sent back to their age, facing today’s intense competition and an AI wave that feels like an opened Pandora’s box, I would probably feel lost too.

That anxiety is not childish. It is real.

The Shock Brought by AI

I started to truly experience large-scale AI applications around the end of 2022.

The first time I used ChatGPT, my feeling was very direct: it was like a teacher who never got tired. As long as you were willing to ask, it could become an encyclopedia, a teaching assistant, a writing partner, or a discussion partner that was always online.

Later, GitHub Copilot’s intelligent completion made me feel that programming had shifted from a manual transmission to something like semi-autonomous driving. Now that agent-based programming is becoming more common, the entry barrier for many development tasks has dropped even further.

I have personally experienced AI’s transformation over these years. It is impressive enough to inspire people, and powerful enough to create pressure.

But if you ask me whether today’s AI will completely replace developers, my answer is still: no.

It Is Not a Silver Bullet

The more I use AI, the more I realize that it is not a silver bullet in the development process.

It can certainly improve efficiency, but it also changes where a developer’s work goes. In the past, perhaps 80% of the time was spent writing code and 20% checking for gaps. Now, in many cases, the ratio is almost reversed. Code generation is faster, but review, judgment, verification, and correction become more important.

When you face unfamiliar code, a complex structure, or an implementation that seems runnable but may not be reliable, the real question is not “did AI write it?” The real question is “can you tell whether it is right?”

AI-assisted programming is useful, but if the operator cannot identify the quality of generated code, the risk concentrates in places that are hard to see. It may be a hidden edge case, a wrong abstraction, or a structural problem that only appears during future maintenance.

So AI lowers the entry barrier, not the ceiling of professional ability. It reduces repetitive labor, not judgment itself.

Tools Change Professions, But Do Not Necessarily Erase Them

I gave the students an example.

After CAD appeared, architects moved from hand drafting to computer-aided drafting. For people who were extremely skilled at manual drawing, many techniques accumulated over a lifetime suddenly seemed less scarce in front of the new tool.

But CAD did not eliminate architects. It freed them from a large amount of repetitive and mechanical drafting work, giving them more time to think about space, structure, materials, and design itself.

The shift from abacus to calculator was similar. Tool changes may make some older skills less valuable, but they also release new productivity.

Software development will probably go through a similar process.

Future developers may not need to handwrite every line of boilerplate code the way they used to. But they still need to understand the business, judge architecture, design interactions, review system boundaries, and take responsibility for the final result.

The stronger the tool, the more important human judgment becomes.

Facing Uncertainty

When facing an uncertain future, I cannot honestly say that I am completely calm either.

Industries change, technologies change, and opportunities change. No one can accurately predict what software engineering will look like in a few years, and no one can give young people an absolutely correct choice.

But there is one thing I am sure of: actively embracing change is more meaningful than pretending change does not exist.

The next time we stand at a crossroads, what gives us confidence may not be a fixed occupation, a fixed industry, or a skill that will never expire. It may be the ability to keep learning, adapt to change, and understand tools and problems again.

In the end, I encouraged them to pursue the future they wanted.

Graduate school, civil service exams, employment: every path has its own pressure, and every path has its own possibilities. What matters is not choosing only out of fear, and not denying your own judgment simply because of outside voices.

I hope they, and everyone else hesitating at a crossroads, can face the unknown tomorrow with hope.