From a talk to senior students at CAU, Seoul, 2024-11-19.

Introduction

AI has changed how software work is valued. Tasks like boilerplate code, API connections, simple unit tests, and basic debugging can now be done quickly with coding assistants. This does not mean engineering is less valuable. Instead, the value has shifted from routine tasks to good judgment.

This change is clear in hiring and mentoring. It is now easier to access explicit knowledge. What is still rare is the skill to judge, adapt, and explain technical work in real situations.

Knowledge Is Cheaper, Judgment Is Scarcer

Large language models put a lot of knowledge into an easy-to-use tool. This changes how experience is valued, but it does not make experience useless. Senior engineers still have an edge because experience is more than memory. It means knowing which solution fits, which shortcuts are risky, and which requirements are not worth following exactly.

In practice, AI often widens the gap between engineers who can check their results and those who cannot. A less experienced engineer might create code that looks right, but looking right is not the same as being truly useful.

Creativity Means Reframing and Selection

People often say that humans matter because of ‘creativity,’ but this word is used too broadly. In engineering, creativity is not about chaos, random new ideas, or personal style. It is about seeing problems in new ways, coming up with options, and choosing the best approach within limits.

AI can create different versions and sometimes even surprising ones. But it often cannot tell which version matters most in a real business or user situation. Human creativity is most important when the problem is not clearly defined. The real job is not just making options, but deciding what makes an option good in the first place.

Engineering Is Applied Economics

Software engineering is often taught as if technical elegance is the main goal. In reality, engineering is more like applied economics. Decisions must be made with limited time, money, reliability, risk, attention, and maintenance resources.

This is why AI output needs to be checked. A model might create a solution that works but is too costly, too hard to maintain, or does not fit the business needs. Good engineers ask questions that models do not: Is this worth building? Is it too complex? What hidden costs might show up later?

Context Turns Code into Consequence

Code does not exist on its own. It must follow legal, security, operational, and company rules. AI systems are getting better at finishing patterns, but they still rely on what you tell them. They usually do not know which licenses are not allowed, which privacy rules must be followed, which systems are weak, or which team habits cannot be changed unless you say so.

This is why human engineers are still responsible for what happens, not just for writing correct code. Context is what makes code ready to use in the real world. It is also why debugging, maintenance, and handling problems are still human jobs, even when some parts are automated.

What This Means for Junior Engineers

The idea that AI will just replace junior engineers is too simple. Some entry-level tasks are being automated, but companies still need to train new talent. Good engineers are not made by skipping the early stages of learning and responsibility.

What is changing is how junior engineers stand out. They cannot rely only on writing routine code. They need to show they can think through problems, test ideas, review AI output, see the results of their choices, and explain the pros and cons.

I have seen this in interviews. Candidates stand out not by writing more code, but by showing they can question a solution. The best early-career candidates spot performance issues, find extra complexity, or point out risks that were not directly mentioned. These actions show good judgment, not just skill with tools.

Conclusion

Competing with AI is not about producing more output than it can. That is not the right goal. A better approach is to build skills that matter more as AI gets cheaper: seeing problems in new ways, making trade-offs clear, and bringing context that makes code truly useful and responsible.

In this way, the future will favor engineers who make good decisions, not just those who can type fast.