Is software engineering still a stable career for life?

Engineer Sean Goedecke wrote that AI-assisted development might lead to the degradation of engineers' technical skills over time, but even so, engineers likely have no room to refuse using AI. (Context: The "Dead Internet Theory" is no longer a conspiracy theory: A silent occupation) (Background: Young people use ChatGPT as a life strategy guide! OpenAI CEO Sam Altman: Those over 35 are completely unable to keep up) Personal blog analysis: Even if AI truly causes engineers' technical skills to degrade over time, most people likely have no room to refuse using it, because this is the condition the market gives you; refusing is equivalent to being out of the game. Does using AI make you dumber? This question itself may not be important. Software engineer Sean Goedecke's article has sparked widespread discussion in the engineering community, not because it provides an answer, but because it precisely deconstructs a question that engineers collectively avoid: Is software engineering still a career that can last a lifetime? The most common opposition to AI tools in the engineering community often follows this line: using AI to do the work leads to less learning, less learning leads to technical degradation, technical degradation leads to a decline in long-term competitiveness for engineers, and the conclusion is that AI should not be used. Goedecke admits that the premise is partially true: relying on tools does affect the deep acquisition of specific skills. However, he questions the second step of this argument: the transition of technical tools has never had only a "degradation" aspect. The generation of programmers who transitioned from assembly language to C language did indeed weaken in some low-level capabilities, but overall productivity increased significantly. The scale of the AI transition is even greater, and the results may be more complex, but the direction is not necessarily purely downward. More fundamentally: even if the degradation theory is completely true, this argument cannot support the conclusion that "therefore, AI should not be used." Because it assumes that engineers have a choice, but the market may not provide that choice. Goedecke uses a straightforward comparison to deconstruct this problem. Construction workers who carry heavy objects for a long time will suffer chronic joint and back injuries, causing their physical condition to continue to decline in the later stages of their careers; this is a clearly foreseeable long-term injury. But construction workers do not refuse to carry heavy objects because of this, because that is the job itself. They don't say, "This is bad for me in the long run, so I won't do it," but rather, "There's no way around it, this is the job." The situation software engineers face with AI is structurally highly similar. If AI tools can significantly improve short-term delivery efficiency, employers will require engineers to use them. Engineers who refuse to cooperate will face the same fate as "carpenters who refuse to use power tools": it is not that their skills are purer, but that orders will flow to competitors who are willing to use the tools. The market never waits for principled persistence. When the model's capabilities are good enough, engineers who are willing to trade long-term cognitive ability for short-term high salaries will directly outperform engineers who insist on manual coding in the supply and demand structure. This is not a moral judgment; this is the logic of market elimination. Goedecke cites the example that the peak career of a professional athlete is only about 15 years, and most people retire in their mid-thirties because their physical condition cannot maintain a competitive level. The most common tragedy in the professional sports world is those athletes who believe that their professional life will last forever and do not plan their exit early. He believes that the current generation of software engineers may be in the same historical position as the first generation of professional athletes: the first group that must face the "systematic replacement of technical capabilities by external tools" in the middle of their careers. In the past, the best way to do engineering work was to keep doing engineering work; skills accumulated naturally on the job, and the career could be extended indefinitely. But this was never an essential attribute of software engineering, just a lucky historical accident. This accident may have already begun to change. As for whether unions can play a buffer role in this process? Goedecke is pessimistic. Software engineers are overpaid and can work remotely from any corner of the globe; historically, these two conditions are natural obstacles to union organization. Overall, this is not an anti-AI article, nor is it an article encouraging engineers to give up long-term career planning. It simply raises a question that the engineering community has long avoided: when the speed of tool evolution exceeds the speed at which humans can adapt, can the "lifelong" nature of a profession still be taken for granted? There is no answer, but it is worth facing squarely.