Software engineer Sean Goedecke published a blog post that cuts straight to a question the industry has long avoided: Is software engineering still a lifelong career? The post has sparked widespread discussion, not for providing answers, but for exposing the brutal structure between market forces and individual choice.
Skill Atrophy vs. Market Reality
The common argument against AI tools follows this chain: Using AI reduces learning → learning reduction leads to skill atrophy → atrophy hurts long-term competitiveness → conclusion: don't use AI. Goedecke admits reliance on AI does affect deep skill acquisition, but he challenges the second step: Tool shifts have never been one-sided. When programmers moved from assembly to C, they lost some low-level ability but gained massive productivity. AI's shift is larger; the outcome may be more complex, but not necessarily downward.
More fundamentally, even if the atrophy argument holds, it doesn't justify rejecting AI. Because the argument assumes engineers have a choice — but the market may not grant one. Goedecke uses a blunt comparison: construction workers develop chronic joint and back damage from heavy lifting, a clear long-term harm. Yet workers don't refuse to lift. They say: "That's the job."
Software engineers face an analogous situation. If AI tools boost short-term delivery, employers will demand their use. Engineers who refuse will see projects go to competitors who embrace the tools. The market doesn't wait for principled stances. Engineers willing to trade long-term cognition for short-term salary will simply outcompete those who insist on manual coding. This isn't moral judgment; it's basic market elimination logic.
The Pro Athlete Model: A 15-Year Window
Goedecke draws on the pro athlete career: peak performance lasts about 15 years, with most forced out by their mid-30s due to physical decline. The classic tragedy is the athlete who believed the career would last forever and failed to plan an exit.
He suggests today's software engineers may be at the same historical inflection point: the first generation forced to confront systematic replacement of technical ability by external tools mid-career. In the past, the best way to do engineering was to keep doing it, skills accumulated naturally, and careers could stretch indefinitely. But Goedecke argues this was never an inherent property of software engineering — only a fortunate historical contingency, one that may now be ending.
As for unions providing a buffer, he is pessimistic: software engineers earn too much and can work remotely from anywhere — historically, both are natural barriers to unionization. The essay offers no conclusion, only a question worth confronting: When the speed of tool evolution outpaces human adaptation, can any job's 'lifelong' status still be taken for granted?