Ethical không đồng nghĩa với “phải giữ mọi việc làm” Mục tiêu đạo đức của doanh nghiệp là tạo ra sản phẩm tốt, phục vụ khách hàng và sử dụng nguồn lực hiệu quả. Nếu AI có thể hoàn thành một phần công việc của developer với chi phí thấp hơn và chất lượng tương đương hoặc cao hơn, việc doanh nghiệp sử dụng AI không tự động trở thành phi đạo đức. AI không nhất thiết phải thay thế hoàn toàn con người Ngay cả khi AI viết code, con người vẫn có thể đảm nhiệm architecture, security, testing, product decisions và accountability. Vì vậy, “AI replaces developers” có thể được hiểu là AI thay thế một số lượng công việc lập trình, chứ không nhất thiết xóa bỏ hoàn toàn vai trò kỹ sư phần mềm.

I agree that developers do far more than type code—but that actually weakens your argument. If the valuable part of a developer’s job is understanding ambiguous requirements, system architecture, debugging, and making judgment calls, then those are precisely the skills companies can retain while automating the parts that don't require that level of expertise. And the fact that AI-generated code still needs maintenance doesn't prove AI is inefficient. Humans also generate code that needs debugging, refactoring, testing, and maintenance. The relevant question isn't whether AI produces zero mistakes; it's whether the total cost and time of producing and maintaining software is lower with AI-assisted development. Your ‘house built on sand’ analogy also assumes AI-generated code is inherently unreliable. That's not necessarily true. Companies can require automated tests, code review, security checks, monitoring, and human approval before deployment. If those safeguards make AI cheaper and faster overall, refusing to use it simply to preserve developer positions isn't more ethical—it may just make companies less competitive. So no, AI doesn't need to completely understand a chaotic product meeting or recognize a 2009 database bug for automation to be valuable. It only needs to handle enough of the developer workload reliably that fewer humans are required. The ethical responsibility is to manage that transition fairly, not to pretend technological efficiency is unethical because some jobs disappear.

But you're assuming that using AI automatically means companies will remove enough humans to make oversight meaningless. That's a management failure, not an ethical argument against replacing some developer roles. If one developer is expected to approve a thousand AI-generated pull requests they can't understand, the problem is clearly inadequate staffing, testing, and review processes—not the fact that AI was used. And you're right that an algorithm can't be held accountable. But neither can a compiler, a database, or an automated deployment system. Accountability has always belonged to the people and organizations that choose how technology is designed and deployed. A company can't escape responsibility by saying, ‘The AI did it.’ In fact, AI can make accountability more structured. Companies can require automated tests, logging, monitoring, rollback systems, security checks, and human approval for high-risk changes. If a billing system fails at 3 AM, those systems can detect and reverse the problem without requiring a developer to understand every line of generated code. So the ethical question isn't, ‘Can humans understand every line AI writes?’ It's, ‘Has the company designed a system where humans have meaningful control over the outcomes?’ If the answer is no, that's irresponsible engineering. But if the answer is yes, reducing the number of developers needed for routine work isn't inherently unethical. You're arguing against irresponsible automation. I'm arguing that responsible automation can still be ethical—even when it reduces the number of jobs required Don't confuse a failure of management with a failure of AI. If the solution to AI is simply ‘keep more developers no matter what,’ you're protecting headcount, not necessarily protecting users