You’re raising some important questions, but I think the argument that coding skills are becoming useless is missing a key point. Knowing how to code isn’t just about writing lines of syntax. It is about understanding how software works, how to structure logic, and how to debug when things inevitably break.
For the foreseeable future, AI models will still require users who can clearly articulate what they want, interpret the results, and diagnose failures when the output is not what was expected. Coding knowledge will be crucial for that. It is one thing to say, “build me a SaaS app,” but unless you have a solid technical understanding, how will you judge whether what AI gives you is functional, efficient, or even secure? The ability to communicate with AI models effectively will be a major differentiator between those who get great results and those who end up with broken or inefficient systems they do not understand.
That said, I do think we will see a massive flood of low-effort apps. Many of these will be clones of existing ones or ultra-niche personal projects that would not have justified hiring a developer in the past. But these projects were never going to be big business anyway. The real impact is that professional software development will likely shrink, focusing more on high-complexity, high-value projects where AI still struggles. These include cutting-edge research, mission-critical infrastructure, and AI alignment itself.
The dystopian concern of AI replacing all intellectual labor is valid in the long term, but it is not happening overnight. AI-generated software still has limitations, including hallucinations, security risks, inefficiencies, and lack of deep domain expertise. Even if AI can generate software, there will still be a need for people who can direct, validate, and refine it.
The real question is not “why learn coding if AI can code?” It is “who will be best positioned to use AI coding tools effectively?” And for now, that still points to people who understand code.