There is an ongoing debate about agentic AI and using generative AI to write code or autonomously edit files. It usually boils down to one anxiety-inducing question: Has coding been solved?
What that question is really asking—and what is stressing out so many people in the tech industry—is: Do we even need developers anymore?
Here is my take.
Whether a developer has memorized the exact syntax and APIs required for their environment (like the browser API for JavaScript) or they just write a plain prompt in English, Spanish, or Persian—they are still developers.
They are developing a solution. They are producing an outcome. If the project goals are reached, they are being productive.
It doesn't matter if you open up a terminal and restrict yourself to coding in Vim without any plugins, auto-completion, or syntax highlighting, pressing every single key to specify the exact characters in a file. And it doesn't matter if you say "yuck" to traditional code and only use natural human language to instruct an AI to do the dirty work.
In both scenarios, you are building software. In both scenarios, you can create amazing things—and in both scenarios, you can completely mess things up.
Even if we reach a point where we never need to memorize syntax again, we still need to understand the big picture. We need to grasp the system architecture and how different components are interconnected.
Why? Because you have to be able to judge the AI's output.
When the AI hallucinates or breaks something, the human in the loop has to help it clean up and fix its mistakes. AI frequently reaches "dead ends"—complex logic traps that, no matter how many agents you throw at the problem, simply cannot be resolved without human intervention.
At the end of the day, we still live in a world that requires human supervision. Even if a human decides not to write a single line of traditional code, they still have to use their non-artificial intelligence. They have to form thoughts, evaluate the system, and transfer those thoughts to the AI—whether through a voice-to-text prompt or by typing out instructions. The medium doesn't matter; the human cognitive effort does.
It is true that AI excels at straightforward tasks. If a project manager says, "Hey, change the background color on the landing page," you can just hand the codebase to your agentic AI. It will find the right CSS element, make the change, and you can move your Jira ticket to 'Done'. You get paid, you're happy, and your PM is happy.
But not all problems are that straightforward.
Most real-world engineering requires you to actually talk to the AI, iterate on the logic, and think deeply about the problem. You have to be intimately involved in the problem-solving process. As long as software requires navigating complex, ambiguous business logic where a human must remain in the loop, we have by no means gotten rid of the developer.
In fact, this very article is a perfect example of this dynamic. I used AI to help structure, format, and refine the wording of the post you are reading right now. But the actual raw thinking, the core opinions, and the initial draft that served as the input to clean up this article? That came entirely from non-artificial, real human involvement. The AI didn't invent my perspective; it just helped me compile it. Just like in modern coding, the AI handled the boilerplate, but the human drove the logic.
Maybe the job title will change. Maybe our daily workflow will look completely different in five years.
But thinking—specifically, non-artificial intelligent thinking—is still strictly required to solve real problems. Humans still exist in the loop, and we aren't going anywhere anytime soon.