The Hard Part of Programming
The hard part of programming was never typing; it was typing backed by thinking. We have been so much embedded into AI that we forgot to think and just focused on speed. AI tools did not solve problems—they changed them. I mean, think of it: are we thinking even a bit after the generation of code?
When Claude writes thousands of lines of code in one message, all we do is paste it into another AI and ask it to debug it or find errors. Back in the days when I was learning web development, we had to manually type every line of code—which is the real coding—and every bug took hours of debugging. If you have a thousands-line code script, it feels like an eternity.
AI has completely changed our role; we are not the owner of the code, rather we are supervising it only. Think of it: when you ask AI about some specific topic, you are thinking like AI, and as the AI model is upgraded, you will be upgraded too. It is as if AI had limited our thoughts.
The Illusion of Speed
Even speed is not guaranteed when it comes to AI. If you keep on asking things, it will start hallucinating. I will tell you what happened to me last month. I was working on a WordPress plugin (it's a secret, yet to be revealed soon), and it was kind of a very, very code-heavy plugin, so I used Claude to help me. It was working good until I decided to use DeepSeek for more guidance.
That's where the loop begins.
When I pasted it into DeepSeek, it gave me a new list of bugs. And not only DeepSeek—even Gemini gave another list. And when I pasted the bugs list back into Claude, both agreed that this was right! Now, whom to trust? We don't know.
Moreover, most of the AI tools are designed to maximize user satisfaction, so they respond with "sorry," "yes," etc. Although Grok is working on it, and Claude has kind of a good model released last week—i.e., Fable 5—it was disabled within 4 days on the directions of the US Government. So, I stucked in a loop, and after a month, I am on the same place that I was a month ago.
The RLHF "Flattery Trap"
Now to my readers: AI models are trained on RLHF (Reinforcement Learning from Human Feedback). It means it will keep on agreeing with you instead of arguing, because their primary goal is to keep the user happy. Even if a dumb person asks AI about a dumb idea, it will say:
"Yes, you are a genius!" 😂😂😂
This RLHF got on my nerves. I mean, I was like, I am going towards the completion of the plugin, but in reality, I was stuck on the same stage. Now, many of you can say that if I am a developer for a long time, how did I not know how AI works? I know how AI works, but I did not test the new model. I was over-relying on a new model like Opus 4.8.
The Death of Critical Thinking
Now, if we see the pattern of where we are heading, I think it is evident that AI crosses its limits. I mean, it not only replaced jobs; it began to replace our thinking. I am too much in agreement with the fact that senior developers or coders can't be replaced by AI because they have spent years debugging without AI.
But on the other hand, junior programmers like me became lazy and began to rely on AI. As a result, that critical thinking vanished, which was necessary for programming.
Conclusion: Moving From Janitor to Director
In conclusion, I would say AI did not replace programmers. It replaced that spirit of the programmer—that thinking of the programmer that was necessary to survive. As already mentioned, if you keep relying on AI, it is as if you are thinking with the intelligence of AI, not yours.
So, I made a decision: I will not use AI for debugging anymore.
I will use it to search for ideas and make an implementation plan, and then I will direct it to write code. But the final part will be done by me—i.e., debugging. In this way, I will be the director rather than the janitor.
What do you think? Did AI replace our thinking? Drop your comment below and let's discuss!
Top comments (1)
Thanks for reading, everyone. I wrote this because I felt like I was losing my mind spending weeks copy-pasting code back and forth instead of actually programming.To my fellow devs: Have you noticed your critical thinking slipping since relying heavily on these models? Or are you finding a way to balance it without becoming a 'code janitor'? Let’s be completely honest here—leave a comment and let me know your experience.