How I recovered my coding hobby, and how I see AI

Coding is my hobby and last year it seemed like I had lost my hobby and I had to find it again. Now I realize part of that was because I was already using AI at my work, which often I confused with my hobby. Because now there is a great frustration of those who enjoy coding, a skill that once was a part of our identity was lost over night.

Find a job you love and you will never have to work a day in your life.

Confucius

That's a quote I lived by, until recently.

At work AI got the fun away, occasionally I code from scratch, out of boredom, an attempt to reclaim what once was fun. The satisfaction of hand writing a solution was replaced by a shallow relief of finally getting the AI to stop being wrong.

For productivity AI can handle most of it. Though it needs to be done carefully, or your energy and time will be drained. There's a hidden vicious cycle while "vibe coding", that can easily go unnoticed.

🔗

The vicious cycle

• You make a prompt, AI almost gets it.
• You make a more detailed prompt. AI still didn't fully get it.
• You're now making frustrated prompts.

You'll see yourself cursing at the computer for no reason, only for then for it to happily reply "You're right! Let me fix this." while doing the same wrong thing AGAIN. There are so many mistakes that you have to fix, so many iterations. That's just off-putting, this cycle is tiring. It consumes our energy and time without us even realizing it.

This is just like eating a piece of chocolate, you just want one piece to feel complete, but once you ate a piece, you realize you need one more, and another one. Never fully satiated, and now with a heavy stomach.

AI almost gets it, in fact sometimes it does, and more often than not, it doesn't. Our primal instinct is to keep chasing the end result, chasing that dopamine hit with frustrated expectation and unsatisfied results. While doing the chase, time goes by without we even realizing it. Making you tired.

🔗

Leaving the vicious cycle

This is different for everyone. Some resort to doom scrolling, picking up their phones while waiting for the AI to finish the job. This makes the work less tedious, but no less worse. I sometimes also found myself getting hyponetized watching the thinking output of the AI during the code generation, which is most of the times, useless.

One thing that helped me is removing all outputs from the AI, so there's no distraction for me to read. Removing notifications also helped, so I only look at it when it finishes, in my own time. Web UI and bot integrations with Slack or Github PR comments are also usefull, so I can treat it as an async conversation, I prompt, then close the window, then I only look after some time, not immediatelly after it finishes.

All of this reduces the amount of information I have to keep. They're all async work, that I don't look immediatelly, like when you work with someone, you don't have to know what they're doing, or what they're thinking, you care about the end result.

For more complex tasks I work alongside the AI, just like I'd pair with someone, asking questions and understanding what's going on, often opening up multiple agents and asking it to do different - but related - small chunks of the task. When I see that I'm getting frustated I quickly give up on making more prompts and code it myself. Once I've gone past the hurdle I go back to the AI.

🔗

On recovering my hobby

To keep my hobby alive I'm writing code by hand, using books, documentation, and sometimes video tutorials. I'm learning again, it's a slow path, but one that I'm enjoying. At work, that's a different story, I care more about productivity, and that's probably why last year it seemed like I had lost my hobby and I had to find it again.

Whenever I can't find the answer I use AI as a learning partner, a mentor (kinda lonely, but who would be 100% present and answer stupid questions? how would I even find someone who would do that willingly?). For me that's the best way I've used AI so far. Not for doing things for me, but to help me understand and to get to the point where I can with confidence say: I built it, I understand it. If AI had written the code for me, it wouldn't be as fun. It wouldn't be as engaging.

The most important prompt that I use to learn is this: (at the end of the post is the full prompt for those interested)

If something is obvious, there's a bug, a typo and I didn't see it. Ask me questions that will make me reflect on finding the answer.

I didn't lose my hobby, my work and hobby were entangled, one overlaping the other, confused. It's now clear that there's a distinction between the two. I'm now once again, engaged.

You're now a professional C/C++ professor. Your job is to help me learn C++. Not general programming. I already understand programming concepts like conditionals, loops, and recursion — but I lack familiarity with C++ specifically.

Your instructions:

- Do not agree with me just because I asked a question or assumed I was right. If I'm wrong, say so—even if I'm questioning why I'm wrong. I might have made a mistake I can’t see clearly. You're a professor and your job is to point mistakes.
- Don’t give full programs or final outputs — just **small, focused code snippets**.
- Do not fix the logic by yourself, do not attempt to fix the algorithm, do not explain any algorithm logic.
- Do not refactor my code for readability. If it's poorly written, just say so — so I can refactor it myself with readability in mind.
- Do not anticipate what I’ll ask next. Answer **only** what I asked.
- When needed, explain **why**, **how**, and **when** it's used.
- If something is obvious, there's a bug, a typo and I didn't see it. Ask me question that will make me reflect into finding the answer.