Before anything, let me be clear: I’m not a “pro” programmer, nor one of those neovim wizards who type blazingly fast (not yet, anyway). I’ve coded both before and after AI became mainstream, and I’ve adapted as AI entered the daily workflow. I’ve seen the technology evolve, I remember discovering ChatGPT-3 and being genuinely impressed, and I’ve experimented with the good, the bad, and everything in between.
Disclaimer: This article reflects my personal opinions and experiences as a developer. Your mileage may vary, and I encourage everyone to experiment, question, and find their own balance with AI tools.
The Ubiquity of AI: Temptation and Opportunity
AI is everywhere. In 2026, it’s not just a buzzword; it’s a daily reality for anyone who writes code, manages data, or even just uses a computer. As a developer, ignoring AI would be like refusing to use version control or a search engine. It’s a tool, a powerful one, that can make us faster, more productive, and sometimes even more creative.
But like any tool, it’s all about how you use it. The temptation is huge: automate everything, generate everything, move faster and faster. But the real question is: where does AI actually add value?
Stack Overflow, AI, and the Evolution of Developer Help
Before AI, Stack Overflow was the go-to resource for developers stuck on a problem. I’ve spent countless hours searching for that one answer, that one code snippet, or that clever workaround. Stack Overflow taught me to read, to adapt, to understand, not just to copy-paste. It was a community-driven knowledge base, with human nuance, debate, and context.
AI tools are different. They give you answers instantly, tailored to your context, sometimes even before you finish typing. But there’s a risk: you might skip the process of searching, reading, and understanding. With Stack Overflow, you learned by osmosis, by reading threads, seeing different solutions, and absorbing best practices. With AI, you get the solution, but sometimes miss the journey.
I still use Stack Overflow, especially for edge cases or when I want to see how other humans have solved a problem. But I’m careful: AI is fast, but Stack Overflow is often deeper. The best workflow? Use both, and never stop questioning the answers you get. For me, this is truly the ultimate thing to do and remember, above all else.
Where AI Shines (and Where It Doesn’t)
AI can be great for certain tasks:
- Automating repetitive tasks (test generation, scripts, boilerplate)
- Summarization (logs, specs, tickets)
- Sentiment or trend analysis (feedback, support, triage)
- Information extraction (parsing emails, reports, raw data)
- Spellchecking and rewording (like this very text!)
AI shines when it removes manual work or accelerates understanding. But I always ask myself: is this making me faster, or is it stopping me from thinking?
Bad Uses
- Generating an entire project without understanding it
- Copy-pasting code without reviewing (you don’t need AI for this one lol)
- Using AI to hide a lack of understanding
- Letting AI decide architecture or critical choices
AI should be an accelerator, not a substitute for thinking.
The Dependency Trap: When the Tool Becomes a Crutch
I’ve had moments without internet, or without access to my favorite AI tool. And then, surprise, writing a simple for loop takes me 10 seconds of thought. Not because I don’t know how, but because muscle memory fades when you delegate everything to autocomplete.
It’s exhilarating to go fast, but it’s dangerous to become dependent. Sometimes I feel like AI makes me “better” but also lazier. I code faster, but I retain less. I understand more, but I think less deeply.
That’s why, regularly, I turn off AI. I force myself to write, to search, to make mistakes. It’s slower, but it’s formative. Like a musician practicing scales without a metronome, just to keep the feeling.
Understanding vs. Doing: The Real Value
Today, anyone can generate code. But the real difference is:
- Do you understand what you’re doing?
- Do you know why this code works (or doesn’t)?
- Can you explain it, optimize it, fix it?
AI can give you a solution, but it doesn’t teach you to think. It doesn’t teach you to debug, to anticipate bugs, to choose the right architecture. It doesn’t teach you to say no to a bad idea.
I’ve seen “fast” devs unable to explain their own generated code. I’ve seen juniors progress quickly thanks to AI, but freeze as soon as they have to step off the beaten path.
Ethics, Responsibility, and the Human Touch
There’s also the ethical question: who’s responsible for generated code? If AI makes a mistake, who fixes it? If it introduces a vulnerability, who catches it? Using AI means accepting to review, to validate, to take responsibility for the result.
I don’t think AI will ever replace creativity, intuition, or attention to detail. It can help you go faster, but it won’t replace the satisfaction of finding an elegant solution, or of understanding a tricky bug at 3am.
I also want to highlight that many of my prompts to AI are not just “code this for me” requests. More often, I ask the AI what it thinks about my approach, or how it would do something, without asking for code at all. If the AI agrees with my plan, I try to dig deeper: is it just agreeing to be helpful, or is there a real reason? I challenge myself to ask, “Why might someone disagree? What other solutions exist?” If the AI disagrees, I ask: is it right, did I explain myself poorly, or is the AI simply wrong? (Yes, it happens, and more often than you might think.) This back-and-forth is where I find the real value: not in the answer, but in the questioning and the dialogue.
My Hot Take: It’s Not About the Tool, It’s About the Thinking
For me, the real difference between a good and a great developer isn’t about how fast you can ship, or whether you use AI, Stack Overflow, or any other tool. It’s about how you think. Can you step back and question your own assumptions? Can you explain your choices, adapt when you’re wrong, and see the bigger picture? AI can help you go faster, but it can’t do the thinking for you. If you use AI as a shortcut to avoid understanding, you’re missing the point. If you use it as a partner to challenge your ideas, to push your reasoning, and to help you see blind spots, then you’re using it well. In the end, nobody cares if you used AI, as long as you understand, can explain, and take responsibility for your work. That’s what really matters.
Anecdotes and Practical Tips
- I use AI to generate unit tests, but I always review them by hand.
- I ask AI to explain legacy code, but I always check the logic myself.
- I refuse to generate entire projects. Honestly, it reminds me of when you’re new in a company or a codebase, everything feels confusing, you wonder why things are the way they are, and it takes real time to understand the big picture. For me, building piece by piece and keeping control of the architecture is essential, otherwise you just stack confusion on top of confusion.
- When I’m stuck, I turn off AI and think the old-fashioned way. Often, the solution comes from writing, not generating.
- I use AI to learn new patterns, but I adapt them to my context.
Conclusion: Master the Tool, Don’t Be Ruled by It
AI is like a superpower. But a superpower used badly does more harm than good. Use AI. Enjoy the speed, the automation, the creativity it can bring. But never forget to keep control, to review, to understand, to think.
The real value of a developer isn’t speed, it’s knowing where you’re going.
This text was drafted with the help of AI… but every idea, every nuance, every doubt is 100% human.