The most hyped topic on the internet today is vibe coding. A lot of people are talking about it, and opinions vary widely. Some ideas contradict others, and some folks are even worried about losing their jobs—concerned that people will soon be able to build apps without the foundational knowledge we, as engineers, spent years acquiring, long before artificial intelligence looked the way it has for the past two or three years.
To be honest, I wasn't impressed with artificial intelligence at first. When I initially paid $20 for ChatGPT, it didn't help me optimize my coding process in any significant way. I remember even the unit tests it generated were, well, pretty mediocre. So, back then, I concluded that I had just paid $20 for a personal English translator and a writing assistant for emails.
But over time, artificial intelligence evolved. A lot has changed. Companies invested in chip development, engineers worked on expanding context windows and giving models more resources. And now we have an incredibly powerful tool—one that knows way more than we do and can implement things 10 times faster than the human brain. But more on that in a bit.
Just recently on the platform X, a guy started sharing how he managed to launch an online business using Cursor, without any programming background. Honestly, I didn't dive into the project's specifics, but the fact is, he built and launched something using just prompting. Someone with zero knowledge of JavaScript, or any language for that matter, managed to launch a functioning product. I want you to pause and really reflect on that. This is already happening.
Eventually, someone hacked this guy's database, injected SQL, and started launching DDoS attacks on the project. The guy began posting about it, asking for help, and actively prompting to find ways to protect his project. I stopped following the story after that, but even this much is enough to start drawing some conclusions.
I believe we're at a crossroads. Not everyone has adapted to this new reality yet, but we've already begun developing in new ways. At the same time, we're still assessing developer knowledge the old-fashioned way, while approaching proof-of-concept creation with a completely new mindset.
The most important thing we, as engineers, need to internalize is this: we now have a permanent assistant who can quickly offer solutions to the problems we're trying to solve. And having that assistant? That's perfectly normal. It doesn't mean we no longer need to learn new technologies, or that people like the guy from Twitter will steal our jobs, or that artificial intelligence will take over completely.
Instead, this means we now have more time to dive deeper into product context—thinking about features, user experience, and edge cases. These are things we used to sacrifice when building tech solutions took months. Now, you can get a working piece of code you can share in just a few hours.
At last, we can become true product engineers—those who solve actual product problems.
When it comes to the quality of solutions artificial intelligence provides, it's time to talk about our own knowledge and how deeply we understand the technologies we use—whether it's a programming language, framework, or library. In my opinion, deepening your technical understanding is more important than ever. That way, you can steer your assistant with greater precision by giving it more context, and in return, you'll get a much higher-quality output.
Technical expertise is still highly relevant. This guy's project shows that lacking basic knowledge of security and how to protect databases from attacks is a critical issue. Artificial intelligence cannot cover every edge case—it doesn't understand the full picture or your expectations, which are often only in your head.
Let's be clear: the human brain is still far more efficient than artificial intelligence. Our brains can handle a much greater volume and depth of context. We use short-term memory to juggle 5 to 9 items at once, and long-term memory has virtually unlimited capacity. We effortlessly form abstract associations, analogies, emotions, and social context—all of which are essential when building products.
Artificial intelligence, on the other hand, has strict limits. For instance, GPT-4 can retain about 128,000 tokens, which is roughly 150 pages of text. Its understanding is fairly surface-level, built on static relationships, not actual comprehension or awareness. The biggest problem is that it can't remember past conversations beyond the current session without additional tools or techniques.
Here are my takeaways:
-
Artificial intelligence is not a replacement—it's an enhancement of who we are as professionals. Yes, we can ask it to write TypeScript types, create unit tests, or optimize rendering. But it doesn't know what actually matters in your specific context. Only you know that.
-
The more you know, the better AI works for you. Artificial intelligence is like a mechanic—and that's a great analogy because my dad worked with cars for years, so it really resonates with me. If you understand how things work, you can clearly say, "Replace this hose, don't touch the transmission, and check the brakes—I heard a weird noise there." But if you don't know anything, all you can say is, "It's not running right," and you'll get vague feedback in return. Your understanding isn't an alternative to AI—it's what makes AI's help more precise and effective.
-
The best developers aren't the ones who know everything—they're the ones who understand deeply. Even if you don't remember every detail about server-side rendering or accessibility, understanding why it matters, how it impacts the product, and what tradeoffs exist puts you a step ahead. You become someone who leads the process, not someone who just reacts to problems.
-
Being the best isn't just about your technical level. It's about seeing the big picture, Being independent in problem-solving, Being able to lead others, And having the courage to make decisions.
Now it's the industry's turn to adapt. Companies need to respond accordingly. In my opinion, the hiring process is broken. Many companies assess developers through live coding sessions, which rarely reflect the full range of skills or a candidate's true potential. The fact that tools are being developed to cheat interviews using AI proves this. We need to rethink how we evaluate candidates.
Applicants should be allowed to use all the tools they would have access to on the job. Interviewers can still validate depth of knowledge through well-crafted questions and thoughtful problem scenarios.
We're living in an extraordinary time, and I truly believe that what's ahead will only get more interesting.