This is quite an interesting topic.

Yesterday over dinner with a friend, we discussed Bloomberg’s recent article about Vibe Coding, which claimed the technology is “killing” junior engineers and might even dismantle the entire developer talent pipeline. Honestly, as someone in tech, my first reaction to such headlines is always: “Here we go again with the fearmongering.” But upon closer thought, there might be some truth to it.

Vibe Coding, in simple terms, lets you write code using natural language, with AI translating requirements directly into executable code. Sounds amazing, right? No need to learn syntax, no more debugging until 3 AM—just speak your mind. But here’s the catch: if the barrier to writing code disappears entirely, who’s going to bother learning the fundamentals?

My friend’s company has been hiring recently and noticed a growing trend among fresh graduates: they’re getting “too comfortable.” Ask them a basic algorithm question, and the response is, “Why not just use Copilot?” Even worse, some don’t even know basic Git operations, arguing that “AI can handle code merging.” That’s downright alarming. Tech stacks can be rushed, but problem-solving thinking and engineering discipline? AI can’t spoon-feed those.

The Bloomberg article highlights a critical point: junior engineering roles are vanishing. Big tech used to hire rookies to groom future tech leaders. Now? With AI as a safety net, they’re axing entry-level positions and poaching seasoned engineers instead. Short-term cost savings, sure—but long-term, the industry’s talent pipeline collapses. Where will we find people capable of handling complex systems in five years? Rely entirely on AI to iterate on itself?

Some argue this is the inevitable cost of progress. But let’s be honest, this kind of “progress” feels like robbing Peter to pay Paul. Code can be auto-generated, but who ensures quality? Who understands the business logic? When AI-written code fails, human engineers still end up cleaning the mess. The irony? Many AI coding tools are trained on decades of code written by human engineers. Once this generation retires, and the next lacks foundational skills, will AI start fabricating its own training data?

Beneath this lies a deeper conflict: companies want plug-and-play talent but refuse to invest in training. Education systems are still teaching outdated curricula, while students cram AI tool crash courses for job prospects. The likely outcome? A hollowed-out middle layer, leaving only a handful of elites and a sea of “pseudo-developers” who just call APIs.

Don’t get me wrong—AI coding isn’t inherently bad. It boosts efficiency and unleashes creativity. But the crux is, we can’t equate “using tools” with “solving problems.” When calculators became ubiquitous, math education grew more important, not less. The issue now is the industry’s reckless sprint toward “quick fixes,” with no one pausing to ask: Will human engineers even be needed in tech’s future?

(PS: Sent this draft to my friend, who replied: “Relax, by the time AI can debug itself, we’ll be retired.” Me: “
Fingers crossed.”)