That’s the question that is stirring debate across developer communities ever since Andrej Karpathy coined the term in a now-viral tweet.
Because the tweet alone had sparked a full-blown developer identity crisis. Suddenly, everyone is asking: Will Vibe coding make developers obsolete? Some say it is democratizing coding. Others believe it is going to unleash superpowers in coders.
As a leader in AI services, if Trigent were to take on this phenomenon: Vibe coding definitely presents a potential route for 10X development, but in the process, it signals a drastic shift from actual coding. While it definitely opens fresh avenues for low-code creativity, it also raises questions about shedding control, maintaining code quality, and the evolving definition of what it means to be a developer.
Let’s break down what Karpathy actually said and what it could mean for the future of software engineering.
What Is Vibe Coding?
There’s a new kind of coding I call ‘vibe coding,’ where you fully give in to the vibes, embrace exponentials, and forget that the code even exists.
At a glance, vibe coding seems magical. Karpathy describes a workflow where he speaks to Composer, an AI coding assistant via SuperWhisper (a voice-to-text tool) — which means he is barely touching the keyboard. He prompts for things like “decrease the padding on the sidebar by half” and accepts whatever changes the AI suggests—no questions asked.
On the surface, this looks like supercharged productivity. But beneath the surface, it invites three big questions:
1 Control vs. Convenience: Are Developers Giving Up Too Much?
Karpathy openly admits he no longer reviews diffs and rarely touches the code:
I ‘Accept All’ always. I don’t read the diffs anymore.
That might be fine for a throwaway side project, but in production environments, skipping code review could introduce serious risks. Accepting AI-generated suggestions without oversight might save time in the short term, but could cost much more in technical debt down the line.
GitClear’s research confirms the same. According to the software analytics company1, there is an 8-fold increase in the number of code blocks that were duplicated right next to similar blocks. Duplicated code makes your system harder to maintain, debug, and update in the long run – a classic sign of technical debt.
2 Comprehension vs. Completion: Are We Still Thinking Critically?
When I get error messages I just copy paste them in with no comment, usually that fixes it.
Debugging is one of the hardest and most human parts of development. In another research, software developers claim that they spend more time debugging and resolving vulnerabilities present in AI-generated code.
So while vibe coding might help you complete a project faster, is it helping you understand what you’re building? If the AI writes the code, and the coder no longer reads it, who’s really in control?
Even Karpathy seems to backtrack:
The code grows beyond my usual comprehension… I’d have to really read through it for a while.
So understanding the code might still be part of vibing—even if we pretend it isn’t.
3 Playground or Production? Where Does Vibe Coding Fit?
It’s not too bad for throwaway weekend projects.
Andrej draws a boundary here, whether he realizes it or not. He uses vibe coding for personal projects, not for mission-critical software. That’s telling. It suggests vibe coding is great when stakes are low, expectations are relaxed, and experimentation is the goal.
At Trigent, we believe in innovation, but not at the cost of engineering discipline. We have to say we are already experimenting with vibe coding tools such as Cursor and V0.dev They are helping us accelerate our POCs and prototypes, but the guardrails are always in place. We pair innovation with rigorous code reviews to avoid black box complexity.
Suffice to say, Vibe coding may very well be the spark behind the next wave of low-code creativity—but it’s not a replacement for strong software architecture, testing, and review.
The Real Work? Still Thinking, Still Designing
One of the best responses in Karpathy’s thread nailed it:
Articulating the problem requires sufficient mental effort… AI helps with the busywork, but thinking through the architecture and abstractions is still 90% of the battle.
That’s the heart of it. Vibe coding shifts the manual work onto the AI, but the creative, structural, and logical thinking still falls on the human.
Final Take
Vibe coding indeed encourages a faster, more conversational way to build software. And in that regard it may sound that you are coding less — but it is really AI sharing your cognitive load so you can build more in less time.
But then, if we were to answer the burning question —will Traditional coding really die out?
If Vibe Coding becomes the norm, the spotlight is bound to shift. Writing new code could become a commodity — easy, fast, and widely accessible. But the real value will lie with those who can go deeper: developers who can decode logic, troubleshoot complex systems, and understand what’s happening under the hood.
Software may become easier to build, even ubiquitous. But building quality software — scalable, reliable, and fit for real-world problems — will remain as critical as ever. At the end of the day, we’re still solving problems with code. Vibe Coding might just be the way to do it 10X faster.
