Vibe Coding: Still Learning, Still Vibing

When I first wrote about vibe coding, I saw it as a fresh, almost playful way of thinking about development—coding less like grinding through syntax and more like shaping an idea in conversation with an AI. Since then, I’ve gone deeper: I’ve tried Replit, Cursor, and now Kilo inside Visual Studio with Copilot and Claude. Along the way, I’ve burned down and rebuilt multiple instances, hit roadblocks with authentication, ripped out Clerk when database connections got messy, and wrangled with Docker, Digital Ocean, Spaces, and Stripe.

It hasn’t been plug-and-play. Far from it. But persistence pays off, and I’ve come to see the process less as failure and more as learning. Each rebuild teaches me something about how these pieces fit together, and—surprisingly—it gets kind of fun.

The Evolution of Vibe Coding

The term vibe coding was coined by Andrej Karpathy in 2025. At its core, it means coding by describing—using natural language prompts to build and refine software, rather than painstakingly typing every line of code. (Wikipedia)

• For developers, it’s about efficiency: letting AI handle boilerplate, integrations, and UI scaffolding, while you focus on higher-level architecture and creativity. (IBM)
• For non-coders, it’s about accessibility: removing the barrier of syntax so anyone can build something useful, whether it’s a personal app, a prototype, or a workplace tool. (Business Insider)

Adoption has been fast. Y Combinator reported that a quarter of its Winter 2025 startups used vibe coding for most of their codebase. Tools like Cursor, Lovable.dev, Bolt.new, and Replit Agent are lowering the threshold for both seasoned devs and citizen developers to get something running quickly.

But the story isn’t just hype. As WIRED and others point out, AI-generated code raises real questions about quality, security, and maintainability at scale. (Wired, Economic Times)

Where My Journey Fits

That tension—between possibility and friction—is exactly what I’ve been living.

On the one hand, I’ve seen firsthand how vibe coding accelerates things. In minutes, I can scaffold projects that would have taken hours or days before. I can test ideas, break things, and rebuild without sweating the small stuff.

On the other hand, I’ve learned that integration is where the magic (and pain) happens. AI can help you spin up an app skeleton, but connecting it to real-world systems like authentication, payments, and cloud infrastructure is still hard. Sometimes, you just have to rip something out and go back to basics.

So, What’s Vibe Coding For?

I’ve come to think of vibe coding as a bridge:

• For developers, it’s a way to work faster and stay in flow—like pair programming with an AI that never gets tired.
• For non-coders, it’s a way to get closer to building something real, without needing a CS degree.

It’s not “done” yet. It’s clunky, it’s imperfect, and it can’t (yet) replace the hard parts of system design. But the direction is clear. What feels fragmented today will evolve into seamless, integrated workflows tomorrow.

And if you’re willing to persist, rebuild, and embrace the learning curve—it really does start to feel like fun.

Photo by MARK ADRIANE on Unsplash