Operational

Thinker ∴ Researcher ∴ Hacker ∴ Coder ∴ Founder ∴
Based in Bucharest.

Back to Blog

On Vibe Coding / Coding with AI

The shift from methodical engineering to flowing with AI. What changes when the computer can write most of the code.

November 20, 20254 min

I've been coding for over 13 years. The last two years have been different.

The Old Way

Before AI, coding was methodical. You held the entire architecture in your head. You knew every file, every function, every edge case. You had to-because writing code was slow and expensive. Every line mattered.

You'd spend hours on StackOverflow, reading docs, debugging cryptic error messages. The feedback loop was long. The cognitive load was high.

The Shift

Now I code differently.

I start with a rough idea, bounce it off Claude or Cursor, let it scaffold something. I read the output, spot what's wrong, redirect. The code emerges through conversation, not careful planning.

It's less like construction and more like sculpting. The AI generates clay-lots of it-and I shape it into something that works.

What Actually Changes

Speed: What took days now takes hours. What took hours now takes minutes.

Leverage: I can explore ideas I would've dismissed as "too much work." The opportunity cost of trying something has collapsed.

Focus: I spend less time on boilerplate and syntax, more time on the problem I'm actually trying to solve. The craft shifts from writing code to directing systems.

The Uncomfortable Part

I don't memorize APIs anymore. I don't stress about perfect variable names. I let the AI handle the tedious parts.

Some people see this as degrading the craft. I see it as evolving it.

We didn't mourn assembly when high-level languages arrived. We built bigger things.

What Matters Now

The bottleneck isn't writing code-it's knowing what to build and how to architect it. Taste, judgment, experience. Those still matter. Maybe more than before.

AI makes you prolific. But prolific without direction is just noise.

The ones who win are the ones who vibe with the tools but stay grounded in the fundamentals: knowing what actually needs to exist, what users need, what systems can handle.

We're still scratching the surface. And I'm here for it.