You vibe code,
I conduct code.

There’s a new taxonomy forming in software development, and most people are landing on the wrong side of it without realizing it.

Everyone has an opinion about vibe coding right now. Half the internet thinks it’s the future of software. The other half thinks it’s how you accidentally HIPAA-violate yourself into a federal consent decree. Both are right. But neither has named the alternative properly.

The opposite of vibe coding is not “not using AI.” That ship has sailed. The opposite of vibe coding is conductor coding—and the distinction will define who still has leverage in this industry five years from now.

What vibe coding actually is

Vibe coding is the practice of describing what you want to an AI, accepting what comes back, and repeating until something runs. The practitioner’s relationship with the code is essentially aesthetic. Does it feel right? Does the demo work? Ship it.

The tell is what happens when it breaks. The vibe coder pastes the error back into the chat. They iterate on vibes until the error goes away. They cannot tell you why it works when it finally does.

This is not a criticism of beginners. Beginners have always needed scaffolding. The problem is when experienced practitioners—people who should know better—adopt this posture because it’s faster. They trade understanding for throughput. And throughput is great, right up until something goes wrong in production at 2 a.m. and nobody knows where to look.

What conductor coding is

A conductor does not play every instrument. But they know what every instrument is supposed to sound like, when it’s off, and why. They set the tempo. They make architectural decisions in real time. The orchestra executes; the conductor is responsible for the outcome.

Conductor coding is the same relationship with AI-assisted development.

You know what you’re building. You know why it’s structured the way it is. You can read the output critically—not just run it and hope. When the AI hallucinates an API that doesn’t exist, you catch it. When it produces code that works but creates a security boundary problem you’ll regret in six months, you see it.

The AI is fast. You are correct. The combination is how you actually ship something worth shipping.

The leverage question

Here’s what the vibe-coding evangelists miss:

AI doesn’t eliminate the value of expertise. It amplifies it.

A vibe coder with no domain knowledge and an LLM produces faster wrong answers. A conductor coder with deep domain knowledge and an LLM produces better outcomes faster than either could alone. The gap between those two practitioners is not closing. It’s widening.

Every AI coding tool that gets released makes the conductor coder more productive. It makes the vibe coder more confidently wrong.

The accountability gap

There’s a professional dimension here that nobody wants to talk about.

When a vibe-coded system fails—and it will—nobody can explain the failure mode, because nobody understood the system well enough to anticipate it. The code is an artifact of a conversation, not a design. There’s no mental model to reason from.

Conductor coders can explain their systems. They can write the postmortem. They can tell you exactly which decision created the blast radius and why a different decision would have contained it. That accountability is not just good practice—in regulated industries, in security-sensitive environments, in anything with contractual SLAs—it’s the entire job.

How you know which one you are

You don’t need to pass a test. Just answer one question honestly:

Neither is permanent. But only one of them scales.

cablepull is a principal security architect with 25+ years in the field. He builds things that have to work when it matters.