Linus Torvalds is Vibe Coding Now. Yes, That Linus Torvalds.

If you told me a year ago that the creator of Linux and Git would be casually mentioning "vibe coding" in a GitHub README, I would have assumed you were trolling.

Linus Torvalds is famous for a lot of things. Creating the Linux kernel. Inventing Git. Telling Nvidia where they can stick their closed-source drivers. Having extremely strong opinions about code quality. The man once called a contributor's patches "garbage" that were "making the world worse."

So when he dropped this line in the README of his new hobby project, AudioNoise, the developer world collectively did a double-take:

"Also note that the python visualizer tool has been basically written by vibe-coding. I know more about analog filters — and that's not saying much — than I do about python. It started out as my typical 'google and do the monkey-see-monkey-do' kind of programming, but then I cut out the middle-man — me — and just used Google Antigravity to do the audio sample visualizer."

The Emperor Penguin himself is vibe coding. In 2026. And the world is somehow still intact.

What He Actually Built

AudioNoise is a hobby project. Torvalds describes it as creating "random digital audio effects" using the "random guitar pedal board design" he built last year. The goal was learning about digital signal processing, not shipping production software.

The project has two parts:

• The core audio filters (delays, phasers, reverb, etc.) — written in C, by hand, by Linus
• The Python visualizer — vibe coded with Google Antigravity

This split is the interesting part. Torvalds stuck to his strengths for the critical logic. The audio processing math? That's where his expertise matters. He wrote every line himself in C, the language he's mastered over decades.

But the visualization script? The part that just needs to render some graphs so he can see what the filters are doing? He outsourced that to AI.

His reasoning is refreshingly pragmatic: "I know more about analog filters — and that's not saying much — than I do about python."

The Pattern Worth Stealing

There's a Hacker News comment that nailed what's actually happening here:

"There's a big difference between vibe-coding an entire project and having an AI build a component that you lack competency for. That is what is happening here. It's the same principle as a startup that builds its core functionality itself in-house and then uses off-the-shelf libraries for all the other uninteresting details."

Torvalds established a replicable pattern that any senior developer can use:

1. Core logic (manual): The critical path stays handwritten. This is where your expertise matters, where bugs have real consequences, where performance is non-negotiable.
2. Presentation layer (AI-generated): The "show me the data" parts. Graphs, visualizations, quick UIs. Speed matters more than elegance.

For AudioNoise, if the Python script fails, the graph doesn't load. No servers crash. No security vulnerabilities get exposed. The stakes are zero. That's exactly the right environment for vibe coding.

What This Doesn't Mean

Before anyone panics: Linus is not vibe coding the Linux kernel.

At the Open Source Summit Asia last November, Torvalds said he was "OK with vibe coding as long as it's not used for anything that matters." He maintains strict standards for kernel contributions. AI-generated patches still need to meet the same quality bar as human-written ones.

This is a hobby project. A learning exercise. The README literally says: "These are — like the analog circuits that started my journey — toy effects that you shouldn't take seriously."

The distinction matters. Using AI to scaffold a visualization script for a weekend project is fundamentally different from merging AI-generated code into an operating system that runs most of the internet.

The Real Takeaway

Torvalds' approach here is a masterclass in knowing when to use tools and when not to.

He wanted to learn about digital signal processing. That meant writing the DSP code himself, understanding the math, grappling with the implementation details. That's where the learning happens.

But the Python visualization? That's not what he was trying to learn. It's a means to an end. He needed to see the output of his filters. Fighting with matplotlib documentation wasn't going to teach him anything about audio processing.

So he described what he wanted, let AI handle the syntax, and got back to the interesting problem.

That's not laziness. That's prioritization. It's recognizing that your time and attention are finite, and spending them on the parts that actually matter.

Try This Yourself

Here's the Linus Torvalds approach to vibe coding, distilled:

1. Identify your core problem. What are you actually trying to learn or build? That part, you do yourself.
2. Identify the supporting infrastructure. What's just scaffolding? What's a means to an end? That's where AI can help.
3. Match the stakes to the method. If failure means a broken graph, vibe code it. If failure means a security vulnerability, write it yourself.

For me, this has meant things like:

• Writing the actual security logic myself, but vibe coding the quick dashboard to display the results
• Hand-coding the parsing logic for threat data, but letting AI generate the report formatting
• Building the detection rules manually, but outsourcing the "make this data exportable to CSV" part

The goal isn't to replace your expertise. It's to stop wasting your expertise on things that don't need it.

Even Linus Torvalds, the man who built Linux and has opinions about everything, knows when to cut out the middle-man.

---

What's your split? What parts of your projects are worth doing by hand, and what's just scaffolding? I'm curious how other people are drawing that line.