Linus Torvalds tries vibe coding, world still intact somehow

The most famous low-level systems programmer in the world has tried "vibe coding" for himself—and he seems to be enjoying it. Linus Torvalds, best known as the leader of the Linux kernel project, revealed that he used Google's Antigravity LLM to write a Python visualizer tool for his latest side project, AudioNoise.

What happened

Torvalds unveiled AudioNoise as a project for creating "random digital audio effects" using a "random guitar pedal board design" he shared last year. Buried in the project's README file was an unexpected admission: the Python visualizer tool was "basically written by vibe-coding."

"I know more about analog filters—and that's not saying much—than I do about python," Torvalds wrote. "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."

This marks a notable evolution in Torvalds' public position on AI coding assistants. In November, during an on-stage chat with Dirk Hohndel at the Open Source Summit Asia, Torvalds expressed a moderate stance: he's OK with vibe coding as long as it's not used for anything that matters.

Legal and regulatory context

While Torvalds' personal hobby project falls outside commercial software development, his acceptance of AI-assisted coding touches on emerging regulatory questions around AI-generated code. The European Union's AI Act, which entered into force in August 2024, establishes risk-based requirements for AI systems, including transparency obligations for generative AI models.

Companies using AI coding assistants like GitHub Copilot, Amazon CodeWhisperer, or Google's Antigravity must now consider:

• Data provenance: Where did the training data come from? Are there licensing conflicts?
• Intellectual property: Who owns code generated with AI assistance?
• Liability: Who is responsible if AI-generated code contains security vulnerabilities or bugs?
• Compliance: Does the use of AI tools trigger disclosure requirements under GDPR or other privacy regulations?

The GDPR doesn't directly address AI-generated code, but if personal data is processed during training or if AI tools handle user data, companies must ensure compliance with data minimization, purpose limitation, and security requirements.

Impact on developers and companies

Torvalds' pragmatic acceptance signals that AI coding tools are moving from controversial to commonplace. For professional developers, this raises several practical concerns:

For individual developers: Using AI assistants for non-critical code may become normalized, but the responsibility for code quality, security, and licensing remains with the human programmer. You can't delegate accountability to a machine.

For companies: The legal landscape remains murky. Using AI-generated code in production software could expose companies to:

• Copyright infringement claims if training data included proprietary code
• Security audits that fail to identify AI-introduced vulnerabilities
• Regulatory scrutiny under emerging AI governance frameworks

For open source projects: Torvalds' position matters because he controls the Linux kernel, one of the world's most critical open source projects. His acceptance of AI tools for hobby projects doesn't extend to kernel development. The kernel maintainers have strict rules about code provenance and licensing—AI-generated code could introduce license contamination or security issues that are unacceptable in critical infrastructure.

What changes

Torvalds' admission reflects a broader industry shift toward pragmatic AI adoption. The question is no longer "should we use AI coding tools?" but "where and how should we use them?"

For hobby projects and non-critical applications, AI-assisted coding appears increasingly acceptable. Torvalds' AudioNoise project—built with Raspberry Pi-driven audio effects for his homegrown guitar pedal—fits this category perfectly. It's experimental, personal, and carries no risk to users or systems.

However, for production software, especially critical infrastructure like the Linux kernel, the standards remain unchanged. Code must be reviewed, understood, and maintainable by human developers. AI-generated code that can't be properly audited or explained has no place in systems that power the internet, financial systems, or medical devices.

The regulatory environment will likely evolve to address these distinctions. Until then, Torvalds' approach offers a reasonable framework: use AI tools where they help, but never abdicate human responsibility for the final product.

Related projects and resources:

• AudioNoise project (Torvalds' GitHub repository)
• Google Antigravity LLM documentation
• Linux Kernel Documentation on code contribution standards
• Subsurface dive logging software (Torvalds' other open source project)
• EU AI Act official text

The Emperor Penguin has spoken, and the world remains intact. For now, vibe coding stays in the hobby corner where it belongs—while the serious business of systems programming continues under the watchful eye of human maintainers.