Linus Torvalds Tries Vibe Coding (Just a Little)
▼ Summary
– Linus Torvalds used an AI tool called Google Antigravity to write the Python code for a visualizer in his latest hobby project, AudioNoise.
– This project is a digital audio effects generator for guitar pedals, which he describes as another “silly” personal repository.
– Torvalds clarifies that his use of AI for this specific task does not mean he broadly endorses “vibe coding” for all programming.
– He has previously expressed nuanced views, valuing AI primarily as a tool for code maintenance, review, and finding errors.
– However, he is generally less interested in AI for writing new code and is strongly opposed to the hype surrounding AI technology.
The creator of Linux and Git, Linus Torvalds, recently revealed he used an AI coding assistant for a personal project, sparking fresh discussion about the role of these tools in software development. While the code for a Python visualizer was “basically written by vibe coding,” Torvalds’ broader stance remains characteristically pragmatic and far from a wholesale endorsement of AI-generated code.
Torvalds often dedicates holiday time to small, personal engineering projects. Following last year’s work on guitar pedals, he turned his attention this year to a repository called AudioNoise, which he describes as another lighthearted endeavor related to digital audio effects for guitars. Within the project’s documentation, he openly disclosed his method for creating a specific tool. He admitted to knowing very little about Python, starting with his usual approach of searching for solutions online and replicating them. This time, however, he decided to “cut out the middle-man, me” and utilized an AI tool named Google Antigravity to build the audio sample visualizer.
Google Antigravity is a specialized fork of the AI-integrated development environment Windsurf. Although Torvalds did not specify the exact underlying model, his use of Antigravity strongly implies he leveraged a version of Google’s Gemini AI. This practical experiment stands in contrast to his previously expressed, nuanced views on large language models in programming, which have often been more measured than the polarized debates found online.
Torvalds has publicly highlighted the potential of AI as a valuable tool for maintaining and reviewing existing code, citing instances where automated tools successfully identified issues he had overlooked. His primary interest lies in leveraging AI for tasks like automated patch checking, which can enhance code quality and reliability. Conversely, he has consistently expressed significantly less enthusiasm for employing AI to write original code from scratch. His position is not rooted in an ideological opposition to the technology but rather a well-known skepticism toward the excessive hype that often surrounds it. For Torvalds, AI is just another tool, useful in specific, well-defined contexts but not a replacement for fundamental understanding and careful engineering.
(Source: Ars Technica)
