At the recent Open Source Summit India 2026, Linux creator Linus Torvalds offered candid insights into the current state and future trajectory of the operating system. Alongside his friend Dirk Hohndel, Torvalds discussed everything from his evolving role in development to the challenges and opportunities presented by artificial intelligence.
He emphasized that the Linux kernel’s success stems from a philosophy of “steady progression and continued improvements,” rather than chasing “blockbuster” releases. Ever since the creation of the Git version control system, the focus has been on incremental progress, a model Torvalds actively champions.
The Evolving Role of a Development Lead
One of the most striking revelations from Torvalds was his changing perception of his own role. “Let’s be entirely honest. I hardly read code at all anymore. I’m not a programmer, I’m a development lead,” he stated, highlighting a significant shift after decades at the helm.
While he still drafts small patches, these are offered as suggestions rather than definitive code. He clarified, “I expect the maintainers of the code to be the ones who then send me the fix back. So I very seldom commit my own code anymore.” His primary focus now is on understanding the “bigger picture” and the intent behind pull requests, often diving into code only when forced by issues like build breaks or merge conflicts.
Torvalds admitted that the human element can be far more challenging than technical hurdles. “Code is easy to fix. Personality is not always as easy to fix,” he mused, acknowledging his own past contributions to such issues while noting his efforts to improve.
When it comes to his personal workflow, Torvalds keeps it remarkably simple. He revealed, “Git and email are the two really only tools I use.” He occasionally uses Google for research, but largely avoids the myriad of tools other maintainers, including those exploring AI, might employ. “I work with people, not tools,” he clarified.
Navigating AI and Legacy in the Linux Kernel
The conversation inevitably turned to Artificial Intelligence, which Torvalds noted is increasingly pressuring the kernel’s development workflow. While AI has proven adept at finding interesting bugs, this has also “stressed out people in the community” due to the volume of reports.
A significant pain point has been the influx of AI-generated bug reports that are entirely bogus or “hallucinations,” requiring considerable human effort to debunk. Torvalds stressed the need for human oversight: “if you find a bug with an LLM, it’s not enough to just ask the LLM to make a bug report… We want to see a suggested patch; we want to see the human who ran the LLM act as a kind of back-and-forth.”
Despite the challenges, Torvalds acknowledged that some AI-identified issues have been “absolutely, stunningly… interesting in a painful kind of way,” particularly security vulnerabilities that have quickly made headlines. He maintains a pragmatic view: “I think we’re much better off with LLMs finding bugs, even when they are embarrassing.”
On the topic of supporting older technologies, Torvalds revealed a decidedly unsentimental approach. “I’m not very sentimental when it comes to technology,” he stated, discussing ongoing efforts to remove support for “museum” hardware and obsolete code. Recent changes include dropping support for:
- Machines without hardware floating point on x86, such as the Intel 486 SX.
- Obsolete networking standards like ISDN and ATM.
This push is driven by the growing burden of maintaining support for hardware that virtually nobody uses anymore, freeing up resources for more relevant development.
C, Rust, and the Future of Code
The ongoing discussion about integrating Rust into the kernel also came up, with Torvalds pushing back against some of the hype. “I’m not sure Rust is going to take over the world. I still think Rust is very interesting, [but] I still find C to be a much simpler tool,” he explained.
He expressed more enthusiasm for the robust verification tools available for C code, such as automated patch and email checking utilities. Torvalds underscored that while Rust addresses certain “easy bugs” in C, it doesn’t solve fundamental “logic errors.” He added, “when you write incorrect code, the language does not matter. The end result will be incorrect.”
Furthermore, he cautioned that Rust’s safety guarantees only apply within its own code sections. As soon as Rust interfaces with C code, those guarantees largely diminish, especially when interacting with the highly tested and scrutinized core kernel C code. Many recent high-profile bugs, he noted, have been logic errors arising from “bad programming.”
Torvalds concluded his thoughts on AI by highlighting its practical, if sometimes unconventional, uses. For his own toy projects, he utilizes LLMs as prototyping tools, acknowledging that while the generated code might not be production-ready, it’s excellent for initial experimentation. He also shared a humorous anecdote about using AI to add Godzilla to his travel photos, proving that even the creator of Linux finds diverse “useful and less useful uses for AI.”
Source: ZDNet – AI
