Why John Carmack Thinks Building New Operating Systems Is a Losing Battle
John Carmack argues that building new operating systems is often impractical due to high costs and limited benefits. He advocates improving existing platforms with AI rather than starting from scratch.
The Futility of Building New Operating Systems
John Carmack, known for his programming work on iconic games like Doom and Quake, recently shared a clear-eyed view on the challenges of creating new operating systems today. Responding to Jonathan Blow’s reflection on why fresh OS projects are scarce, Carmack draws from his own experience to explain why building a new OS is often impractical and inefficient.
He recalls being pitched LIBBA, a custom OS for smart glasses, and expresses admiration for minimalist, efficient systems like Oberon, Plan 9, and TempleOS. Carmack values “clear, efficient programs that do their job without baggage.” Still, he stresses that launching a new OS in the current tech landscape rarely makes sense. The high development cost, short lifespan, and heavy demands on developers outweigh any benefits.
The Meta XROS Experience and Custom OS Challenges
Carmack’s position is grounded in real outcomes. While at Meta, he strongly opposed the company’s investment in XROS, a fully custom OS for extended reality devices. He warned that such efforts drain resources and get bogged down by corporate politics, slowing innovation. Meta’s experience with XROS confirmed these concerns.
The LIBBA pitch Carmack mentioned also failed to justify itself, reinforcing a common industry pattern. Custom operating systems promise optimization but struggle to compete with established platforms like Android or Linux, which provide extensive ecosystems and developer familiarity.
AI and the Future of Software Foundations
Carmack’s thoughts intersect with AI development challenges. Despite AI’s advances, real-time processing demands—such as those required for self-driving cars—remain a hurdle. This shows why rebuilding OSes from scratch is often misguided. Modern AI benefits from scalable, existing infrastructures rather than isolated, custom builds.
He suggests that AI could improve software by cleaning up legacy code and making codebases more efficient, rather than replacing entire systems. This approach aims to reduce complexity without starting from zero.
Lessons from Past OS Experiments and What Lies Ahead
Systems like Plan 9 serve as cautionary examples. While innovative, such custom OSes rarely succeed in consumer markets where compatibility and rapid iteration are critical. For XR and AI, companies might achieve more by adapting proven platforms instead of betting on new OSes.
Carmack also points to upcoming challenges in 2025, where AI may favor textual interfaces over graphical ones. This shift could force software architectures to rethink how they support intelligent systems.
Balancing Innovation with Practicality
The key takeaway is that innovation should be grounded in practicality. The idea of dedicated OSes for devices like smart glasses remains appealing but often overlooks opportunity costs. Modern hardware complexities and AI-driven processes demand ecosystems that evolve incrementally.
For professionals in IT and development, this means focusing efforts on enhancing existing systems and integrating AI to optimize what’s already built. The next breakthroughs will likely come from AI-augmented improvements rather than new OS foundations.
