Introduction: openKylin in Prague
The Open Source Summit Europe 2026, held in Prague from October 7‑9, served as the backdrop for a major announcement from the openKylin community. As a Chinese initiative aimed at creating an optimized, free‑to‑use operating system for artificial intelligence, openKylin presented its bold vision: to transform Linux into an AI‑ready platform through its new KART runtime and a system‑level agent environment.
The event gathered developers, researchers and companies from around the world. Executive Director Min Liu’s keynote sparked interest by detailing how the convergence of KART and system agents could revolutionise the way AI applications interact with hardware.
1. KART: The Unified Runtime
KART (Kernel AI Runtime) is designed to be a single execution engine capable of handling multiple AI models simultaneously while adapting to the resource constraints of embedded systems. By integrating a just‑in‑time compilation layer, KART optimises performance while reducing power consumption.
Key Features
One of KART’s strengths is its native support for TensorFlow Lite, Pytorch Mobile and other frameworks. It also offers a dynamic update mechanism that lets developers push improvements without rebooting the system.
- Multi‑framework support
- JIT optimisation for CPU/GPU/TPU
- Real‑time updates
2. System Agents: A New Architecture
The agent architecture proposed by openKylin decentralises AI tasks to the kernel level, creating an environment where every software component can become an autonomous agent. This enables better process isolation and enhanced security.
Application Example
Imagine a surveillance system that continuously analyses video streams, detects anomalies and automatically triggers corrective actions without human intervention. With system agents, these tasks can be executed directly in the kernel, reducing latency.
This approach opens doors to use cases such as real‑time energy optimisation or predictive industrial maintenance.
3. Linux RISC‑V: Towards a More Open Ecosystem
Jianfeng Li, a technical committee member, showcased a ready‑to‑deploy Linux AI workstation on RISC‑V. This initiative highlights openKylin’s commitment to supporting open‑source processors to foster innovation and technological sovereignty.
Using RISC‑V not only reduces costs but also accelerates the development of custom AI solutions, as manufacturers can tailor hardware to their specific needs.
4. Integration with Cloud Services
openKylin is not limited to embedded systems; it also offers seamless integration with major cloud providers. Through standardised APIs, AI models can be deployed in the cloud and invoked locally via KART.
This hybrid approach provides an ideal balance between local performance and global scalability, allowing businesses to keep data privacy while benefiting from massive compute resources.
5. 2026‑2028 Roadmap: What’s Next
“We stand at the dawn of a new era where Linux becomes the foundation of artificial intelligence,” declared Min Liu during the keynote.
The roadmap foresees a public beta release of KART by late 2026, followed by official support for system agents in 2027. openKylin also invites the community to contribute through bug‑bounty programmes and open‑source development.
Conclusion: Join the Movement
The vision presented at Open Source Summit Europe 2026 shows that Linux can be reinvented to meet the growing demands of AI. Whether you’re a developer, researcher or company, today’s tools and resources let you participate in this revolution.
To stay informed about upcoming updates, subscribe to openKylin’s newsletter and download the 2026‑2028 roadmap now. Together, we’ll build the future of AI on Linux.
