As a technologist deeply immersed in the evolution of machine learning and natural language processing, Laurent Giraid has spent years analyzing how digital intelligence reshapes our daily interactions. His perspective offers a unique window into the ethical and structural shifts occurring within the mobile industry, particularly as global giants pivot toward autonomous systems. This discussion explores the transition from traditional mobile operating systems to sophisticated agent-centric models, the tactical maneuvers used to capture regional markets, and the technical benchmarks that define modern hardware performance. We examine how geopolitical friction and regulatory landscapes are inadvertently fostering a new era of localized innovation that challenges established global hierarchies.
HarmonyOS 7 marks a significant departure from the traditional app-based interface by introducing an “intent-as-service” model. How do you see this restructuring changing the way a typical user interacts with their device on a daily basis?
This shift represents a fundamental redesign of the user experience, moving away from the cumbersome process of manually navigating through layers of menus to find specific functions. By leveraging the HarmonyOS Intelligent Agent Framework 2.0, the system essentially collapses complex workflows into a single natural-language command that feels almost intuitive. Imagine a user who previously had to open three different apps to plan a trip; now, they can simply state their goal, and the system coordinates across its capabilities to get it done. This isn’t just a minor update, but a generational inflection point that Richard Yu has mapped out from the birth of the OS in 2019 to this new “Agent era” in 2026. The emotional weight of this change is profound because it transforms the smartphone from a tool you operate into a proactive partner that understands your context and needs.
The heart of this new system is Xiaoyi, which has evolved from a basic voice assistant into a system-level intelligence agent. What does it mean for the broader ecosystem when an assistant can control over 2,100 system-level capabilities?
When an assistant gains control over 2,100 system-level functions while simultaneously coordinating with 2,000 third-party AI agents, the entire concept of an “app” begins to dissolve. This creates a deeply integrated web where services from partners like Ctrip for travel or Ant Medical for health data are woven directly into the fabric of the OS rather than sitting in isolated silos. For the developer, it means their service is no longer just a destination but a capability that can be summoned by the OS whenever the user’s intent requires it. This level of coordination is remarkably ambitious, aiming for a task execution rate of over 90%, which suggests a level of reliability that could finally make voice and intent-driven interfaces the primary way we use technology. The sheer scale of this integration provides a structural advantage in a market that demands highly localized and efficient digital services.
Looking at the technical foundation, the openPangu 2.0 model features some staggering numbers, including a Pro version with 505 billion parameters. How do these specifications, alongside the upcoming on-device models, impact the actual performance and privacy of the hardware?
The technical specifications of openPangu 2.0 are designed to handle immense complexity, with the 505 billion parameter Pro version and the 92 billion parameter Flash variant providing the heavy lifting for sophisticated queries. What is particularly impressive is the 512K context window, which allows the AI to maintain a massive amount of information in its “short-term memory” during a conversation. By autumn 2026, we will see 30 billion parameter models running directly on Kirin chips, which is a massive win for privacy and speed since data doesn’t have to leave the device. Huawei’s benchmarks already show a performance improvement of more than 15% over the previous version, making the entire interface feel snappier and more responsive. This hardware-software synergy ensures that the intelligence isn’t just a gimmick but a core component that makes the device faster and more secure for the end-user.
In terms of market dynamics, HarmonyOS has recently overtaken iOS in China, reaching a 19% market share in early 2026. Why is the absence of certain global AI features in the Chinese market such a pivotal moment for local operating systems?
The timing of this market shift is critical because it highlights a rare moment where a global leader like Apple has been forced to leave an AI gap that local players are more than happy to fill. When it was confirmed that Siri AI would not launch in China, it created a vacuum for a sophisticated, localized AI agent that understands the specific nuances of the Chinese digital landscape. HarmonyOS capitalized on this by offering a platform that is fully optimized for domestic services, which explains why they grew to 19% while iOS sat at 16% in Q1 2026. This trajectory isn’t just about features; it’s about being present and functional in a market where regulatory friction has slowed down international competitors. It is a classic case of necessity driving a homegrown version to become a market leader, with 90% of Huawei devices now running this independent architecture.
While the growth is impressive, there are still hurdles like the app count and international expansion. How do you reconcile the aesthetic similarities with other major operating systems against the very different underlying architectures?
It is fascinating to see that while the underlying architectures are pulling in opposite directions due to sanctions and regulations, the visual language is actually converging. HarmonyOS 7 has adopted a “Liquid Glass” aesthetic that mirrors what we see in iOS or Samsung’s One UI, suggesting that there is a global consensus on what a “premium” digital experience should look like. However, the real challenge lies in the ecosystem’s scope, as 400,000 applications is a significant achievement but still only a fraction of what the global App Store offers. For the platform to move beyond its current aspirational international status, it will need to prove that its “intent-as-service” model is so superior that users are willing to overlook the smaller app library. The tension between this familiar visual surface and the radical, sanctions-built engine underneath defines the current state of mobile innovation.
What is your forecast for the future of mobile AI and the global OS market?
I believe we are entering a period where the “OS wars” will no longer be fought over who has the best grid of icons, but who has the most capable and trustworthy digital agent. By 2026, the success of HarmonyOS in China will likely serve as a blueprint for other regions looking to build “sovereign” digital ecosystems that are independent of Silicon Valley’s giants. We will see a shift where on-device processing becomes the gold standard, as evidenced by the push for 30 billion parameter models on local silicon, which balances the demand for high-level intelligence with the absolute necessity of data privacy. Ultimately, the winners will be the platforms that can turn 2,000 disparate services into one seamless, invisible assistant that feels like a natural extension of the user’s own mind.
