China’s AI Agent Playbook

Grace Hsiao, founder of the AI Pro‑Am newsletter, shared her view on a sweeping view of China’s AI landscape. Grace’s unique blend of investor‑level analysis and reporter‑level storytelling reveals a sector that is less a monolithic “China‑vs‑the‑U.S.” race and more a mosaic of ecosystems, business models, and geopolitical cleverness. In this post we break down the most compelling takeaways from their chat: the rise of AI agents, how the country’s three tech giants are turning their existing platforms into AI‑native workspaces, the divergent strategies of emerging labs, and why Singapore has become the unexpected launchpad for Chinese AI talent.

AI Agents: From Hype to Real‑World Tools

The buzz around “agentic AI” peaked in 2025, with industry leaders promising assistants that could act on our behalf. Grace admits she tasted the hype herself, spending a weekend tinkering with Cloud Code, a user‑friendly “agent‑builder” that lets non‑technical folks create custom assistants. The experience convinced her that the tech is finally moving beyond prototype.

In China, the rollout looks different from the U.S. Rather than waiting for a single giant model, companies are embedding agents directly into the services users already love:

• Minimax (one of the “four tigers”) launched the Minimax Agent – a free, AI‑native workspace that mimics Cloud Code but at a fraction of the cost (its API is ~8 % of the price of OpenAI’s Claude‑4.5).
• Alibaba repurposed its consumer app Tongyi Qin (renamed “QIN”) into a shopping assistant that can search, compare prices, and complete transactions without leaving the app, leveraging its massive ecosystem (Taobao, Fliggy, AMap, Alipay, etc.).
• Tencent is integrating agents into Yuanbao, its AI chat platform, while also exploring deeper ties with WeChat to turn the messenger into an “AI operating system.”
• ByteDance is pushing its AI assistant Doubao (the most popular AI app in China with ~300 M MAU) and plans to sprinkle agentic capabilities across TikTok’s Chinese counterpart Douyin, creating a seamless bridge between content and action.

These moves highlight a core insight: the most powerful agents are those that can pull data from an existing ecosystem, not those that live in a vacuum. By letting AI operate inside their own shopping, travel, or social platforms, Chinese giants bypass the “plug‑and‑play” limitations that OpenAI still faces.

Big Tech’s Ecosystem Play

Grace points out that the three tech titans—Alibaba, Tencent, and ByteDance—are each playing to their strengths:

Because each firm already controls a “walled garden,” the agents can act as the new operating system for their users—much like iOS is for iPhone apps. Grace notes that this “AI‑OS” model is different from the traditional notion of an OS because the AI layer is the interface, not a separate platform you install. It’s the glue that lets a user ask, “Find me a flight to Shanghai tomorrow and pay with Alipay,” and have the whole transaction complete instantly.

Start‑Ups: Niche Focus or Global Ambition?

Beyond the big three, a new wave of labs—Minimax, Moonshot, Zhipu, and Baichuan—has emerged. Their strategies fall into two camps:

• Cost‑Leader Global Play – Minimax and its sibling Drupal (yes, the AI lab) promise ultra‑cheap APIs (e.g., “if Anthropic charges $200, we charge $20”) targeting the global south, SMEs, and government clients in Southeast Asia.
• Research‑Heavy Niche Play – Moonshot (creator of the Kimi models) focuses on frontier AI research and avoids consumer apps, while Baichuan doubles down on vertical AI for healthcare.

Grace observes that the capital constraints of these start‑ups force them to either piggy‑back on the ecosystems of the big three or become ultra‑specialized service providers. The result is a vibrant but fragmented market where each player carves out a narrow moat instead of trying to out‑scale Alibaba.

Singapore: The Quiet Bridge Between East and West

One of the most surprising threads in the conversation is the role of Singapore as a safe harbor for Chinese AI firms eyeing global capital. The most high‑profile example is Manus, an agentic AI startup that moved its headquarters from Beijing to Singapore, attracted a major US investment from Benchmark, and was later acquired by Meta for roughly $2 billion.

Looking Ahead: Agents, Interfaces, and Wearables

Grace’s crystal ball points to three themes for 2026 and beyond:

• Mature Agents: The next generation of agents will finally reduce mental load, delivering end‑to‑end task completion across shopping, travel, and even workplace productivity.
• AI‑Centric Interfaces: We may see an “AI operating system” that lives on phones, voice assistants, or even smart glasses—a trend highlighted by a Grab executive who envisions AI‑enabled glasses helping delivery workers navigate streets safely.
• Hardware Synergy: China’s deep hardware supply chain gives local firms a natural advantage in bringing AI from the cloud to the physical world, from wearables to robot arms, creating new value chains beyond pure software.

Conclusion

China’s AI ecosystem shows a sector that is less a monolithic race and more a sophisticated game of ecosystem leverage, niche specialization, and geopolitical navigation. Big tech turns its existing platforms into AI‑native workplaces, start‑ups either chase low‑cost global APIs or deep research, and Singapore quietly becomes the launchpad for cross‑border exits. As agents mature and AI begins to replace the traditional operating system, the real winners will be those who can seamlessly blend data, hardware, and user experience—whether that happens on a smartphone, a WeChat mini‑program, or a pair of AI‑enhanced glasses. The AI story in China is still being written, but the chapters we’ve explored already hint at a future where intelligent assistants are as ubiquitous—and as indispensable—as the apps we use today.