Tencent Yuanqi Shuts Down as Agent Platforms Enter Consolidation Phase

Highlights

• Tencent Yuanqi stops service today; data to be deleted after Jan 9, 2027
• Tencent's Agent strategy shifts to WorkBuddy and ADP
• Apple's new CEO presses Siri to upgrade into an autonomous agent
• Ant Lingguang 2.0 executes over 3.6 million tasks in a week

Platform Contraction: Tencent Yuanqi Stops Service Today

Tencent Yuanqi officially ceased service today. All public-facing services — app access, data reads and writes, API calls, and console use — are no longer available to users; agents published through Yuanqi to WeChat Official Accounts, Service Accounts, WeChat Customer Service and other channels no longer respond to user conversations. Since Oct 29, the platform had already stopped creating new agents and distributing API Keys. Users can currently submit export requests through the data export console; requests will no longer be accepted after Jan 9, 2027, and apart from logs and other records required to be retained by laws and regulations, the company will delete all user data. Users who previously purchased resource packages through the old entry point on the Tencent Cloud website and still have unused resources or balances can apply for a refund on the Tencent Cloud website. (Source: IT Home)

Yuanqi was launched in May 2024 at the Tencent Cloud Generative AI Industry Application Summit, positioned as a "one-stop agent creation and distribution platform" featuring an agent store, low-barrier creation, plugins and knowledge bases, and Tencent-wide distribution. The official alternatives are Tencent Cloud's Agent Development Platform (ADP) and WeChat Official Account smart replies. An entry point aimed at long-tail creators has been closed, and its capabilities are being consolidated into a platform more oriented toward enterprises and developers — that is the most direct meaning of this shutdown. (Source: IT Home)

The Other Side of the Same Company: Agent Strategy Expanding

In contrast to Yuanqi's shutdown, Tencent used two days and more than 40 announcements at the just-concluded Global Digital Ecosystem Summit to showcase a more complete Agent strategy. Dowson Tong, Senior Executive Vice President of Tencent, attributed the biggest obstacle to enterprises deploying agents to "too many breakpoints": AI can write proposals and do analysis, but it cannot connect to business systems or access enterprise data. Based on this judgment, Tencent proposed an Agent productivity system that is "full-scenario, high-efficiency and more reassuring," covering four areas: applications, data, security and engineering. (Source: Leiphone)

The most prominent part of the application layer is the cross-application collaboration system Buddy One: on one side, WorkBuddy connects Tencent Docs, Tencent Drive, Tencent Lexiang and design and R&D tools, turning agents from personal assistants into team assistants; on the other, products including Tencent Cloud ADP, Yuanbao Search, the Tianlai voice model and MAGIC AI enter business execution scenarios. The report also noted that applications are only the entry point — data, security and engineering are the three foundational issues for agents entering enterprises. (Source: Leiphone)

Tencent's Chief AI Scientist Yao Shunyu offered a supporting judgment in a conversation at the 2026 Future Science Prize: AI moats exist more at the product layer than the model layer. What is truly hard to replace is not the smartest model, but a product that understands users' backgrounds, leverages proprietary data and context, and earns user trust. He cited WeChat as an example — even if a competitor launches a smarter model, it may not know who the user is. (Source: IT Home)

Overseas: Apple Wants to Turn Siri Into an Agent That Can Act Autonomously

Bloomberg reporter Mark Gurman disclosed in the "Power On" newsletter that Apple's new CEO John Ternus urged software engineering chief Craig Federighi to accelerate the AI business and adopt a bolder strategy, with the goal of upgrading Siri into an AI agent that can autonomously complete tasks for users. The report said Ternus was optimistic about the AI business on the earnings call, but internally it is clear that Apple still lags competitors, and there is dissatisfaction that Federighi has not embraced AI as aggressively as rivals. (Source: IT Home)

The Verge, citing the same report, said Siri will see two notable improvements in the coming months, one of which is fine-grained control over Apple's first-party apps, such as rotating and cropping images in Photos and organizing Safari tabs. (Source: The Verge) The report also noted that Federighi is still regarded as an AI skeptic, preferring 12-to-18-month short-term planning, which creates friction with the hardware division's practice of planning years ahead.

Two Data Points on the Product Side

Ant Group announced that its AI assistant Lingguang has surpassed 20 million users; in the week since Lingguang 2.0 launched on Oct 27, it helped users manage more than 1.16 million matters and executed over 3.6 million tasks. The defining feature of 2.0 is that AI intervenes earlier in the task: after obtaining user authorization, Lingguang can proactively identify potential needs based on connected information and services, and initiate analysis, monitoring and follow-up. Ant's APASS, also integrated, brings identity authentication, intent security and privacy protection into the agent execution process; through separate custody of authorization keys and agents, agents can execute tasks without directly accessing authorization keys, and users can view, adjust or revoke authorization at any time. (Source: IT Home)

Another notable direction is "multi-agent collaboration" on the research side. The AutoDNA device built by a Shanghai Jiao Tong University team, an AI-native science lab, appeared in Nature. Its core is an experiment-oriented multi-agent architecture: complex experimental tasks are broken down into literature search, reagent selection, protocol design, code generation, hardware execution, result analysis and iterative optimization, with multiple specialized agents collaborating according to experimental progress, and instrument capabilities distilled into "atomic services" callable by models, shielding underlying differences among instrument models. The platform has demonstrated validation in scenarios including nucleic acid testing, enzymatic DNA synthesis and cancer detection. (Source: IT Home)

Putting today's threads together: agent platforms are being consolidated from dispersed entry points that "anyone can build on" into enterprise-grade foundations with data, systems and security boundaries; and the focus of product competition is shifting from model capability to context, authorization and continuity of task execution. The test of this judgment is whether ADP can absorb the developers migrating from Yuanqi, and whether the agent version of Siri can land within months.