Tencent Yuanqi is heading toward a shutdown that turns a creator tool into an immediate operations problem. Users who built agents on the service have a limited period to pull out configurations, workflows, and conversation records, then determine what can actually be rebuilt once Yuanqi’s publishing and API functions go dark.
Sharpa has brought three pieces of a tactile-manipulation system to IROS: the D01 integrated tactile manipulation robot, the W02 dexterous hand, and the AE01 exoskeleton data glove. The consequence is more practical than a new hardware catalog.
People.cn has placed an AI psychological-support model inside new school settings in Hebei and Henan, alongside a humanoid robot and a smart-campus terminal. The donation gives students and educators another point of contact around stress, emotions, and family communication.
ByteDance has turned a product consolidation into a near-term operating decision for TRAE users. The company says it is merging TraeWork and TraeCode into a new TRAE platform, and TraeWork will stop service on October 23. Anyone who has not upgraded by then will not be able to continue using TraeWork.
LivSyn Robotics has raised a Series A of more than RMB 100 million to pursue a stubborn industrial-robotics problem: work cells often need their data, training, and control logic rebuilt when the hardware changes. Its RUDA platform is meant to give robot builders a common layer between physical machines and AI models.
A chat response can be useful and still create more work: the user must pull material out of a scrolling exchange, move it into a document or design tool, and then explain the revision in a new prompt. IT Home and Leiphone reported that Doubao Work added an infinite canvas to task mode.
Enterprise AgentOS Separates Operations From the Model openJiuwen has put an enterprise-focused AgentOS into public code, giving teams that want to move beyond an agent demonstration a concrete stack to install and examine.
The meaningful part of XPENG’s latest robotaxi move is the invitation code. By opening online registration for XPENG YOYO, the automaker has begun taking a bounded ride-service test beyond its own employees.
RobotEra has put VPP2 first on RoboDojo-Sim’s non-agent table, a result that gives robot builders a concrete new policy to inspect but not a reason to assume reliable physical task completion. VPP2’s 39.26 average Score and 32.26% success rate lead that table.
The next consumer AI agent may not begin with a new destination app. Tencent’s Xiaowei is being tested inside Weixin, Tencent’s Chinese version of WeChat, where conversations, mini-programs, and businesses already meet.
Striding AI has set its retail robotics bet around a division of labor that is familiar to any convenience-store operator: the company positions one machine for human-facing, open-space work, another for goods movement and shelf work, and M1 for coordination.
Xiongan is asking AI and robotics companies to make a commercial bet: choose a new city as a place to build, deploy, and win customers. Its three-year 511 drive supplies a 2027 scorecard. The program seeks five unicorns, 10 companies in the RMB5 billion tier, and 100 backbone firms, according to Xiongan official reporting.
A large embodied-AI funding number can make a market look settled before a robot has spent a shift doing a customer task. For readers trying to distinguish investment activity from commercial evidence, the useful question is not which label wins a chart.
For a venture investor deciding how to assess physical AI, a polished machine is only part of the question. Business Insider reported on October 5 that U.S. venture capital investors have traveled to China, or are organizing trips there, to examine robotics, manufacturing, and physical-AI competitors firsthand.
Kuaishou’s Kling AI is reported to be putting its prospective corporate identity under sharper examination just as its video-generation business begins to produce revenue at a consequential scale.
A preliminary investigation has documented a concentrated burst of automated Amap queries whose operator remains unidentified. In its October 4 report, Swarmchasers counted 2,048 Amap reports across 216 places between September 28 and October 4. The activity was heavily concentrated on October 4, when investigators counted 1,810 reports spanning 213 places.
VA-Bench puts camera choice, robot commands, and feedback into one simulated evaluation loop. The arXiv preprint, submitted September 17, comes from authors affiliated with Dalian University of Technology and Nanyang Technological University. PanDaily reported on the work October 5. The paper follows what it calls an observe-reason-act-revise sequence.
On September 23, Malaysian Prime Minister Anwar Ibrahim met AgiBot Innovation management in Hangzhou and said the company had pledged a potential RM1 billion investment in Malaysia by 2035, alongside 5,000 potential jobs. The Edge Malaysia reported that Anwar met AgiBot and representatives of four other Chinese companies.
A review-stage IPO built around working robot categories Deep Robotics’ proposed STAR Market offering is still an inquiry-stage application. The Shanghai Stock Exchange listing record was updated September 28. The company’s application was accepted May 18, and its inquiry status dates to May 27.
China’s agricultural AI story is moving beyond the controlled work of breeding labs. A seven-part China Daily feature dated October 5 describes machinery using positioning, machine vision, artificial intelligence, and real-time data systems across breeding, irrigation, harvesting, aerial application, and phenotyping. The examples remain technology showcases rather than evidence of nationwide transformation.
