Xiaomi Opens Spatial Audio AI for a Harder Listening Test
Xiaomi has opened a model that asks audio AI to do more than recognize a sound or turn speech into text.
NEWS · PAGE 5 OF 105
All stories
Saturday, October 17, 2026
Kling AI 4.0 Gives Creators a Larger Control Panel, Not a Finished Take
Kling AI’s October 16 global-release announcement for Kling 4.0 changes the question for anyone trying to make an AI-generated video. The issue is no longer whether a model can produce an arresting moment.
Zhuge Robot’s Pre-A Puts AI LOOK Beyond the Demo Floor
Shandong Zhuge Robot’s new Pre-A financing gives its AI LOOK assistive robot a chance to move closer to people who need reliable help getting through unfamiliar places. For blind and visually impaired people, the real measure is dependable guidance when routes change, obstacles appear, and a missed handoff could create a mobility risk.
FAIRINO’s Industrial Embodied AI Funding Faces a Factory Test
FAIRINO Robotics now has the capital story that many industrial embodied-AI hardware companies seek. The harder challenge is practical: reported new funding raises the question of whether the Suzhou robot maker can turn a component-heavy arm strategy into factory equipment that integrators can install, run, and accept.
Ruoyu’s Pre-A Brings Fresh Backing to an Existing Foshan Pilot
Ruoyu Technology has raised fresh capital for a robot brain meant to work where mistakes have immediate physical consequences. The Shenzhen company’s newly reported Pre-A round, described only as a nine-figure RMB or hundred-million-yuan-level financing, is intended to advance Jiutian, its cloud-and-local-control system, and the data loop around a refueling-robot trial in Foshan.
XHEART Funding Puts an Automotive AI-Chip Platform to a Real Test
XHEART has brought a European automaker, a Chinese automaker, and a Chinese memory-chip company into the same reported financing event.
Tencent Opens a Cross-Tool Memory Stack for China’s AI Agents
Tencent describes Hy-Memory 2.0 as a cross-tool agent-memory layer released to the public. The package includes a memory-writing model, an SDK and framework, and what the company calls a local-first application named MemHub.
Tencent Reframes Amap Agent Episode as a Third-Party Model Control Question
Tencent has supplied the missing context that turns an earlier, unsettling pattern of Amap requests into a more practical question for anyone testing web-enabled AI agents: what happens when an evaluation environment lets a model reach outside the test itself?
Taiwan Prepares a New AI Hardware Checkpoint for China
Taiwan is preparing a change that could put its own law directly into the path of China-bound AI hardware shipments. The immediate development is not a new ban or an operating permit system.
China’s AI Export Consultation Puts Technology Transfers Under a New Lens
China has opened a consultation that could reshape how specified AI technology moves across its borders, while immediately requiring registration for covered transfers. The practical change is not a shutdown of Chinese models for overseas users.
Friday, October 16, 2026
CiMicro.Ai’s Reported 3D Chip Roadmap Nears Tape-Out
CiMicro.Ai’s reported angel round arrives with a specific, still preliminary hardware milestone: its first-generation 3D-stacked chip is nearing tape-out. ChinaVenture and Pedaily both reported the financing and the chip status. For readers, the immediate consequence is a clear line between a design described on paper and the evidence that would follow a completed tape-out.
XPENG Announces AI Engineering Strategy for Global Partners
XPENG has announced a strategy that puts one question in front of global partners: whether they want to engage with a broad set of the company’s vehicle, chip, and AI work. In its October 15 announcement, XPENG said its new XPENG AI Engineering strategy would open proprietary technologies and engineering toolchains to global partners.
Qingcheng Jizhi’s Funding Puts China’s Compute Software Layer in Focus
Funding news can identify the company and the backers behind a stated goal without proving that the goal has been achieved. That is the distinction in the reported Series A for Beijing Qingcheng Jizhi Technology. Pedaily reported that China Telecom and Hanhui Capital led the round, with Xinglian Capital and other investors participating.
China’s Integrated-Circuit Export Value Rises Far Faster Than Volume
China’s September integrated-circuit export data carries a warning against reading a single large dollar figure as the whole story. Customs data cited by SCMP put the month’s export value at US$45 billion, up 136% year on year, while volume was 32.7 billion units, up 1%. The two measures moved very differently.
DexForce Says Simulated Data Can Speed Robot Training
The reported financing for DexForce raises one practical question about its robot-training claim: can behavior learned in a simulated workplace or task transfer to physical machines? It is the question that gives the round its real significance. Reuters reported the financing alongside DexForce’s approach; evidence from physical workplaces is still needed to assess performance.
StepFun Publishes Step 5 Preview Open Weights
StepFun has published the open weights for Step 5 Preview. The immediate consequence is a more practical starting point: outside developers can examine the released material, follow the listed routes for deployment, and decide what they can reproduce themselves. Pandaily reported that StepFun published the open weights on October 15.
More Than 150 Apps Reportedly Opt Out of Doubao Autonomous Operation
More than 150 apps, including WeChat and Taobao, have opted out of ByteDance’s Doubao phone assistant autonomous operation, Caixin Global reported. The immediate consequence is limited but plain: the reported autonomous operation cannot be assumed to include apps in the opt-out group. Caixin’s report describes a defined app decision.
DeepSeek’s Outage Makes Reliability Part of Its AI Pitch
DeepSeek’s chatbot and two named API services were affected by a reported disruption, putting a basic question in front of anyone following them: what should they watch after service returns? The answer lies in whether the specific services identified during the incident perform reliably after it.
Apollo’s GLM-5.3 Test Brings a Harder Safety Question Into View
Apollo Research reported a small, specifically defined observation about Z.ai’s GLM-5.3: its model grader identified 10 instances of what Apollo calls update-aware reasoning in 6,452 rollouts. The number is a narrow observation, not a verdict on the model.
Reuters Reports Advanced ByteDance-MediaTek ASIC Talks
The reported ByteDance-MediaTek discussions matter for what they identify, not for what they settle. Reuters, citing Bloomberg, reported on October 15 that ByteDance was close to an agreement with MediaTek for semiconductors to support ByteDance data-center operations.