AI is moving from generating a picture that looks real to understanding space, predicting change and executing action. At ECCV 2026 that trend was impossible to miss. Why Spatial Intelligence Became AI’s Route Into the Physical (Source: LeiPhone) The programme chair reported 10,473 submissions with 2,834 accepted. By topic, image and video synthesis led with … Read more
One model-architecture diagram featured both Kimi and DeepSeek at once. The GLM-5.3-Flash project config uses 34 layers of Kimi’s KDA linear attention and 11 layers of DeepSeek’s KPool-DSA sparse attention across its 45 layers. Kimi and DeepSeek Showed Up in the Same Model Architecture (Source: LeiPhone) This year a clear shift appeared in large-model architectures: … Read more
Huawei’s self-developed optical engine Hi-ONE has surfaced at its first landing point, a kilo-card-scale supernode. At a recent optoelectronics expo, HiSilicon showed the engine’s bandwidth of up to 7.2 terabits and said Hi-ONE would enter the next-generation supernode in NPO (near-packaged optics) form. Huawei calls it the world’s first AI compute supernode to use NPO. … Read more
A gimbal robot arm is a camera mount that tilts and pans. The author connected one to a Mac mini and let three top models, GPT-6 Astra, Claude Fable 5.1 and GPT-5.6 Sol, each drive it, with the same instruction: make the robot sweep its full range of motion, however you like. GPT-6 Astra, Fable … Read more
DeepSeek has open sourced a bundle of operator tools and, together with Huawei Ascend, is taking aim at Nvidia’s CUDA lock in. The release centres on TileLang, a language for writing the low level operators that actually run on silicon. TileLang lets developers write chip operators in a compact Python like syntax. (Source: Leiphone) The … Read more
D-Robotics’ Sunrise S600 packs brain and motor control on one chip, its answer to Nvidia’s Jetson (Source: OFweek Robotics) The “picks and shovels” model On 17 September D-Robotics (Digua Robotics) closed a USD 400 million Series C. Mirae Asset led, with Meituan strategic, Hefei Guotou, Nanshan Industrial, Jingquan Capital and Cathay Capital following, while Hillhouse, … Read more
On 19 August, OpenAI open-sourced Codex Harness, the layer that had long kept its name more promise than practice. For a developer community that had treated harness engineering as private craft, the mood was relief: a best practice once locked inside a few top agent products was now open. OpenAI Open-Sources Codex Harness, and the … Read more
In an AI chip market that refreshes constantly, last generation’s GPU has not fallen in price the way legacy hardware did. ‘The H100 batch we deployed in 2023 now rents for more than it did at launch,’ Stephen Balaban, co-founder and chief technology officer of cloud provider Lambda, said on The MAD Podcast. Why a … Read more
LeiPhone has learned exclusively that Yao Xingcheng, who led pre-training for Tencent’s Hunyuan large language model, has joined North American start-up Thinking Machines Lab. Tencent’s Hunyuan Pre-Training Lead Joins Mira Murati’s Thin (Source: LeiPhone) People familiar with the matter said Thinking Machines Lab offered Yao a package clearly above his Tencent income, which was already … Read more
On 22 to 23 September, Intel held its ‘Same Core, Far Intelligence’ technology and ecosystem conference in Suzhou, putting the two main AI deployment lines on stage: industry AI, and physical AI for robots. Intel’s Physical-AI Play: One Chip for the Robot Brain and t (Source: OFweek Robotics) Wang Jingjia, general manager of Intel China’s … Read more