On July 31, 2026, Geely Auto Group CEO Gan Jiayue announced the establishment of a '2030 Lab' to strengthen forward-looking technology research in areas including acoustics, optics, comprehensive safety, power semiconductors, digital chassis, embodied intelligence, data science, large models, and AI agents. The lab aims to reserve disruptive technologies for the company's 2030 strategy. Geely also plans to fully convert its fuel vehicles to HEV, targeting monthly sales of over 30,000 i-HEV units by year-end as part of its 'Million China Star' goal. Two new methanol hybrid models, the Galaxy Starshine 6 EF and Galaxy Starship 7 EF, will launch soon and support flexible fueling with any ratio of methanol and gasoline. The Xingrui i-HEV, the first model with the i-HEV system, was priced at 10.77–11.97万元 during presale and is equipped with the Qianli Haohan H3 driving assistance system. This move is part of Geely's All-Domain AI strategy, which has evolved to version 2.0 with the World Behavior Model (WAM) for cross-domain vehicle coordination.
Industry insiders and investors from Qianjue (robot brain), Shunheng (tactile sensors), Mengyou (companion robot Ropet), and Inno Angel Fund discussed the real-world hurdles of embodied AI deployment. Key facts: Qianjue COO Xiao Haimo revealed a 1-second response gap from user expectations due to ASR–cloud–TTS latency, already serving ~100k terminals. Shunheng co-founder Shi Yuanyuan noted that precision assembly in factories requires touch for the last 0.01mm, and that large models are not yet used in their POC; she argued B2B must drive supply-chain cost reduction before home robots can scale. Mengyou CEO He Jiabin shared that Ropet faced heat dissipation and wear issues, delaying shipment from March to July/August, yet achieved over 20k units with 80%+ 90-day retention by focusing on emotional value and ‘life-sense’ rather than chat. Investor Wang Sheng stated that the early investment window for World Models is closing, that 2024 will judge real landing capability, and that home robots are at least 3 years away; he believes the brain (world model) commands the most value but the paradigm war remains unresolved. All panelists agreed that embodied AI has moved past storytelling and must now prove paying customers and real-scene performance; a consensus forecast is that robots capable of doing half of household chores at half human efficiency, running a week without failure, is at least 3 years off.
Cerebras has secured a multi-year computing power procurement agreement with OpenAI, with a potential total value exceeding $20 billion, to be delivered by 2028; the company's current wafer-scale engine WSE-3 uses a 5nm process, covers 46,225 mm², integrates 4 trillion transistors and 900,000 AI cores, peaking at 125 PFLOPS. Tesla restarted Dojo 3, relying on TSMC's InFO_SoW wafer-level packaging to integrate 25 D1 chips into a training tile. In China, a Tsinghua University team demonstrated a 12-inch wafer-scale AI chip prototype and published three ISCA papers showing up to a 3.12x average performance improvement over GPU clusters for large-model inference mapping. The Institute of Computing Technology, CAS, developed a 16-module prototype called 'Ying Tian Lake' with 1.45x and 1.78x improvements in linear algebra and inference tasks, and proposed the Ouroboros wafer-scale SRAM compute-in-memory architecture that integrates 54 GB of SRAM on a single wafer, delivering 150,000 tokens/s for Llama 13B inference. New Ziguang Group announced the 'Zixian' 3D near-memory computing architecture achieving 30 TB/s memory bandwidth and planning wafer-level integration. Wafer-level chips are moving from technical demonstrations into industrial validation.
Over the past year, several projects have taken early steps toward recursive self-improvement (RSI), where AI systems help design better AI. OpenAI's GPT-5.3-Codex, released in February 2026, deeply participates in its own training debugging, deployment management, and evaluation analysis. Anthropic disclosed that most of its internal code is now written by its Claude Code tool. Google DeepMind's AlphaEvolve (2025) leverages large models to explore and optimize algorithms for chip design and neural networks. The Darwin Gödel Machines project (2025) enables AI agents to modify their own code via evolutionary algorithms. An 'AI Scientist' system, published in Nature (March 2026), automates the entire research pipeline—idea generation, experimentation, paper writing, and peer review. While these systems mark progress, experts stress that full RSI—where AIs autonomously invent, evaluate, and improve without human guidance—remains out of reach, and human oversight is still required.
Baidu was an early leader in deep learning, bidding over $40 million for Geoff Hinton’s DNNresearch in 2012 before losing to Google. It founded the Institute of Deep Learning (IDL) in 2013 and hired AI luminaries Andrew Ng (2014) and Qi Lu (2017), launching platforms like DuerOS and Apollo. By 2021, Baidu had built large language models such as ERNIE 3.0 Titan and PLATO-XL, possessing key components for a ChatGPT-like product. However, Wenxin Yiyan launched after ChatGPT felt rushed and failed to secure a leading consumer mindshare, ceding ground to rivals like ByteDance’s Doubao and DeepSeek. Today, Baidu’s AI has become a broad industrial platform, but its consumer-facing momentum faded, symbolized by the departures of Yu Kai, Andrew Ng, and Qi Lu.
Chinese embodied AI startup BeingBeyond, founded by PKU professor Lu Zongqing, released Being-H0.8, a world action model trained on 500,000 hours of first-person human video—currently the largest such dataset in the world. The model integrates latent touch information during pre-training, supports control of over 30 robot embodiments, and achieves 30 Hz inference on edge devices. Unlike many competitors, the company focuses purely on a general-purpose embodied foundation model without building its own robots, aiming to license the model to robot makers and system integrators. The release marks a step toward a model that combines visual understanding, tactile interaction, action generation, and cross-body generalization.