AI News

Moonshot has completed a $500 million Series C funding round, with cash reserves exceeding $10 billion. The founder stated that they are not rushing for an IPO, and will focus on long-term technology research and development and talent incentives.

Lexiang Technology launches a new embodied intelligence brand called Yuandian Smart, and introduces its first full-size humanoid robot prototype, officially entering the field of general-purpose robots. The brand has built a product lineup covering scenarios such as outdoor and household, including the tracked outdoor robot W1, and is committed to creating comprehensive smart robot solutions for all scenarios.

Morgan Stanley predicts that by 2030, artificial intelligence will threaten about 20 million banking jobs in the financial sector of Europe, mainly affecting back-end and mid-level office positions, such as data processing and administrative work. Development of financial technology is driving traditional banks to adopt intelligent solutions to improve efficiency and reduce costs.

YuanLab.ai released the open-source multimodal large model Yuan3.0Flash, with a parameter scale of 40B. It uses a sparse mixture of experts architecture, activating approximately 3.7B parameters during inference, significantly improving efficiency. The model provides 16-bit and 4-bit weights, technical reports, and training methods, supporting secondary development and industry customization, promoting the popularization of AI technology.

Qwen AI Glasses completed its first OTA update on December 31st, adding five new features: voice notes, text and image notes, multi-intent understanding and execution, blue ring payment, and community services. Existing functions such as translation and itinerary inquiries were also optimized. With its self-developed voice enhancement model and hardware configuration, the recording function now supports clear audio capture and effective noise reduction within a ten-meter range.

The Tencent Huan Yuan team has open-sourced the HY-Motion1.0 model, which is based on DiT architecture and flow matching technology. It can generate high-quality 3D skeleton animations from text descriptions and is compatible with mainstream 3D tools, greatly lowering the threshold for animation creation. The model uses a full-stage training strategy, optimizing the generation effect with over 3,000 hours of action data.

OpenAI's Employee Compensation Sets a New High for Tech Startups, with an Average Stock Incentive of $1.5 Million, Approximately 10.51 Million RMB, Far Exceeding Google's 2003 Levels.

AI shell products are not simply applications; the key lies in whether they can integrate into users' workflow, accumulate unique data, and compete with giants, which determines their market survivability.

It has been reported that Zhou Gurui, a vice president of Kuaishou, is about to resign and may join Meta or TikTok. His internal status shows he is on leave, and his email signature has been changed to "Log Out." Zhou Gurui holds a master's degree and has a strong technical background.

The 'Two New' policy in 2026 is upgraded, for the first time including smart glasses and smart home products in the trade-in subsidy scope, promoting intelligent terminals to be integrated into the inclusive consumer system. The policy explicitly provides subsidies for four categories of digital and smart products: mobile phones, tablets, smartwatches, and smart glasses.

OpenAI's compensation data is astonishing, with average stock compensation reaching $15 million per employee, setting a new record for tech startups. This figure is 34 times higher than the average employee compensation of 18 major tech companies in the previous year before their public offerings over the past 25 years. It is expected that by 2030, the company will issue approximately $3 billion in stock compensation annually.

Google is accused of using photos from Google Photos to train its AI system without user consent, involving the privacy of 1.5 billion users. The accusation comes from competitor Proton, not a regulatory body, sparking new concerns about photo analysis and storage.

Corporate AI applications are shifting from pilot projects to scaled investments, with budgets expected to grow significantly by 2026 and focus on a limited number of proven suppliers. Investors believe that companies will consolidate resources and select winners, marking the end of the experimental phase.

JD.com and Unitree open the world's first offline store in Beijing, expanding full-channel retail for embodied AI products. JD introduces a 'rent-only' robot leasing service with affordable pricing and flexible terms to boost public adoption.....

Baidu Wenyin AI released the 2025 Annual Prompt Work. Last year's keywords were Answers. Over the past year, users sought answers from AI about life, dreams, and more, driving AI to understand human emotions and thoughts. In 2025, people turned to AI to share their expectations and dilemmas at work.

Luo Yonghao's startup, Xi Hongsixian, launched the AI product Qie Ting, which is positioned as an in-depth voice listening library. The product uses natural language processing and knowledge graph technology to structure books, providing in-depth summaries of 1-2 hours with over ten thousand words, extracting the core content of the books.

Samsung Electronics plans to issue performance bonuses of 43%-48% of annual salary to employees in the semiconductor division in early 2026, a significant increase from 14% in 2024, highlighting the strong development momentum of the semiconductor industry in the AI tide.

Sanhua Holding partners with DingTalk to enhance global collaboration, security, and AI applications, building a smart decision-making system for improved resource efficiency.....

Musk's xAI accelerates the computing power competition, acquiring a third building to expand its infrastructure in Tennessee, aiming to increase model training computing power to nearly 2 billion watts. The new building is named "MACROHARDRR," which appears to be a parody of Microsoft.

According to the latest SuperCLUE-VLM rankings, Google's Gemini-3-Pro scored 83.64 points, showing a significant lead, particularly in visual understanding and reasoning. Domestic models performed outstandingly, with SenseTime's SenseNova V6.5Pro and ByteDance's Doubao ranking second and third, showcasing China's rapid progress in the multimodal field. The evaluation covers three core capability dimensions.