Discover the AI Tools

Huawei has unveiled its AI glasses that combine cutting-edge technology with everyday wearability. Priced at 2499 CNY, these lightweight smart glasses feature a self-developed chip, Xiaoyi Intelligent Agent, and innovative functions like 'Look to Pay'. The design weighs just 35.5g - comparable to regular glasses - making them comfortable for all-day use whether you're working, commuting, or working out.

Microsoft is rolling out an AI-driven solution to the universal struggle of disorganized files. OneDrive's new 'Suggested Rename' feature, powered by Copilot, will analyze document contents and propose three sensible naming options. Launching in June for web users, it supports Office files, PDFs, images, and more—finally bringing order to digital chaos.

Chinese AI unicorn MiniMax has officially begun its journey toward an A-share listing, signing with CITIC Securities for IPO counseling. The announcement sent shockwaves through the market, with related AI stocks surging to daily limits. MiniMax's impressive 158.9% revenue growth and expanding global footprint position it as a rising star in China's competitive AI landscape.

In a bold move that could reshape how AI interacts with the internet, Cloudflare is testing a groundbreaking search API for developers. This innovation allows AI systems to fetch live web data before responding to queries, potentially solving the chronic 'outdated information' problem plaguing chatbots. While giants like Google keep their search APIs locked down, Cloudflare's open approach could democratize real-time web access for AI applications.

NVIDIA has joined forces with robotics company Unitree to create the first open-source humanoid robot design. The partnership combines NVIDIA's powerful Jetson Thor computing platform with Unitree's advanced robot body and Sharpa Wave's sensitive robotic hand. This move could dramatically speed up development of humanoid robots by giving researchers a complete, standardized system to work with. The tech is expected to launch commercially by late 2026, potentially sparking a new wave of intelligent robots for various real-world applications.

NVIDIA has launched Cosmos3, the world's first fully open-source multimodal AI model designed for physical intelligence applications. This breakthrough technology combines visual reasoning, world generation, and action prediction in a single system, potentially slashing training times from months to days. Alongside the release, NVIDIA formed the Cosmos Alliance with leading robotics and AI developers to push the boundaries of intelligent systems.

Douyin saw 44,200 AI-generated short plays added in April - 13 times more than live-action content. Yet only 0.6% cracked 100 million views, revealing a content glut where few AI productions stand out. While the technology lowers production barriers, success remains elusive for most creators in this booming but competitive space.

OpenAI is making waves again—this time by stepping into the physical world. CEO Sam Altman announced the company's robotics venture, recruiting engineers to bridge AI and hardware. Their plan? Start with infrastructure helpers, then evolve to customizable personal bots. This move signals a major shift in how AI might soon interact with our everyday environments.

China's tech landscape just got more exciting with the launch of MiniMax M3, a groundbreaking AI model that's turning heads globally. Not only does it outperform GPT-5.5 in programming tasks, but it also handles massive documents and multi-modal inputs with ease. What makes this release special? The model's revolutionary architecture delivers blazing speeds while maintaining accuracy, promising to shake up how we use AI for complex tasks.

A casual museum visit turned bizarre when a visitor's photo of a priceless artifact appeared for sale on Xianyu - complete with AI-generated descriptions and a 6,000 yuan price tag. The platform claims it was an accidental upload to their new 'Xianyu Space' feature, but the user insists she never listed it. This strange case raises serious questions about how platforms handle user content and the ethics of AI-generated listings.

NetEase Youdao's LobsterAI has made waves with its latest upgrade, bundling four powerful image and video generation models into a single open-source platform. The move brings together Seedream, Seedance, HappyHorse, and MiniMax-Hailuo technologies, offering creators an unprecedented toolkit for visual content generation. This strategic play not only expands LobsterAI's capabilities but significantly lowers the entry barrier for developers exploring multimodal AI applications.

At Computex 2026, Intel unveiled its groundbreaking OpenVINO Physical AI framework, designed to seamlessly merge artificial intelligence with robotics and industrial systems. This innovation promises to slash customization costs by offering a unified hardware-software solution while ensuring real-time local processing for safer operations in dynamic environments. The framework particularly shines in edge computing scenarios where split-second decisions matter most.

China's Xiyu Technology has unveiled its groundbreaking MiniMax M3 large language model, boasting superior programming skills, massive context handling, and native multimodal abilities. In head-to-head tests, M3 outperformed GPT-5.5 and Gemini3.1Pro in key benchmarks, while introducing innovative architecture that slashes computational costs. The company is rolling out developer-friendly pricing and plans to open-source the technology completely within days, potentially sparking a new wave of AI innovation.

As artificial intelligence drives unprecedented demand for computing power, an unexpected metal is emerging as a critical bottleneck - tin. Prices for this unassuming component of semiconductor packaging have skyrocketed 40% in just six months, fueled by AI server requirements and shrinking global supplies. With major producers cutting exports and experts warning of just 12-15 years of remaining reserves, manufacturers are racing to boost recycling efforts in what's becoming a make-or-break resource for the digital age.

NVIDIA has unveiled its groundbreaking Vera CPU, specifically designed to power intelligent agents. The chip delivers an 80% performance boost for AI tasks compared to traditional processors. Tech giants OpenAI, Anthropic, and SpaceX have already signed on as early adopters, signaling strong industry confidence in this new architecture that could reshape how AI systems operate.

A popular ChatGPT plugin for Google Sheets has been found to have dangerous security flaws that could expose your financial data. Security researchers discovered hackers can exploit hidden malicious instructions in spreadsheets, potentially gaining access to sensitive financial models and account credentials. The vulnerability allows attackers to bypass security measures and create convincing fake login screens. Experts recommend users immediately check their installed AI extensions and revoke unnecessary spreadsheet permissions.

Tianqitong, Weibo's weather service, has launched a new tool for AI developers that provides detailed weather data through major cloud platforms. The service offers everything from minute-by-minute rain predictions to 100-day outlooks, making it easier for AI applications to integrate real-time weather information. With affordable pricing starting at just 1 cent per million calls, it's set to transform how businesses use weather data in decision-making.

The open-source GodeX project has released version 1.1.0, bringing significant improvements to AI model interoperability. This update introduces deeper MiniMax-M3 integration with enhanced multimodal capabilities, smarter protocol handling, and improved Zhipu search result reconstruction. Developers can now work with multiple AI services through a unified gateway without worrying about protocol differences.

The open-source API gateway GodeX just leveled up with version 1.1.0, tackling one of developers' biggest headaches - inconsistent protocols across AI services. Its new multimodal capabilities shine with domestic models like MiniMax-M3, while innovative features transform raw web data into structured outputs. This could change how we build AI applications.

A team from The Chinese University of Hong Kong has developed SLIM, a smart framework that helps AI agents manage their external capabilities like a thoughtful gardener pruning plants. Instead of endlessly accumulating skills or forcing everything into the model, SLIM lets agents decide which skills to keep, discard, or expand based on actual performance. Early tests show this approach significantly boosts success rates in tasks ranging from home robotics to complex web searches.