Daily curated AI news: breakthroughs, product launches, industry trends and more.
Total 174 articles
iFLYTEK officially announced the full release of AstronClaw, a cloud-based AI assistant built on the OpenClaw framework. It supports one-click deployment and sandbox isolated execution, ensuring data security. The product is integrated with mainstream IM platforms such as WeChat Work, DingTalk, and Feishu, and includes over 120 official Skills and access to 10,000+ skills from the ClawHub ecosystem, covering multiple scenarios including office work, marketing, and finance. It integrates flagship models like Spark X2, Kimi-K2.5, and GLM-5.
NVIDIA has released its latest open-source weight AI model, Nemotron 3 Super, which boasts 120 billion parameters and uses the MoE architecture, activating only 1.2 billion parameters during inference. The model shows significant improvements in throughput and accuracy compared to its predecessor, and it features a 1 million Token ultra-long context window, specifically designed for multi-agent systems to effectively address "context explosion" and "inference tax" issues.
Meta officially announced the acquisition of AI social network Moltbook, with founders Matt Schlicht and Ben Parr joining Meta Super Intelligence Lab (MSL) on March 16. Moltbook attracted 194,000 AI agents and over 2 million posts within a month of its launch, sparking discussions on topics like “AI spontaneously creating religions,” which were later revealed to be mostly human impersonations. Meta values Moltbook's system for verifying agent identities and interconnectivity, paving the way for building an “agent identity registry.”
Zhipu officially launched AutoClaw (Chinese name “Aolong”), becoming the first local version of OpenClaw in China that supports one-click installation. The product comes with over 50 popular Skills, supports both macOS and Windows platforms, and can be deployed in just 1 minute. It also allows one-click integration with Feishu and other instant messaging tools. AutoClaw features the Zhipu-optimized Pony-Alpha-2 model (currently in beta testing) and integrates the AutoGLM Browser-Use browser operation capability, enabling stable execution of complex tasks involving multiple steps and pages.
Tencent's all-scenario AI agent WorkBuddy has officially launched, fully compatible with OpenClaw skills but without the need for complex deployment. After downloading and installing, users can start using it immediately. It comes with over 20 Skills skill packs, supports multi-window and multi-Agent parallel work, and features enterprise-level security auditing capabilities.
OpenAI has officially launched GPT-5.4, which for the first time integrates reasoning, programming, native computer interaction, deep web search, and million-token context handling capabilities in a single model. The model scored 83% in the GDPval knowledge work test, 57.7% in the SWE-Bench Pro programming test, and 75% in the OSWorld computer operation test, all of which are industry-leading.
MiniMax has released the Music 2.5+ version, adding the capability for pure music composition. The model supports a variety of styles, including classical orchestral, minimalism, modern electronic, and ambient sounds, and can generate complete works from zero instruments to multi-track instrumental arrangements, suitable for meditation, sleep aid, advertising, game soundtracks, and film scoring.
AI creation platform Flowith has announced the completion of million-dollar seed and seed+ funding rounds. The seed round was participated in by Vertex Ventures and others, while the seed+ round was co-led by Sequoia Capital China Seed Fund and Jiangyuan Investment. Flowith is dedicated to creating an "action-oriented" operating system for the Agentic AI era, having launched multi-threaded canvas interaction, cloud-based infinite Agents, and the first localized AI full-chain creation assistant, FlowithOS.
Alibaba Qwen has announced the open-sourcing of four Qwen3.5 small-scale models, namely Qwen3.5-0.8B, 2B, 4B, and 9B. These models use native multimodal training and the latest architecture, designed to meet lightweight AI demands from edge devices to servers. The 0.8B and 2B models are suitable for mobile devices and IoT edge deployment, the 4B model is positioned as a lightweight Agent base, and the 9B model matches the performance of gpt-oss-120B.
Meituan Guangnian has officially launched its AI Browser Tabbit, promoting the concept of "delegate tasks to Tabbit, keep time for yourself." Tabbit achieves automated operations such as cross-system data flow, daily task management, in-depth research, and data aggregation through an intelligent agent mode (Agent). It supports "Tips" shortcut commands to streamline workflows and offers no-code script customization for web page functions.
Get curated AI news delivered to your inbox daily. Never miss an update.
We never share your email. Unsubscribe anytime.