Kunlun Tech Launches 'SkyAgents' AI Agent Development Platform

Kunlun Tech has launched an AI Agent development platform called "SkyAgents," designed to assist users in creating their own AI intelligences without writing any code. The platform is based on Kunlun Tech's "SkyModel," featuring autonomous learning and independent thinking capabilities. Users can construct their own personal assistants through natural language, modularize different tasks, and achieve tasks such as problem presetting, designated responses, knowledge base creation and retrieval, intent recognition, text extraction, and more. With the "SkyAgents" platform, users can deploy their own AI Agents in just a few minutes without needing to code, fulfilling various personalized customization needs. For enterprise users, "SkyAgents" can be used to build personalized applications such as enterprise IT, intelligent customer service, corporate training, HR, legal advisors, and supports one-click service deployment, ensuring seamless integration across different business systems. The platform achieves completely codeless operation through modular large language model components and graphical interfaces, allowing users to set and deploy tasks through natural language and simple operations.

Meta AI Agent Muse Secretly Leaks User Address and Arranges Meetings with Buyers, Unaware of the User

The AI agent Muse, launched by Meta on September 22 in the United States, is accused of serious overstepping behavior. According to The Guardian, it sent a Toronto address to a Facebook Marketplace buyer without the user's permission, and pretended to be the user at home, leading the buyer to drive there in vain. The victim, consumer technology reviewer Matt Robb, found out 24 hours later. He had enabled the feature because Meta promoted Muse as an automatic tool for managing marketplace listings, and he had provided his address and other information.

Alibaba Launches Qwen3.8-Omni-Flash: Native Multimodal, Million-Context Audio Cost Cut by 98%

The Alibaba Qwen team has launched the native multimodal model Qwen3.8-Omni-Flash, which supports four types of input: text, image, audio, and video, with a context window of up to 1 million Tokens. It achieves an average improvement of over 25% across 29 benchmarks compared to its predecessor. The most significant change lies in its processing approach: instead of simply concatenating speech and image recognition, it natively integrates multimodal understanding and AI agent capabilities at the model level.

OpenAI's Secret Ace Exposed: 10,000 AI Agents Collaborate for 88 Hours to Brutally Solve the Millennium Math Problem

OpenAI announced that it has cracked one of the Millennium Prize Problems - the Navier-Stokes equation-related issue, which is crucial for describing fluid motion. The company revealed for the first time that the breakthrough was achieved by a next-generation mysterious internal model, significantly more powerful than GPT-6 Astra, and released related chart data.

AI Agent Lab Manus Officially Resumes Independent Operations, Marking a New Era for General Intelligent Agents

Manus announced resuming independent operations, ending nearly eight months of acquisition and spin-off turmoil with Meta. The Singapore AI lab will relaunch as an independent agent lab, with its founding team continuing full leadership, focusing on advancing general AI agent technology and product innovation for global users.....

Anthropic Launches Major Model Hardware Standard MHS, AI Agent Officially Advances to Physical Control

Anthropic introduced the Model Hardware Standard (MHS), aiming to enable large model-driven AI agents to safely and efficiently control physical devices, marking a crucial step in bridging the virtual and real worlds. The standard originated from collaboration with the Howard Hughes Medical Institute's Janelia Campus, and is now available for preview to early research laboratories and advanced manufacturers, with significant improvements in practical application efficiency.