Anthropic has launched a web version of Claude Code , its rapidly growing AI assistant for programmers, which can now be used directly from your browser. Starting Monday, the new platform is available to subscribers to the Pro, Max, and Max+ plans, which cost between $20 and $200 per month. The web version is accessible from the claude.ai website, by selecting the “Code” tab, or via the iOS mobile app. Previously, Claude Code existed only as a CLI tool launched from the terminal, but the company is now focusing on enabling developers to build and manage their AI agents in a familiar

Researchers discovered 131 extensions for automating WhatsApp Web in the official Chrome store. All were being used to send mass spam to Brazilian users. According to Socket analysts, all these extensions share the same code base, design patterns, and infrastructure. Together, they have approximately 20,905 active users. “This isn’t classic malware; it’s a high-risk, automated spam campaign that violates the platform’s rules,” explains Kirill Boychenko, Socket specialist. “The code is injected directly into the WhatsApp web page, working with WhatsApp scripts to automate mass mailings and schedule them, thus bypassing spam protection.” The ultimate goal of this campaign is to send mass

Google has quietly launched Gemini 3.0 Pro , the latest development in its multimodal language model. The stated goal: to improve contextual reasoning, the quality of results, and integration with Google tools (Workspace, Chrome, Android). Evolution compared to Gemini 2.5 Pro Version 2.5 Pro had already set a standard in multimodal reasoning and handling long contexts, especially across documents in Workspace. Gemini 3.0 Pro builds on these foundations, but introduces faster inference, greater factual consistency, and better understanding of mixed inputs (graphs, PDFs, screenshots). According to internal tests on AI Studio and Vertex AI, the new model reduces errors (“hallucinations”), produces more

The Global Computing Consortium (GCC) officially announced the release of the group standard ” Uniform Basic Input/Output System (UBIOS) Infrastructure Specification” (T/GCC 3007-2025) on October 21. This is China’s first fully independent firmware framework, designed to support distributed architectures and hardware-software cooperation. The initiative represents a historic milestone for China, which thus achieves its first national standard for a complete, standardized and scalable firmware system , laying the foundation for a truly autonomous technology ecosystem. The technical document was prepared by 13 institutions and companies , including the China Electronics Standardization Institute, Huawei Technologies Co., Ltd., Nanjing Baiao Software Co., Ltd., and

OpenAI has released the ChatGPT Atlas browser for macOS. It integrates artificial intelligence directly into the interface and can perform tasks on web pages without having to switch between tabs. Atlas works like a regular browser, but with a ChatGPT sidebar . The assistant can read the page content, answer questions about the text, summarize the content, suggest next steps, and use chat memory for context. The user retains control over their privacy; browsing data is not used to train the model by default, and “browser memory” can be disabled in the settings. Agent mode has been added. This allows the assistant

Kohler enters the digital health market with an unusual product: a toilet-mounted module equipped with a camera and a recognition system that monitors health. The device, called Dekoda , costs $599; pre-orders are open now, with first shipments expected on October 21, 2025. It is the first development from Kohler Health, a new initiative that aims to integrate accurate medical measurements into everyday life. The system analyzes the contents of the toilet after each use and interprets the data to assess hydration levels, gastrointestinal function, and the presence of blood. Artificial intelligence algorithms process the data, and the results are sent to

For the first time, the U.S. Department of Homeland Security has asked OpenAI to disclose the identity of a ChatGPT user whose requests were implicated in a child pornography investigation . The warrant, issued to investigators at Homeland Security Investigations (HSI) and released after being declassified in Maine, marks the first known instance in which authorities formally require a generative AI developer to disclose data on specific requests. The investigation began with attempts to identify the administrator of a large darknet website that distributed child pornography. Undercover agents managed to contact the alleged administrator, who casually mentioned using ChatGPT. During the conversation,