OpenAI has acknowledged user experience and operational cost issues that emerged following the release of ChatGPT Work and GPT-5.6 Sol and has begun making improvements. While the new work environment was designed to boost productivity by centering on chat, projects, and desktop interfaces, criticism was raised that existing user flows were disrupted during the actual transition process. In particular, confusion arose because the location, usage methods, and storage structure of chat and projects were not clearly communicated during the move to a desktop-centric interface. [… ]
OpenAI has begun the public rollout of its latest artificial intelligence model, GPT-5.6. Previously, GPT-5.6 was the subject of regulatory controversy as it was provided exclusively to government-approved organizations during a restricted preview period. About two weeks later, following approval for public distribution from the U.S. Trump administration, it became possible to launch the model for general users and enterprises. OpenAI CEO Sam Altman described GPT-5.6 as the most outstanding model the company has ever created. […]
OpenAI is phasing out its AI browser, ChatGPT Atlas. Introduced as a browser capable of performing tasks on the web on behalf of users, ChatGPT Atlas is entering the process of discontinuation less than a year after its launch last October. OpenAI confirmed the plan to end support for Atlas in an announcement regarding ChatGPT Work, with the target date for service discontinuation being August 9. ChatGPT Atlas is a product that combines web browsing with task automation, where AI agents [… ]
Google has introduced a feature that allows users to check whether ads displayed on Search, Google Discover, and YouTube were created or edited by artificial intelligence. Users can view information regarding how ads are created by opening the My Ad Center panel by tapping the three-dot menu or the info button next to an ad. Newly added items are labeled "Created or edited by AI." On the same screen, as before, you can block specific ads or [… ]
Google AI Studio has expanded the way existing GitHub repositories are brought into the app development workflow by adding the "Import from GitHub" feature to Build Mode. This feature is designed to convert developers' existing repositories into an executable format within Google AI Studio, enabling editing, iterative development, and deployment. The ability to import an existing codebase and convert it into a runtime-compatible structure, instead of creating a new project from scratch, is [… ]
OpenAI has unveiled a new voice model designed to make real-time conversations more natural. The key feature of this model is that the voice mode can listen to the user while speaking. Existing voice AI services often respond only after the user has finished speaking, leading to issues such as interruptions in the conversation flow or delayed responses. Through this new voice model, OpenAI aims to implement interactions closer to human conversation, and prevent interruptions or [… ]
Meta has set out to expand its generative AI services by unveiling a new AI image generation model, 'Muse Image.' Muse Image is a tool that creates images based on text input or modifies existing ones, aiming for diverse applications such as ad production, visualizing interior and decoration ideas, and creator content production. Meta enables users and brands active on its platforms, such as Facebook and Instagram, to create visual content more quickly [… ]
Antropic is expanding its AI collaboration platform, 'Claude Cowork,' to mobile and web. With this update, Claude Cowork will now be available on iOS, Android, and the web, in addition to its existing desktop apps for macOS and Windows. The service expansion begins on Tuesday and will initially be available to Claude Max subscribers. It is scheduled to be rolled out sequentially to users of other plans over the next few weeks. Claude Cowork is designed to allow users to work with AI [… ]
Tencent's Hy team has unveiled 'Hy3,' a large-scale open AI model. Hy3 is a language model with a Mixture-of-Experts (MoE) structure featuring a total of 295 billion parameters, with 21 billion parameters actually active per token. While the overall model is massive, it is characterized by enhanced computational efficiency through a method that selectively utilizes only a subset of expert models during inference. Licensed under Apache 2.0, it offers relatively open conditions for both research and commercial use. Hy3 is [… ]
Google is extending the managed agent capabilities of the Gemini API to help developers build more stable AI agents for production environments. This update focuses on strengthening the execution management, long-running task handling, and external tool connectivity required for developing Gemini API-based agents. Managed agents reduce the burden of developers directly configuring infrastructure, and model invocation, tool usage, and state management [… ]
Microsoft is introducing various features to improve usability in Windows 11 version 26H2, scheduled for release in the second half of this year. This update focuses on refining areas where users frequently experienced inconvenience, rather than making major design changes. The most notable change is the enhancement of AI-based Copilot search. The search experience is improved by integrating more naturally with Copilot when finding necessary information or performing tasks in the search bar. The Taskbar can also be customized more freely [… ]
Google's AI research tool, NotebookLM, has added a feature to summarize uploaded materials into 60-second vertical videos. Users can create short videos on specific topics by uploading documents or materials to Notebook and selecting Video and Shorts from the Studio menu. The output combines AI-generated images with narration. This feature is essentially an extension of the existing Audio Overviews and Video Overviews. NotebookLM already allows users to summarize documents like podcasts or use video explanatory materials [… ]
