Skip to main content


Anthropic has unveiled its new AI model, Claude Sonnet 4.5, which independently ran for 30 hours to develop a chat application similar to Slack or Teams. During this time, it generated approximately 11,000 lines of code. This marks a significant improvement over the company's previous Opus 4 model, which could operate autonomously for seven hours. Anthropic views Claude Sonnet 4.5 as a major advancement in the competitive landscape of AI agents and coding technology.


Electronic Arts is set to be acquired and taken private in a $55 billion leveraged buyout led by Saudi Arabia's Public Investment Fund (PIF), Silver Lake, and Affinity Partners. Scheduled to close in Q1 of EA's 2027 fiscal year (April to June 2026), this deal is one of the largest in the gaming industry. It may significantly influence EA's strategic direction, game investments, and its workforce. The $36 billion cash component comes mainly from Silver Lake and Affinity Partners.


A still from the Critterz short film highlights advancements in generative AI. This update comes from The Stepback, a weekly newsletter analyzing pivotal tech stories. Generative AI, once limited to blurred and dreamlike images, now produces highly detailed visuals based on text descriptions. Following improvements in text-to-image models, text-to-video AI such as Runway's Gen series, Meta's Make-A-Video, and Google's Veo are advancing rapidly, aiming to revolutionize content creation in Hollywood.


Apple is internally testing a new AI chatbot named Veritas to revamp Siri's functionalities. According to Bloomberg's Mark Gurman, Veritas operates like other AI chatbots such as ChatGPT and Gemini, allowing employees to interact with it through messages, perform in-app actions, and search personal data. Despite its advanced features, Apple currently plans to keep Veritas exclusive to employees rather than releasing it to the public, a decision Gurman considers a missed opportunity. Additionally, Apple may integrate Google's Gemini for AI-powered search, signaling challenges in Apple's AI development journey.


YouTube is testing AI hosts in radio and mixes within the YouTube Music app as part of its new Labs program. These AI hosts aim to enhance the listening experience by sharing stories, trivia, and commentary about music. The program allows a limited number of US users to test experimental AI features, although access is not guaranteed. This initiative aligns with Google's broader AI efforts, including NotebookLM. Additionally, YouTube has expanded some Premium-only AI features, like the jump ahead feature, to more platforms such as TV and game consoles.


Meta is investing billions into building humanoid robots, marking it as their next major initiative in augmented reality (AR). CTO Andrew Bosworth revealed that earlier this year, under CEO Mark Zuckerberg's direction, he established a robotics research team at Meta. While the hardware challenges are manageable, the focus appears to be on advancing robotic capabilities within the AR space. This move highlights Meta's significant push into robotics technology, aiming to integrate it with its AR ambitions.


Microsoft is enhancing the Photos app on Windows 11 Insiders with a new AI feature that automatically organizes your images into specific folders such as receipts, screenshots, identity documents, and handwritten notes. This update, available on Copilot Plus PCs, categorizes photos based on their visual content, regardless of the language of any text found in the images, like sorting a Hungarian passport into identity documents. These categorized folders appear under the Categories section in the app's navigation, simplifying image management. Currently, the app recognizes and sorts these four types of images, with potential future improvements allowing users to customize their own categories, such as pets or vacation photos.


Microsoft AI CEO Mustafa Suleyman recently shared insights on the evolving role of AI in web browsers. With the AI browser landscape intensifying—Google launching Gemini in Chrome, Perplexity developing the Comet AI browser, and Atlassian acquiring The Browser Company for $610 million—Microsoft aims to innovate with its Edge browser. Suleyman highlighted Microsoft's focus on evolving Edge with AI capabilities, especially through its experimental Copilot Mode that manages tabs and facilitates tasks like restaurant bookings. The goal is to transform the browser into an AI-controlled platform, enhancing user experience seamlessly.


Google has introduced Mixboard, a new AI concepting tool designed to help users create moodboards for exploring design ideas. This Google Labs experiment offers an open canvas where users can work from pre-made board templates or generate content using text prompts powered by Google's Gemini 2.5 Flash AI model. Currently in public beta in the US, Mixboard resembles other design platforms like FigJam and Adobe Firefly Boards but adds unique AI-driven features. Users can upload images or generate visuals by describing scenes such as “show me cups, bowls, and plates in Memphis style” or “plan an autumn party in my living room.” The tool supports design exploration for home decoration, event planning, or simply curating fun image collections, similar to Pinterest boards. Additionally, Mixboard allows natural language editing to adjust, combine, or create new visuals based on uploaded images, along with one-click options to regenerate or find similar designs.


On Monday, over 200 global leaders, including former heads of state, diplomats, Nobel laureates, AI experts, and scientists, united in advocating for an international treaty on AI 'red lines' to restrict certain AI capabilities, such as impersonating humans or self-replication. This initiative, backed by more than 70 organizations and notable figures like Geoffrey Hinton and OpenAI cofounder Wojciech Zaremba, aims for a political agreement by the end of 2026 to proactively prevent irreversible risks associated with AI. Leading voices emphasize that agreeing on prohibitions for AI actions is crucial even if broader consensus on AI governance is yet to be reached. This call precedes the upcoming UN General Assembly and is spearheaded by the French Center for AI Safety, the Future Society, and UC Berkeley's Center for Human-Compatible Artificial Intelligence. While regional efforts exist, such as the EU AI Act, a global framework and an enforcing institution are deemed necessary for effective AI risk management without hindering innovation.