Meta launches Muse Code AI coding agent for macOS and Linux

9to5mac.com Aug 6, 2026

Meta launches Muse Code AI coding agent for macOS and Linux

Curated by Shubham · TechCapsules · Aug 6, 2026

AI-summarised brief · reviewed before publication

Meta has released Muse Code, a terminal‑based AI coding agent now in beta for macOS and Linux. The tool, powered by the Muse Spark 1.2 model, can plan changes, write and validate code, and coordinate persistent background agents throughout a session. It logs model calls, tool use, approvals, and edits, enabling work to resume after crashes. Installation requires a single command, and the agent supports built‑in commands for planning, stress‑testing, and goal‑driven development across large repositories. The release follows Meta’s earlier Muse Spark 1.1 launch and API expansion.

💡 Why It Matters

· By offering a lightweight, command‑line coding assistant that persists across sessions, Meta provides developers with a new workflow that blends AI automation with traditional terminal tools, potentially accelerating large‑scale software projects.

AI Capsules Meta

Read full article on 9to5mac.com

Google Is Killing Google Assistant on Android Devices Next Month

Google will discontinue Google Assistant on Android devices effective September 4, replacing it with Gemini, its new large‑language‑model assistant. The transition means voice commands like “Hey Google” will route through Gemini, which offers conversational depth but can be slower, less reliable, and prone to errors. Users who previously relied on Assistant for simple tasks such as timers, lighting control, or calls will now experience Gemini’s quirks. The change applies to Android phones, Wear OS, Android Auto, and connected accessories; Google Home and TVs will retain Assistant for now.

💡 The move forces millions of Android users to adapt to a less dependable AI, potentially eroding trust in voice‑controlled ecosystems and accelerating the shift toward LLM‑based assistants across the industry.

9to5mac.com · Aug 6

WhatsApp channel admins will soon get the ability to flag AI-generated media

WhatsApp is developing a feature that lets channel administrators label AI‑generated media to comply with the EU’s new transparency rules. The beta on Android shows a new “Add AI content label” option in the context menu for messages containing AI‑created images, videos, or other media. Once applied, the label appears on the message bubble and cannot be removed. The feature mirrors the process for sponsored posts and may roll out to iOS and globally where legal requirements exist.

💡 The move forces platforms to disclose synthetic content, curbing misinformation and ensuring users can distinguish AI‑generated media from authentic material.

timesnownews.com · Aug 6

Google AI Shakeup: Demis Hassabis Steps Down As DeepMind CEO, Takes New…

Google CEO Sundar Pichai announced that DeepMind founder Demis Hassabis will step down as CEO and become Chair of Google DeepMind and Chief Scientist of Alphabet. Hassabis will retain leadership of Isomorphic Labs, the company’s AI‑driven drug discovery unit. The change, revealed on X, aims to free Hassabis to focus on artificial general intelligence and broader scientific research. Pichai emphasized the importance of Hassabis’s expertise for Alphabet’s future and noted that Hassabis will remain closely connected to DeepMind’s teams.

💡 The shift signals Alphabet’s prioritization of AGI research, positioning Hassabis to steer long‑term AI strategy while DeepMind’s day‑to‑day operations continue under new leadership.

efe.com · Aug 6

Apple acusa a OpenAI de robar secretos y esta tacha la demanda…

Apple has requested a preliminary injunction in a U.S. federal court to prevent OpenAI and two former employees from accessing the company’s confidential information. The motion, filed alongside a request to expedite evidence submission, seeks to compel testimony from ex‑employees Chang Liu and Tang Yew Tan and argues that Apple would suffer “irreparable harm” without the order. OpenAI responded via its corporate blog, calling Apple’s claim “unfounded” and asserting it has no interest in Apple’s trade secrets. The dispute stems from Apple’s July 10 lawsuit alleging that OpenAI stole proprietary data to develop its own hardware, a claim that follows a 2024 partnership announcement and a recent shift in Apple’s AI strategy.

💡 The case pits two tech titans over intellectual property, potentially reshaping future collaborations and setting a precedent for how AI firms handle former employees’ access to proprietary data.

ijr.com · Aug 6

White House Reportedly Plans To Keep Trump’s AI Framework Secret

The Trump administration has completed an artificial‑intelligence oversight framework and presented it to major technology firms, but officials say it will not release the document to the public. The White House met with AI industry partners on Tuesday to discuss the framework’s testing protocols for certain AI models. Sources indicate that the plan will exclude open‑source models and that only “trusted partners” will gain early access to frontier AI systems. The administration’s executive order, issued in June, mandates the creation of this safety testing structure.

💡 By withholding the framework, the administration limits transparency on how AI safety will be enforced, potentially affecting industry standards and public trust.

therobotreport.com · Aug 6

Avnet and Weston Robot partner to launch edge AI inspection platform

Avnet and Singapore‑based Weston Robot have unveiled an edge‑AI autonomous inspection platform that mounts a backpack of Avnet compute modules onto a quadruped robot. The system, powered by AMD Ryzen AI Embedded processors, delivers up to 50 TOPS of AI performance and supports thermal, visual, 3D lidar and GPS‑denied navigation. Integrated with Weston’s fleet‑management software, the robot can patrol predefined routes, stop at inspection points, and run AI‑driven perception tasks that can be retrained and deployed on‑device. The partnership combines Avnet’s industrial compute expertise with Weston’s robotics‑as‑a‑service model, which includes quadruped, humanoid and maritime platforms, to offer real‑time anomaly detection and reduced inspection costs for complex industrial sites.

💡 By processing data locally, the platform enables immediate fault detection in environments where cloud latency or connectivity is unreliable, accelerating the shift from reactive to proactive industrial maintenance.