AI News
Daily Tech Reader - AI Edition
Short 📽
Apple
Apple
Apple Unveils M6 Mac mini and Mac Studio With Siri AI Built In — Pre-Orders Open Now, Ships September 22
Apple surprised with an August desktop drop: the M6 Mac mini (starting $899) and M5 Ultra Mac Studio are now available for pre-order, both shipping with macOS 27 and Siri AI integrated into Spotlight and system-wide context menus.
Sources: Apple · 9to5Mac · CNBC · AppleInsider
M6 Mac mini Is Apple's First 2nm Chip — Neural Accelerators in Every GPU Core for On-Device AI
The M6 is manufactured on TSMC's 2-nanometer process and embeds Neural Accelerators directly in each GPU core, making the Mac mini a competitive on-device AI agent platform for developers who prefer not to run models in the cloud.
Sources: Apple · CNBC · iDropNews · MacRumors
What Comes Next: M6 MacBook Pro and iMac Expected in October — MacBook Ultra Still MIA
With today's Mac mini and Mac Studio out the door, the M6 roadmap points to a 14-inch MacBook Pro and a refreshed iMac arriving in October; Apple is skipping M6 Pro and Max chips entirely, with M7 Pro and Max penciled in for 2027.
Sources: 9to5Mac · MacRumors
Microsoft
Copilot in Excel Now Runs Python for Advanced Analysis — PowerPoint Can Reference Email to Build Presentations
The latest M365 Copilot wave (Aug 11–25) adds Python execution in Excel for statistics and simulations, and lets PowerPoint pull directly from Outlook emails when building a deck — no more manual copy-paste between apps.
Sources: Microsoft · Releasebot · A Guide to Cloud
Microsoft Launches Industry-Specific "Guild Copilots" After Successful Healthcare and Legal Pilots
Following successful pilots in healthcare and legal, Microsoft formally launched sector-tailored Guild Copilots in August, bringing domain-specific workflows and compliance frameworks to enterprise customers without requiring custom builds.
Sources: AI Conference London · Microsoft
Copilot Chat Gets GPT-5.6, Claude Sonnet 5, and Opus 5 — Model Picker Now Live for M365 Users
The Copilot Chat model picker now includes GPT-5.6, Anthropic's Sonnet 5, and Opus 5, giving enterprise users the ability to select the right model for coding, reasoning, or document-heavy tasks without leaving the M365 shell.
Sources: Microsoft · m365copilotconnection · Geeky Gadgets
Claude
Anthropic Launches Model Hardware Standard (MHS) — Claude Can Now Operate Robots, Microscopes, and Lab Equipment
Anthropic today released MHS in research preview, a shared specification that lets AI agents safely control physical hardware — from robotic arms to quantum computing rigs — cutting device integration time from weeks down to hours.
Sources: Anthropic · Bloomberg · Fortune · PYMNTS
MHS Partners Include Genentech, Carnegie Mellon, Doosan Robotics, and Amazon Web Services
Anthropic co-developed MHS with HHMI Janelia Research Campus and is opening the research preview to select partners in biotech, robotics, and quantum computing, with plans to open-source the standard after the preview concludes.
Sources: Anthropic · Fortune · Blockchain.News · AI Chat Daily
Claude Code Gets Auto Mode Permissions Tab, Linux Crash Fix, and Smarter Bash Allow-Rule Warnings
The latest Claude Code update adds a /permissions Auto mode tab, startup warnings for wildcard Bash allow rules, and fixes a glibc 2.44 startup crash affecting Arch Linux, CachyOS, and Fedora Rawhide users.
Sources: Anthropic · Releasebot
ChatGPT
OpenAI Opens São Paulo Office — Brazil Is Now the Third Largest ChatGPT Market by Weekly Active Users
OpenAI launched commercial operations in São Paulo today, citing Brazil's rank as its third-largest ChatGPT market — Brazilians send roughly 215 million messages a day, and ChatGPT Enterprise seats in the country grew fivefold year over year.
Sources: OpenAI · CNBC · Economic News Brasil
ChatGPT Adds More Temporary Chat Controls — Memory, Plugins, and Custom Instructions Now Opt-In Per Session
A new ChatGPT Android update lets users selectively enable memory, plugins, and custom instructions inside temporary chats, or save a temporary chat to history — non-personalized chats still stay fully out of memory.
Sources: OpenAI · Releasebot
OpenAI Cuts GPT-5.6 Sol API Pricing Over 20% for Three Months — Now Rolling Out to Work and Codex Credits
OpenAI dropped GPT-5.6 Sol API and credit pricing by more than 20% for the next three months, available immediately via API and rolling out to eligible ChatGPT Work and Codex credit plans; Plus, Pro, and Business subscriptions are unchanged.
Sources: OpenAI · Releasebot
Gemini
Gemini Robotics Model Shut Down August 31 — Developers Must Migrate Before Weekend Deadline
Google has confirmed the gemini-robotics-er-1.6-preview model is being shut down this Sunday, August 31; developers using it in production pipelines need to migrate to supported model IDs before the deadline.
Sources: Google · Gemini API Changelog
Gemini Omni Flash Goes GA — Video Extension, First-to-Last-Frame Interpolation, and 4K Output Now Stable
Google released gemini-omni-1.1-flash as a generally available stable model, adding video extension, two-image interpolation, and resolution control up to 4K; the existing preview endpoint deprecates September 30.
Sources: Google · Gemini API Changelog
Gemini Enterprise Gets "Transparent Thinking" GA — Real-Time Reasoning Visible Before Tool Calls
Transparent Thinking is now generally available in Gemini Enterprise, showing users the assistant's live reasoning in an expandable panel before any tool or data source is called — improving perceived latency without increasing total response time.
Sources: Google Cloud · Gemini Enterprise Release Notes
China AI
Nvidia Expands Support for DeepSeek and Qwen on RTX and DGX Systems — Warns of Possible White House Restrictions
Nvidia is rolling out day-zero RTX support for Qwen3.8-27B and DeepSeek V4 Flash while simultaneously disclosing in its Q2 SEC filing that potential U.S. government restrictions on Chinese open AI models represent a material business risk.
Sources: CNBC · Nvidia · SEC Filing
DeepSeek Releases Harness — Its First Agent Runtime Framework for Direct Code Repository Interaction
DeepSeek launched Harness, an agent runtime that lets models locate files, modify code, run tests, and iterate on fixes directly inside repositories — the startup's first move into structured agentic tooling beyond raw model releases.
Sources: Global Times · DeepSeek
Big Story
Anthropic's MHS and Apple's M6 Tell the Same Story: AI Is Finally Leaving the Browser and Entering the Physical World
On the same day Anthropic announced a standard for Claude to control factory robots and lab equipment, Apple shipped a desktop designed to run AI agents around the clock — two very different companies both betting that AI's next frontier is atoms, not text.
Sources: Anthropic · Apple · Bloomberg · Fortune · CNBC
Nvidia's China Balancing Act Is This Week's Clearest Sign That the AI Cold War Has No Easy Exits
Nvidia is simultaneously optimizing its hardware for DeepSeek and Qwen while warning investors that U.S. restrictions could gut its business — a posture that makes perfect sense commercially and illustrates why Washington's AI China policy remains paralyzed.
Sources: CNBC · Nvidia · SEC · TechRadar