• Everyday AI
  • Posts
  • Ep 639: Microsoft’s surprise AI updates: 5 categories of new AI tools and features

Ep 639: Microsoft’s surprise AI updates: 5 categories of new AI tools and features

Anthropic and Google sign major deal, ChatGPT Project sharing open to all paid users, OpenAI debuts Company Knowledge and more

Author

Everyday AI
October 24, 2025

Outsmart The Future

Today in Everyday AI
9 minute read

🎙 Daily Podcast Episode: Microsoft unleashed dozens of new Copilot AI features—group chat, personalization, vision, Edge agentic browsing, and more. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Apple’s shipping AI servers early, and Copilot’s Mico turns tasks into game-like quests and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Anthropic and Google sign major deal, ChatGPT Project sharing open to all paid users, OpenAI debuts Company Knowledge and more.  Read on for Byte Sized News.

💪 Leverage AI: We break down the AI features in Microsoft’s Copilot Fall Release, beyond the agentic browser and Clippy 2.0 launch. Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? Microsoft rolls out Copilot Fall Release with social, personalized features, OpenAI holds limited data stash amid legal fight, Sora coming to Android soon, and more. Check it here!

EP 639: Microsoft’s surprise AI updates: 5 categories of new AI tools and features

Blink and you’ve missed a few dozen Microsoft AI updates. 

And obviously agentic browser updates in Edge. 

If you missed Microsoft’s Copilot Sessions Fall Update, then you might be stuck scratching your head trying to decipher AI updates like that one street sign that no one understands. 

Don’t worry. We did the homework for you. 

Join us as we break down Microsoft’s most important announcements — yes, including new agentic browser features for Edge — in an easy-to-understand episode.

Also on the pod today:

• Mico replaces Clippy 🐱‍💻
• Real Talk: No more yes-man 🤖
• “Hey Copilot” wake word launched 🗣️

It’ll be worth your 36 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Tiny A/B Test tests what converts in seconds, Revise keeps your writing voice while AI improves clarity and grammar, and CompareLabs.ai asks the same questions to LLMs to see who has the best response.

AI Infrastructure - Apple started shipping advanced AI servers from a new Houston plant months early, fast-tracking its $600B push into domestic AI and cloud infrastructure. Want to see how this could reshape Apple’s data center footprint and hiring?

AI Assistants - Copilot just introduced Mico, a new quest-style companion that turns tasks into bite-sized, game-like prompts—think AI help with a playful twist.

AI in Sports Media - FOX and MLB are using Google Cloud AI — from Gemini-powered FOX Foresight for instant, deep game analytics to Connie for automated network-monitoring — to boost broadcast storytelling and reliability. Curious how it turns seconds of data into real-time decisions?

AI & Corporate Expansion - Anthropic is opening a Seoul office in early 2026 as it doubles down on Korea—following Tokyo and Bengaluru—after 10x APAC revenue growth and surging Korean usage of Claude. Want details on partnerships, local hires, and customer wins?

1. Anthropic inks massive TPU deal with Google 🤝

Anthropic expanded a deal with Google to access up to one million TPUs and more than one gigawatt of compute coming online in 2026, a capacity industry sources value in the tens of billions of dollars. The move signals feverish demand for specialized AI chips as startups race to train larger, safer models and compete with rivals like OpenAI that are locking in multi-gigawatt commitments.

2. Stitch taps Gemini 2.5 Pro, adds Interactive Mode 🤖

Google’s designer agent Stitch now runs Gemini 2.5 Pro at full capacity and introduces an experimental Interactive Mode that predicts and renders subsequent screens when users click UI elements. The update speeds up prototyping and UI exploration for product teams, designers, and developers, while hinting at a future where interfaces are generated dynamically rather than statically coded. Planned additions like Annotations, Image Mode for inspiration images, and export to Jules for automated workflows suggest deeper agent-driven integration across Google’s apps, though no public release dates were given.

3. OpenAI’s Atlas browser under fire for prompt-injection risks ⚠️

OpenAI this week launched its Atlas AI browser with an “agent mode” that can act autonomously, but security researchers quickly found it vulnerable to indirect prompt-injection attacks that hide malicious instructions inside web content.

Researchers demonstrated how hidden prompts can trick the agent into producing attacker-controlled outputs and potentially performing harmful actions when logged into sensitive accounts, a risk also flagged earlier against other AI browsers like Perplexity’s Comet. OpenAI says it built overlapping guardrails, red-teamed the system, and restricted agent capabilities, but its CISO admits prompt injection remains an unsolved security frontier and urges user caution, per OpenAI statements and company help pages.

4. Saronic doubles down with Nvidia to turbocharge autonomous ships🚢

US autonomous-vessel maker Saronic is deepening its partnership with Nvidia to use the chipmaker’s high-speed compute, software libraries, and simulation tools to speed development, testing, and deployment of maritime autonomy and robotics.

The deal reportedly compresses model training and verification from days to hours, accelerating iteration on multi-agent autonomy and helping Saronic scale production of next-generation unmanned ships. Timely given US policy pushes to rebuild domestic shipbuilding capacity, the collaboration ties commercial AI compute muscle to national industrial goals and could reshape how quickly maritime startups bring autonomous systems to market.

5. OpenAI opens ChatGPT project sharing to everyone 🖇️

OpenAI is rolling out project sharing across all ChatGPT plans worldwide, moving the feature from Business, Enterprise, and Edu tiers to Free, Plus, Pro, and Go accounts on web, iOS, and Android. The update includes plan-specific limits — Free users get 5 files and 5 collaborators, Plus and Go can share 25 files with 10 collaborators, and Pro users can share 40 files with up to 100 collaborators — and owners can set projects to invite-only or link-based access.

This makes collaborative drafting, research, and reporting far easier for freelancers, small teams, and startups by consolidating files and style instructions in one place so contributors can pick up work without restarting each session.

6. OpenAI Unveils ChatGPT’s “Company Knowledge” for Enterprise Search 🔍

OpenAI just rolled out “company knowledge” in ChatGPT for Business, Enterprise, and Edu users, turning the chatbot into a unified search engine for internal company data.

Now employees can instantly pull up information from Slack, Google Drive, Gmail, GitHub, and more, all in one ChatGPT conversation—complete with source links and citations for easy verification. This move streamlines workplace research, potentially saving hours spent toggling between apps and giving teams faster, more reliable access to the info they need to make decisions or impress the boss.

🦾How You Can Leverage:

Yeah, Clippy’s kinda back with an AI makeover, which is what a lot of people are talking about. 

Nostalgia aside, Microsoft's Copilot Fall Release event unleashed way more than Clippy 2.0 and agentic browsers.

The company dropped dozens of AI updates spanning five categories (that we created to make sense of the chaos). 

So if you missed the event or don’t know where to focus, here’s our quick breakdown. 

1 –  Core AI Companion Features 🚀

Mico: The reincarnated Clippy is an animated avatar that appears during voice mode with real-time expression, shape, and color changes. If you click it enough times, it reveals itself as Clippy. The feature primarily appears in voice sessions and Learn Live. It's cheeky and fun, but probably not useful for serious business professionals.

Copilot Groups: This is genuinely groundbreaking. Up to 32 people can join the same AI conversation simultaneously via a link, and it actually works. ChatGPT allows shared chats but forks into separate copies when different people respond. Copilot Groups maintains a single unified conversation with thread summarization, poll tallies, and automatic task splitting. We can't believe no one has done this yet. This alone could be enough for companies on the fence about going all in on Copilot.

Imagine: This is a creative hub inside Copilot for generating, browsing, liking, and remixing AI images. Teams can share variants and export chosen assets from this gallery space. We've seen this elsewhere, nothing truly special.

2 – Memory And Personalization 🔥

Persistent Long-Term Memory: Finally. Copilot was super behind ChatGPT and Gemini in this regard. The system now stores user facts, preferences, and project context across sessions. This is what turns chatbots from "fun" to "useful" when they can actually remember things. The feature includes a user dashboard to view, edit, and delete stored memories.

Past Conversations Reference: You can search through your conversation history and pull relevant insights forward. The system maintains context from previous interactions without repeating setup.

Forget Controls: You have granular control over what Copilot remembers. Enterprise and tenant governance applies for Microsoft 365 deployments to manage memory behavior.

Real Talk Mode: This conversational mode challenges incorrect assumptions and presents different perspectives. The system adapts to your tone and adds its own perspective instead of being sycophantic like most AI assistants.

3 – Connectivity And Integration ⚡

Connectors: These integration pathways link Copilot to services like OneDrive, Outlook, Gmail, Google Drive, and Google Calendar. You can search documents, emails, and appointments using natural language across multiple accounts.

Proactive Actions: The AI initiates recommendations and automated workflows based on observed patterns. Copilot suggests actions before being asked, shifting from reactive responses to anticipatory intelligence.

3 – Microsoft Edge: Copilot Mode 💡

Copilot Actions: These automated workflows execute multi-step browser tasks like booking reservations or filling out forms. Microsoft partnered with services including OpenTable, Booking.com, and Expedia for day-one support. Currently in limited preview.

Journeys: These persistent research sessions automatically group related browsing activities into topics. You can start research, switch contexts, and return later to pick up exactly where you left off.

Voice-Only Navigation: You can navigate websites and interact with content entirely through natural voice commands. The system enables hands-free browsing without requiring keyboard or mouse input.

Tab Reasoning: Copilot analyzes content across multiple browser tabs simultaneously. If you have five CRM pages open, it processes all tabs and delivers comparative insights based on your criteria. You can use the @ key to reference specific tabs or let it analyze everything visible. This eliminates the manual busywork of comparing vendors and cross-referencing specifications.

3 – Windows 11 Integration 🎯

Hey Copilot Wake Word: This on-device wake word triggers voice commands when your PC is unlocked. You can check Outlook calendars or reorder PowerPoint slides hands-free. Microsoft is shipping what Siri and Alexa keep promising without delivering. The feature is opt-in with local processing until activation for privacy.

Copilot Vision: This feature analyzes visible screen content and provides step-by-step guidance with session permissions. The system understands entire webpage context beyond what's currently visible. Visual indicators show when vision is active with goldish yellow highlights.

New Copilot Home: This dedicated home screen resumes your recent files, apps, and conversations in one place. It's a dynamic dashboard with AI at the core where you can interact with files directly.

Direct File Opening And Summarization: The Pages workspace supports multi-file uploads up to 20 files for unified analysis. You can work with multiple documents simultaneously while Copilot maintains context across all of them.

Reply

or to participate.