• Everyday AI
  • Posts
  • Ep 743: The future of AI? 7 New AI Features that Bring us Closer to On-Demand AI Assistants

Ep 743: The future of AI? 7 New AI Features that Bring us Closer to On-Demand AI Assistants

Gemini 3.1 Flash Live Rolls Out Globally, Judge blocks Pentagon’s Anthropic ban, Leak shows new Claude ‘Mythos’ model’s power and more.

Author

Everyday AI
March 27, 2026

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Anthropic launched AI that controls your computer, and Google upgraded real-time voice. Here’s the Updates Give today’s show a watch/read/listen to learn more.

🕵️‍♂️ Fresh Finds: Google Rolls Out Search Live with Gemini 3.1 Flash, Anthropic Leaks New Claude Model, Wikipedia Bans AI-Written Content, and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Gemini 3.1 Flash Live Rolls Out Globally, Judge blocks Pentagon’s Anthropic ban, Leak shows new Claude ‘Mythos’ model’s power and more. Read on for Byte Sized News.

💪 Leverage AI: You missed a bunch of smaller AI updates this week that have a BIG impact on your day-to-day work. We break it all down. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Meta Drops TRIBE v2 Brain Model, Google Expands Gemini Music Tools with Lyria 3 Pro, TurboQuant Shakes AI Memory Market, and more. Check it here!

Ep 743: The future of AI? 7 New AI Features that Bring us Closer to On-Demand AI Assistants


Did Anthropic outship literally everyone else this week? 🚢

And chances are, you missed it.

Even though some of Google's new AI features this week will change how the world searches, Claude's new Computer Use could shape how the world works in the future.

(Or, whatever comes next that Computer Use paves the way for.)

Unless you spend countless hours each day tracking LLM changelogs, you for sure missed a handful of updates that might change everything about how you work.

Join us for Feature Fridays, as we quickly recap the new AI drops you can't miss.

Also on the pod today:

• Google’s Lyria 3 Pro: 3-minute music 🎵 
• AI voice translation in headphones 🎧
• Copilot Tasks launch on mobile 📱 

It’ll be worth your 37 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 – Agentation turns UI annotations into structured context that AI coding agents can understand and act on, Cockpit is The Operating System for AI Agents, 1DevTool puts 9 essential developer tools — console, API client, database, and more — into one window

Google Search Live — Google is rolling out Search Live globally with voice and camera support, powered by the new Gemini 3.1 Flash Live audio model.

Gemini Import Memories — You can now import your memories and chat history from other AI apps into Gemini, so it instantly knows your preferences. Want to pick up right where you left off?

Suno V5.5 Released — Suno released v5.5 with noticeably richer, more expressive vocals and better personalization.

New AI Safety Bill Pushed — Lawmakers are proposing a temporary moratorium on building new large AI data centers until federal rules on energy, water use, and climate impacts are set.

Anthropic Leaked Model — Anthropic accidentally leaked drafts showing it has a far-more-capable model, Claude Mythos or Capybara, now in limited early-access testing.

Wikipedia Bans AI — Wikipedia has banned editors from writing or rewriting articles using AI, citing frequent violations of core content policies.

Voxtral TTS Released — Voxtral TTS: a compact 4B model that generates lifelike, low-latency multilingual voices. Try it in Mistral Studio and hear the difference.

David Sacks Drops AI Advisor Role — David Sacks says he’s out as the White House’s Special Advisor on AI and Crypto. He’ll now co-chair PCAST and advise the president on broader tech issues.

1. Google rolls out Gemini 3.1 Flash Live for real-time voice AI 🗣️

Today Google launched Gemini 3.1 Flash Live, a faster, more natural-sounding audio model that boosts real-time dialogue quality across Search Live, Gemini Live, and developer previews in AI Studio.

The update improves tonal understanding, handles noisy, interruption-filled conversations better, and extends conversational memory, making voice agents more reliable for complex tasks.

2. Codex adds a clean plugin system for reusable workflows 🔌

Codex just introduced a streamlined plugin format that packages reusable skills, optional app connectors, and MCP server settings into a single installable unit, making it faster to share and standardize automation across teams and projects.

The update clarifies local and repo-based installation paths, marketplace metadata rules, and where Codex caches installed plugins, so developers can iterate locally and then promote stable workflows via marketplaces.

3. Judge blocks government ban on Anthropic’s AI use ⚖️

A federal judge barred the Trump administration from enforcing a directive that labeled Anthropic a national security supply-chain risk and banned federal agencies from using its AI models, saying the move trampled free-speech protections.

The decision orders the government to stop applying the designation and to report how it complies by April 6, signaling immediate legal pushback against executive control over AI procurement. The ruling frames the administration’s action as punitive rather than safety-driven and raises questions about how far the government can limit access to private AI tools.

4. OpenAI’s ads pilot tops $100M ARR in under two months 🤑

OpenAI’s nascent ads program has already reached more than $100 million in annual recurring revenue less than two months after its U.S. pilot launched, signaling rapid advertiser demand and a potentially new major revenue stream for the company.

The ads appear labeled at the bottom of ChatGPT answers, do not influence responses, and are withheld from users under 18 and sensitive topics such as politics and health, while OpenAI says privacy trust metrics remain unaffected.

5. Apple will let rival AI assistants run through Siri in iOS 17.1 📱

Apple’s upcoming iOS 17.1 change, revealed in recent reports, makes it possible for third-party AI assistants to integrate with Siri and be invoked through the voice assistant, marking a timely shift in Apple’s tightly controlled voice ecosystem.

The move opens the door for competitors to offer their models and features on iPhones without users having to switch apps, while Apple keeps control over key privacy and platform rules.

6. Anthropic leak reveals powerful ‘Mythos’ Claude model and security worries 💪

Anthropic accidentally exposed draft materials showing it has trained a new, more capable model called Claude Mythos, also described as part of a larger Capybara tier, and the documents say the system is being trialed with early access customers.

The leak, discovered in a public content cache and reported by Fortune, forced the company to lock down the data and acknowledge a CMS configuration error that left nearly 3,000 assets public. Anthropic’s draft warns the model raises “unprecedented” cybersecurity risks and says the rollout will prioritize defenders to help harden code before wider release.

Anthropic dropped nearly 10 new features this week, and one went so viral it clocked 75 million views without even being a new model.

Wild, fam.

Google, Microsoft, and OpenAI were all cooking too, and if you blinked you missed stuff that could change how you work Monday morning.

1. Claude Now Uses Your Actual Computer 🔥

Anthropic dropped computer use inside Claude CoWork and Claude Code, meaning Claude can now point, click, type, and navigate your entire Mac like a human at your desk.

Not a virtual browser like every other "agentic" tool out there. Your actual computer, which is a completely different paradigm for anyone running legacy software that'll never get a native AI connector.

Buggy in research preview and macOS only for now, but paired with Dispatch so you can assign tasks from your iPhone and come back to finished work.

Try This

Pro and Max subscribers: enable computer use in Claude Desktop settings, pair it with Dispatch on iOS, and send yourself a task from your phone while you step away from your screen.

2. Copilot Tasks Lands on iOS ⚡

Microsoft's agentic feature that executes multi-step tasks using its own cloud computer and browser just rolled out to the Copilot mobile app on iOS, and this is the kind of quiet drop that absolutely should not fly under the radar.

This is NOT the enterprise Microsoft 365 Copilot. It's the free consumer app, which means you can text it a task right now without asking your IT department for anything, schedule it weekly, and connect it to Gmail, Outlook, or Google Drive.

Research decks, email summaries, competitive briefs. Stuff that used to take you an afternoon.

Try This

Update the Copilot app on iOS, look for Tasks in the menu, and schedule one recurring research task to run automatically every week.

3. Google AI Music Is Actually Usable Now 🚀

Lyria 3 Pro upgrades Google's AI music from gimmicky 30-second clips to full three-minute songs with real structural awareness, so you can prompt for intros, verses, choruses, and bridges and get something that sounds like an actual track instead of royalty-free elevator music that got lost on the way to the elevator.

Available in the Gemini app now for paid subscribers: Plus gets 10 tracks a day, Pro gets 20, Ultra gets 50. Also rolling out to Google Vids, ProducerAI, Vertex AI, and the Gemini API.

Try This

In the Gemini app, open Tools and select Create music, then prompt with a specific genre, tempo, vocal style, AND song structure instead of just a vibe.

4. Any Headphones Now Translate in Real Time 🔥

Google Translate's live headphone translation finally landed on iOS, and the key thing here is that it works with literally any pair of headphones, not just $250 AI-native earbuds, which means billions of people can now use it today.

Point your phone at a conversation happening in another language and hear the translation in your ears across 70-plus languages while the speaker's tone and cadence are preserved. Previously Android-only; now iOS too, expanding to France, Germany, Japan, Spain, the UK, and more. Free.

Try This

Update Google Translate on iOS, connect any headphones, and tap "Live translate" to try Listening mode for one-way or Conversation mode for both sides in real time.

5. Google Search Live Hits 200 Countries ⚡

Gemini 3.1 Flash Live, Google's new audio model, upgraded Gemini Live with better noise filtering, tonal awareness that adjusts when you're frustrated, and double the context window, AND it's enabling Search Live to go global to 200-plus countries after being US-only since launch.

Search Live lets you open the Google app, tap Live, and have a real back-and-forth voice and camera conversation with Search instead of typing. Developers and contact center teams can access it via the Gemini Live API and Google AI Studio now.

Try This

Open the Google app on iOS or Android, tap the Live icon under the search bar, and try asking something while pointing your camera at whatever you're looking at.

6. ChatGPT's File Problem Is Finally Fixed 🚀

OpenAI launched the ChatGPT Library, a persistent file hub in the sidebar that auto-saves every document, spreadsheet, presentation, and image you upload or ChatGPT generates, and those files now survive across conversations forever until you manually delete them.

No more re-uploading the same Q3 report to five different chats. Available for Plus, Pro, and Business subscribers globally except the EU, Switzerland, and UK. Quick privacy note: turn off "Improve the model for everyone" in settings or your files may go toward model training.

Try This

Look for Library in the left sidebar at chatgpt.com, then use "Add from library" in the attachment menu instead of re-uploading next time.

7. Genspark Voice Can Run Your Whole Workflow 🔥

Genspark's Speakly ain't just voice-to-text. It's a voice entry point into the full Super Agent ecosystem, meaning you speak a task and it executes multi-step workflows across your calendar, email, Slack, Google Workspace, Notion, and Salesforce without you babysitting every step.

Genspark claims four times the efficiency of keyboard input for complex tasks, and Workspace 3.0 earlier this month wove Realtime Voice into the full agent stack. If ChatGPT's voice mode has let you down on actual execution, this is the contender worth testing next.

Try This

Free trial at genspark.ai or download Speakly at speakly.ai, connect your calendar, and try voice-commanding a scheduling task to see if the agent execution actually ho

 

Reply

or to participate.