Your keyboard just got demoted.

This week, OpenAI, Google, and Anthropic shipped the same idea in different wrappers: stop typing, start talking, let AI do the work.

Tell ChatGPT what you need and it'll run multiple agents at once. Show Claude a task one time and it'll repeat it forever.

Tell Gemini to handle your week and it does.

That's not some future roadmap, it's live today.

Plus a ChatGPT built around your medical history, and a smarter voice mode for Claude with real brains behind it.

Biiiig week.

If your team's still hand-typing every prompt, y'all are prolly behind whoever isn't.

We broke down all seven live on today's Everyday AI. Let's get into it.

OpenAI rolled ChatGPT Health out to every US user, ending its January waitlist. Connect Apple Health and your medical records, and it lives in a private space away from your regular chats.

It's free on every plan, though you need to be 18 or older.

That's the unlock: one place for health info scattered across different portals, EMRs, and apps, ready to compare lab results, flag medication changes, or prep you for your next appointment.

It won't diagnose anything, but it'll make you the most informed patient in the room.

Try This

Connect Apple Health to ChatGPT and ask it to summarize what's changed since your last physical. You'll walk into your next appointment already knowing what to ask.

Anthropic upgraded Claude's voice mode to run on Opus and Sonnet, not just the Haiku model it launched with, and you can now swap models mid-conversation while voice mode works with your connectors.

Paid accounts get the upgrade, free stays on Haiku with one connection, and it covers 11 languages.

It's still walkie-talkie style, listen, pause, respond, not full duplex like ChatGPT's GPT-Live, but pulling live data from Gmail or Slack mid-conversation is a big jump for anyone doing hands-free work between meetings.

Try This

Open Claude voice mode, connect one app like Gmail, and ask it something that needs that app's live data. If it delivers hands free, make voice mode your default for thinking out loud between meetings.

Microsoft launched MAI Image 2.5 Pro, its highest fidelity image model yet, built for detailed editing and precise in-image text. It's live in Microsoft Foundry, inside Copilot PowerPoints on a Microsoft 365 Copilot plan, and through the API at $5 per million text input tokens and $106 per million image output tokens.

If Copilot is the only AI tool your organization allows, this closes a gap for building decks and marketing visuals without leaving Microsoft's ecosystem. It's a clear upgrade over MAI Image 2.0, though the market's top image models likely hold an edge on raw quality.

Try This

Next time you're stuck with clip art in a Copilot deck, generate the hero image with MAI Image 2.5 Pro instead. See if it's finally good enough to skip your usual image tool.

Google released three new Gemini models in one shot: Gemini 3.6 Flash and Gemini 3.5 Flash-Lite for everyone, and a restricted Gemini 3.5 Flash Cyber for governments and trusted partners. All three are live now across Gemini, the API, AI Studio, Android Studio, and Gemini Enterprise on paid plans.

3.6 Flash burns 17% fewer tokens than its predecessor with fewer reasoning steps per workflow, and Gemini 3.5 Pro still hasn't shipped while Google quietly pretrains Gemini 4.

Cheaper wins, and it adds up fast.

Flash-Lite will power Google's search AI overviews, where speed is everything.

Try This

If you're running Gemini 3.5 Flash in production, swap in 3.6 Flash for one week and track your token spend. If costs drop and quality holds, make the switch permanent.

Google widened Gemini Spark, its 24/7 personal AI agent, from Ultra subscribers only to Pro subscribers too. It's live for paid Gemini Pro users in the US, with more countries coming soon, though Workspace business plans aren't guaranteed access yet.

Spark runs multi-step tasks on its own, research, planning, monitoring, built around tasks, skills, and schedules, and it can open, read, and edit Docs, Sheets, and Slides directly, including shared team files.

That's a meaningful shift for teams.

Try This

Give Spark one real multi-step task, like planning a trip using your Gmail and calendar. See how much busywork it clears before you even open your inbox.

Anthropic shipped Record a Skill inside Claude Cowork. Screen-record yourself doing a task once, narrate what you're doing, and Claude turns it into a reusable skill, no SKILL.md file, no prompt engineering required.

It's rolled out to all paid Claude plans, though it's desktop only, tucked inside the Cowork tab, which kills the biggest barrier to building a skill: explaining it in writing.

And because skills are shareable, one you record in Claude can run in Codex or ChatGPT too, which turns this into less of a Claude feature and more of a new way to hand off repeat work.

Try This

Screen-record yourself doing one task you repeat every week and narrate your way through it. Save it as a skill and hand it off next time instead of doing it yourself.

OpenAI launched ChatGPT Voice on desktop, powered by GPT-Live, so you can control your computer and direct multiple agents in ChatGPT Work or Codex using just your voice. It's rolling out on macOS and Windows for Plus, Pro, Business, Edu, and Enterprise plans, and it works through ChatGPT's remote feature from your phone.

On Mac, Appshots pulls in full context from whatever program you have open, not just what's visible on screen, so you can juggle several apps hands-free.

No hands.

This is the first way to direct AI agents doing real work with your voice, not just chat with one.

Try This

Open ChatGPT Voice on desktop, pull up two apps you use daily, and direct a task in each one using only your voice. Once you see it work hands free, rethink how much of your day actually needs a keyboard.