My Google I/O 2025 Recap » Gadget Flow

Google I/O 2025 unveiled Gemini 2.5 Pro as the centerpiece of its AI push—powering everything from coding tools and search to real-time assistants and creative media. With breakthroughs like Project Astra, Google Beam, and Android XR glasses, the company is racing to make AI immersive, personal, and ubiquitous.

Look, I’ve been to my fair share of Google I/O keynotes—watched the first Duplex demo with my jaw on the floor, fiddled with Flutter back when it still needed an intro—but this year? This year hit different. If 2024 was about teasing generative AI potential, 2025 is Google standing on stage, flexing like it just caught all 8 Pokémon badges. (Spoiler: Gemini 2.5 Pro literally did.)

So let’s rewind to the beginning. Sundar Pichai opens the keynote at Mountain View with the line: “Every day is Gemini season at Google.” And honestly, I believe it. The pace they’re shipping things is what I’d call “borderline concerning” if I weren’t also extremely impressed.

Gemini 2.5 Pro: The Star Player

Gemini 2.5 Pro is sweeping everything from LLM benchmarks to Reddit karma. Coders are raving about it across platforms like Cursor, and someone gave it a Pokémon emulator and—no joke—it earned all 8 badges. We’re calling it API: Artificial Pokémon Intelligence now, apparently.

It’s showing up everywhere: Android Studio, Firebase, Jules (which is now in public beta!), even popping up in Chrome. “Deep Think” mode is available to trusted testers—which I’m not yet, but a person can dream. Gemini Flash 2.5 is on the way too, boasting performance in nearly every dimension with better efficiency and cost controls (hello, Thinking Budgets).

TPU Ironwood & Token Mayhem

And then there’s TPU v7: Ironwood—Google’s custom silicon that delivers 10x performance gains over the previous gen. The sheer volume of tokens processed every month—480 trillion—is 50x more than last year. This isn’t just scale; it’s absurdity in motion.

Gemini App: The New Google Assistant?

The Gemini app now boasts 400M+ MAUs (take that, Clippy), and it’s getting smarter. “Gemini Live” is now rolled out to Android and iOS—conversations last 5x longer than traditional text interactions. Plus, it now supports screen sharing, camera input, and eventually integration with Maps and Calendar. It’s basically becoming your AI life manager.

Oh, and “Agent Mode” is coming to subscribers. Show a task once, and it’ll learn it forever. Gemini is officially that intern who remembers everything and asks nothing.

Meet Beam, Astra, and Mariner

Google Beam: This one’s wild. Beam turns 2D video calls into 3D immersive conversations. Combine that with Project Starline’s real-time language translations and you’ve got the next-gen Zoom replacement—coming soon via HP collab.
Project Astra: This universal AI assistant sees, hears, understands your surroundings, and answers intelligently. Astra’s the smart glasses sidekick we’ve been promised since Google Glass flopped—and it’s actually useful now.
Project Mariner: A research prototype for agents that browse the web, multitask, and automate actions. It can learn tasks after seeing them once and will soon be doing more for us than our own browser bookmarks.

Agent-to-agent communication is coming too (Agentception?), and yes, it will support external service integrations.

Search Just Got Real

Search is entering its AI Mode era. Powered by Gemini 2.5, it will deliver full AI experiences in the results and the search bar. Think longer queries, deeper answers, and even personal context, rolling out this summer.

Shopping? You can now virtually try on clothes and even have Google handle checkout via agents to find you the best deal. Is it creepy? Yes. Is it convenient? Also yes.

Creators, Rejoice

Between Imagen 4 (for photos), Veo 3 (for videos), and Lyria 2 (for music), Google’s got a full-blown studio suite for creators. You can generate scenes, soundtracks, even dialogue—all with AI. The toolset, including a new production platform called Flow, is pretty much Pixar in a box.

And yes, there’s SynthID watermarking embedded in generated content, because Google is still trying to keep AI content identifiable (bless them).

XR, Android & Glasses

Now let’s talk hardware. Android XR is officially here—Google’s AR/VR push, finally with a name. Lightweight glasses are on the way thanks to partnerships with Samsung (Project Moohan), Gentle Monster, and Warby Parker. The demos from the backstage team looked promising, even if I’m still haunted by the ghost of Google Glass past.

The Mic-Drop Moment: FireSat

Finally, Sundar closed with FireSat—a new initiative that uses multispectral imaging to spot wildfires faster. Google claims it’ll bring down response times from 12 hours to 20 minutes. Honestly, this might be the most world-saving demo of the day.

Google I/O 2025 was equal parts futuristic, frantic, and fascinating. If Gemini 1.0 was the intro, 2.5 Pro feels like the sequel no one thought would drop this soon—but here it is, reshaping coding, search, comms, and even how we play Pokémon.

And as a company watching this unfold, all I can say is: we’re already building on top of Gemini’s momentum. From smarter workflows to creative tools, the ecosystem is maturing fast—and we’re not missing this wave.

Madhurima Nag is the Head of Content at Gadget Flow. She side-hustles as a parenting and STEM influencer and loves to voice her opinion on product marketing, innovation and gadgets (of course!) in general.