Big milestone: 100 episodes of GBO Live! 🎉 Huge thank you to our 20,000+ team working tirelessly every day to make our customers and partners successful. In 2019, we realized our town halls weren't dynamic enough for the world’s fastest-moving industry. Over the last 7 years, we built a format to empower the team when it counts: - Less broadcasting, more real talk - Deep insights from 450+ live guests (including customers & partners turning @Google tech into real-world impact, Nobel prize winners, Nooglers & GBOers) - Non-corporate Q&A answering 3,500+ of the team’s toughest questions Up next: GBO Live 200. Game On!

Spending time with 100+ of Google’s customers and partners @Google Zeitgeist was a reminder that our greatest moonshots are as much about people as they are about the cool tech we’re building. Huge thanks to @LorraineTwohill and the team on an incredible event! Watch a few of this year’s conversations with amazing speakers: youtu.be/EHVyk9Zc3GA?si=TIfB…

Starting today, Gemini 4 Argon is available more widely, including in Gemini for business and Gemini platform. I can’t wait to see the impact more teams will be able to deliver!

Stanford might know a little something about technology :) Fantastic discussion with GSB students from @AmyWilkinson and Charles O'Reilly’s class about companies, change, leadership, and the pace of technology!

Great customer stories! Nearly 90% of the Fortune 100 use Gemini Enterprise.

You don’t really know how resilient a system is until you push it to its absolute limit. We're taking our stress-testing to the extreme: Space 🚀 We're known for our moonshots @Google. Project Suncatcher is literally as close as we have gotten to the Moon so far. We just sent 4 TPU chips into orbit on a prototype satellite built with @Planet aboard the Transporter-18 rideshare mission with @SpaceX. To say these chips are going to go through a lot is an understatement. They have to survive radiation, cooling in a vacuum, and the intense, violent g-forces of a rocket launch. All of this helps us test how our TPUs perform in the super harsh conditions of space. Watching the team tackle these challenges is incredible. Check out this video to see what they built. It’s stunning. What’s the most extreme stress test you’ve ever put your own work through? youtu.be/o1JK79jszqo

Here's an early look at our new frontier model. Been testing it recently and you can feel the difference!

Great roundup of what's ahead on @YouTube from @nealmohan !

Single engine. Single pilot. Somewhere over Botswana, my brain asked: if the pilot passes out, could I land this thing? That night, from a tent, I built a flight simulator to find out. This summer my family and I went on the trip of a lifetime: our first-ever safari in Africa. What an awesome experience: the vastness of nature, the beauty and intensity of the animals, the kindness of the people. It was also our first flight in a tiny single-engine bush plane, which breaks my pilot friends' golden rule: double engine, double pilot. We landed safely, but with my brain trained to think through backup plans B, C, D… I was a bit nervous about the return flight, wondering what I would do if the pilot were suddenly incapacitated. (Feel free to comment that this is statistically unlikely and a waste of mental tokens!) So that night in our tent I got to work (thanks, Starlink!). I fired up Google Antigravity @_mohansolo, fed it pictures I'd taken of the cockpit, and built a working flight simulator. It wasn't one-shot: I used an older 3.x Flash model and didn't let it use tools like Blender, Three.js or Unreal. But after a few rounds of adjustments, the results were amazing. Overall it took me an hour to build, and far too many crashes to finally land. Could I have landed a real plane? Who knows. See the fun warning message on the attached screenshot. Why am I telling this story? 3 learnings: 1. Get your hands dirty. With every new generation of models, tools and harnesses, you have to push the boundaries yourself. You'll learn so much. 2. Start where nothing's (really) at stake. Tinkering privately is often easier and more fun, and what you learn transfers to your work more than you'd think. 3. This is the slowest AI will ever be. A custom simulator, from a few photos, in a tent in Botswana, in an hour. Imagine what people will build for the things that really matter. I just rebuilt the same simulator with our latest internal model, and it's incredible. Stay tuned @koraykv 😀 What's the coolest way you've privately pushed the boundaries of AI (and learned something for your work)? Let me know in the comments!

As we rightly debate the future of AI and safety, it’s great to be reminded of the many good things happening. blog.google/innovation-and-a…

Hello @NBA Draft! Loved seeing Gemini Cam in action last week – it leverages our image and video generation models to place fans into personalized, dynamic filmed scenes. Game on!

Who will craft the story that inspires a generation? And what tools will they lean on to build, scale, and distribute it to the world? At #CannesLions, the conversation is converging around the realization that technology is no longer the bottleneck. The true differentiator is mindset and a willingness to adapt. @demishassabis captured this so well on Wednesday - how AI can act as a catalyst to “express even more human creativity.” It’s not about replacing the artist; it’s about expanding the canvas! So - what are you building next?

Nothing better than your team using the latest technology to make fun of you. They surprised me with this video at a team meeting. @JoeyHelmsPhoto produced it in less than a day with our latest Gemini Omni model in Flow. The timeline for high-fidelity execution is shrinking.

At Google, we don't just build cool, cutting-edge technology in our labs; we bring it straight into business solutions that help our customers and partners grow. Look at Gemini Omni: we announced it at #GoogleIO and, the very afternoon, customers could already try some of its magic. I sat down with a few CMOs at Google Marketing Live. Hear what they’re already doing.