🥰

Here are a few of the pieces of advice that I shared:

Announcing two new features in @GoogleColab! 🚀 1️⃣ Visualization Mode: Build dashboards using plain language with Gemini. 2️⃣ Fullscreen Sharing: Share interactive outputs and dashboards instantly with one link. Learn more in our launch blog below 👇

Designed to capture the eye. Engineered to awaken every sense. Lamborghini Temerario in Blu Marinus. #Lamborghini #Temerario

Learn more about SOLAR-1 and how to see the observations of the solar wind from our new NOAA operational spacecraft by watching this SWPC video posted 10 June, 2026 @NWSWPC @NOAA @NOAASatellites

You can read more about this here: blog.google/innovation-and-a…

There is a lot going on right now on the @Space_Station, but fortunately we are all safe and witnessed a spectacular southern aurora show yesterday thanks to a recent solar event.

Bring the power of Colab straight to your terminal! ⚡️ Access GPUs/TPUs, run scripts remotely, and supercharge your AI agents with the new Colab CLI. Dive into the details from @googledevs below! 👇

Meeting notes just got a serious upgrade. 📝 You can now toggle specific sections (like Decisions or Next Steps) with Take notes for me in Meet. Plus, our new “Decisions” tracker helps you categorize outcomes like Aligned or Needs Further Discussion so nothing falls through the cracks. Learn more in our latest drop → goo.gle/4o28ias

What seemed impossible yesterday is what we put into orbit today. At the NRO, we are charged with turning visionary concepts into operational realities, ensuring the capabilities we will rely on tomorrow are being pioneered today.

And congrats to everyone who worked on the initial research for NeRF! 🎉

glTF 2.1 github.com/KhronosGroup/glTF…

Actually the implementation is just one loss function with less than 30 lines. You would replace the nerf rgb loss with this loss instead. The snippet will be uploaded shortly to GitHub.

1/ Today at #GoogleIO, we’re releasing Gemini 3.5, our latest family of models combining frontier intelligence with action. We’re starting by releasing 3.5 Flash, which is built to help you execute complex, long-horizon agentic workflows. Gemini 3.5 Flash is our strongest model for coding and agent yet.It outscores 3.1 Pro on agentic and coding benchmarks like Terminal-Bench and MCP Atlas, while running 4x faster than other frontier models. Used in Google Antigravity, 3.5 Flash is even further optimized to be up to 12x faster. It’s a powerful engine to deploy sub-agents that collaborate, run high-frequency iterative loops, and solve real-world problems at scale. Some highlights we’re excited about 🔽

Scientists INTENTIONALLY Trigger 8,000 Earthquakes Deep Beneath Swiss Alps modernity.news/2026/05/12/sc… Controversial experiment raising questions about unintended consequences.

💕

🥰

Our team ran a verifiable quantum algorithm that probes how parts of a quantum system interact, from molecules to magnets and beyond. On our Willow chip, it ran 13,000× faster than the best classical supercomputers. A first in quantum computing → goo.gle/42z9E2d

Today, we’re thrilled to announce the launch of Shapefile and 3D model import support on Google Earth. Plus, we’ve also added elevation profiles to the measure tool just for… good measure! goo.gle/4w1MPC0 You’ve made it clear. You want to be able to bring more of your data and models to Google Earth’s real-world canvas. We're excited to take the next step in delivering on that promise with these new features: 🔶 Shapefiles (SHP): Render industry-standard geospatial data as performant, cloud-native layers. 💡 3D Models (GLB): Place custom architectural mock-ups, massings, and more in a visually rich context. 🗻 Elevation Profiles: View detailed terrain data in the measure tool you already know and love. These new features are live now on Google Earth!

The video of my conversation with Amin Vahdat, @gilbert, and @djrosent at Cloud Next last week is now up. youtu.be/BpnJYJmbXcM?si=vUY3… Thanks for a great conversation!

I actually really enjoyed this article on continual learning. Finally had a chance to read through over the weekend. Thanks to @a16z for the spotlight on our work at @adaption_ai I think what the deep dive gets right is it correctly pinpoints one of the core questions researchers are thinking about in continual learning: what types of knowledge belong in different types of memory stores. Our own memory is stored at different rates, with short term and long term memory stored in different parts of the brain. How we achieve this type of dual storage in our current algorithms is going to be one of the most fun questions of next 2 years. The balance between parametric and non-parametric techniques and rapid innovation on the architectures, optimization and data for both will be important.