Google Open Source (@GoogleOSS) header
Google Open Source (@GoogleOSS) profile picture

Announcing new open source releases, exploring projects, sharing how we approach FOSS, and supporting communities around the world.

Mountain View, CA
Joined March 2017
Photo by Google Open Source (@GoogleOSS)Photo by Google Open Source (@GoogleOSS)Photo by Google Open Source (@GoogleOSS)Photo by Google Open Source (@GoogleOSS)Photo by Google Open Source (@GoogleOSS)Photo by Google Open Source (@GoogleOSS)Photo by Google Open Source (@GoogleOSS)Photo by Google Open Source (@GoogleOSS)Photo by Google Open Source (@GoogleOSS)Photo by Google Open Source (@GoogleOSS)
With 73M+ monthly PyPI downloads, kubernetes-client/python powers AI/ML from Kubeflow to Ray. How Google Cloud & @kubernetesio SIG API Machinery are modernizing it for hyperscale clusters—which feature matters most? 👇 Link in first reply
2
5
1,077
A dartastic_opentelemetry sighting in the wild. 🔭 The GenKit 1.0 announcement includes "with OpenTelemetry" and depends on dartastic_opentelemetry. This is their example using OTel with GenKit. Easy, right? github.com/genkit-ai/genkit-… GenKit uses the newly moved and updated OpenTelemetry AI semantic conventions, making sure observers (platforms, agents) talk the same language no matter which tool is using generative AI. github.com/open-telemetry/se… I made an OTel pilot for GenKit for Google but this is not my work. I am open for consulting if you want to add OTel magic on your project. For Dart and Flutter OpenTelemetry Magic, including AI that knows your users, frustration signals, janky widget identification to the source code line, and pixel-fidelity session replays of real users using your production app, see Dartastic.io
Catch up on what's new in Flutter 3.50 → goo.gle/4iTcdEf 🔗 Type-safe Native Interop is in stable 🛡️ Block untrusted Android Intent extras in release ⏱️ Performance improvements for iOS and more!
3
4
517
Google Open Source retweeted
Flutter 3.50 and Dart 3.14 are here and Emma @Twerske has the full scoop! Access native platform APIs with full type-safety from day one with Native Interop, and get ahead of the CocoaPods shutdown with Swift Package Manager. Plus, build full-stack agentic apps with Genkit Dart 1.0. #FlutterCreate
4
25
171
7,970
Google Open Source retweeted
Google has released the open-source code for the XLA Serving Engine! 🚀 This technology can process up to 6 millions requests per second and allows you to run ML tasks without Python in production. According to Google's tests, the performance is 2-5 times higher than traditional Python servers.
1
3
270
Google Open Source retweeted
Ad-hoc scripts don't cut it for proper, reproducible, reliable #AI #agent evals: for #Java & #Kotlin projects, you want agent evaluations in src/test & running in CI! So I contributed Google #ADK #Java & #Kotlin support to open-source #Dokimos (0.31.0) 🧵👇
11
5
1
15
939
A2UI 1.0 is out! 🎨 An Apache 2.0 open protocol for generative UI: agents send declarative data, and @FlutterDev, React, Angular, Lit & SwiftUI render it natively. With @GoogleCloudTech & @googledevs. 👇 Links in first reply
3
14
2
65
4,536
Launch post: goo.gle/a2ui-1-0 Docs & quickstart (npx skills add a2ui-project/a2ui): a2ui.org 6 architecture Q&As from @GoogleCloudTech: teapawt.net/GoogleCloudTech/status…
Today we are releasing A2UI 1.0, now faster, easier, and stable, an open protocol and framework for generative UI across web, Android, iOS, wearables, TVs, and any other surface. 👇 Check out what shipped
3
522
Google Open Source retweeted
Google DeepMind made its robotics simulation engine completely free and open-source. MuJoCo lets you simulate robots, physics and complex movements before testing anything in the real world. Get started in 3 steps: 1. Install: Run pip install mujoco 2. Build: Load a robot or physics model 3. Simulate: Run experiments and test how it moves You can start experimenting with robotics without owning a single robot. Repo: github.com/google-deepmind/m…
4
6
107
52,643
Why build a competing protocol when you can co-maintain an open standard? 🤝🌐 @GoogleCloudTech's Kurtis Van Gent shares how MCP Toolbox for Databases led to co-maintaining MCP & Agent Skills in @linuxfoundation's AAIF. 👇 Link in first reply
3
7
1,303
Honestly, if a tiny robot dog physically walked up to my desk and stared at me until I drank water or stood up after a 4-hour coding session, my posture would be cured immediately. 😭🐕💧 Love seeing @GoogleGemma 4 running 100% locally on an @Apple iPhone (airplane mode!) to keep wearable health data completely private, all while driving @PetoiCamp's open-source Quaddle. Incredible hackathon build! 👏
Your health companion has legs. 🐕 Abhinav Agarwal, Rithika Gurram, and Shobhit Dhamania built Bolt at the AGI House(@agihouse_org) Gemma(@googlegemma) hackathon in Silicon Valley. A Clair Health bracelet streams live data to an iPhone, Google's Gemma 4 answers questions about it right on the phone, and Quaddle, our open-source robot dog, is the companion's body on the floor. It works in airplane mode, with nothing sent to a server. Bolt's code (MIT): github.com/abhinav30219/bolt… Video by the Bolt team, who started from our Gemma + Petoi robots starter: starter repo: github.com/PetoiCamp/opencat… What would you like your AI health companion to do? cc  @DynamicWebPaige @make @irvinxyz @WeeklyRobotics @VantixAgency #PhysicalAI #Gemma #OnDeviceAI #Quaddle #OpenSourceRobotics #Robotics
3
2
1
15
2,297
Google Open Source retweeted
🥽 Write a 3D scene in Three.js, then bring it into native mobile rendering and XR without rebuilding the scene separately for every platform. Three-Live is an experimental project introduced by @GoogleOSS. It compiles supported Three.js code on-device through a sandboxed Rust runtime and connects it to Google Filament and Jetpack XR on Android, Apple RealityKit on Apple platforms, and WebGL2/WebXR in the browser. The project explicitly says it is not an officially supported Google product. Its current release is v1.4.0, published November 10. 🛠️ The useful part for AI-assisted 3D work is the feedback loop around the generated code: ▪️ Render a scene into a four-view contact sheet without requiring a GPU. ▪️ Validate it and return structured spatial information to an agent. ▪️ Compare a procedural model with a reference image using overlays and silhouette differences. ▪️ Convert a scene into GLB or USDZ. ▪️ Expose rendering, validation and conversion through an MCP server. That gives an agent something more concrete than “the code ran”: it can inspect what the object looks like, see spatial problems and revise its program. 🎮 There is also an agent gym interface for stepping through an interactive scene with actions and observations. Here, “gym” means a reset/step environment for testing behavior, not a claim that the project trains a new AI model. The XR side includes spatial traits such as grabbing, scaling and anchoring, plus runtime interaction and physics support. Browser studios let you explore the workflow before embedding the libraries in an application. This is a compiler and integration toolkit, not a promise that every arbitrary Three.js application will work unchanged. Platform capabilities, supported APIs and scene behavior still need testing on the intended device. The code is Apache-2.0. AI image generation and other connected model services have their own configuration and costs. For procedural assets, small interactive prototypes and agent-generated XR scenes, the interesting combination is portable scene code plus tools that let the agent actually inspect its output. 🧩 github.com/google/three-live #ThreeJS #XR #Procedural3D #MCP
1
5
734
185 open source leaders across 133 projects gathered in Munich for the GSoC 2026 Mentor Summit. As AI-generated PRs strain maintainers, community resilience and shared contributor guidelines are leading the way. 👇 Link in first reply
1
2
7
1,302
Google Open Source retweeted
Announcing a major milestone achieved by my team @GoogleDeepMind: we have proven key security properties of the machine code underlying the Signal messaging protocol. This is the first time the actual production code of a widely deployed protocol has been formally verified – by any method, with or without AI – and at this depth. More than a research breakthrough, it’s a prime example of how frontier AI can be a force multiplier to counter the speed and scale of modern cyber threats. Formal verification offers a higher standard of security; however, it’s extremely complex and time consuming, which is why in many cases it’s considered to be impractical. Similar efforts could take teams decades of combined effort. We did it in a month’s time. Using AlphaProof Code, Google DeepMind's new agentic framework for formal verification that runs on LLMs including Gemini Carbon, we verified the Signal protocol down to the actual code running on devices.  This milestone lays the groundwork for a future in which vital systems, from medical device firmware to power grid software, ship with mathematical proof that their protections work. Reaching that future will take continued research from our community. In that spirit, we're open-sourcing our formal specifications and verification artifacts so others can build on this work. On a personal note, this moment is of deep significance to me. I started my own research journey more than two decades ago as an intern working on checking the consistency of complex software systems. Seeing this grand challenge made practical with modern AI is extraordinary. I'm extremely proud of my team members: Swarat Chaudhuri, Andrew Ferraiuolo, Anton Kovsharov, Tej Chajed, Sam Schlesinger,  Karn Seth, Jonathan Katz, and Charles Sutton for taking on this grand challenge and delivering, and grateful to the Signal team for their partnership. deepmind.google/blog/formal-…
225
702
233
5,807
592,046
Google Open Source retweeted
📦 Run remote package executables directly using the dart run command. Starting in Dart 3.12, you can execute one-off command-line utilities without adding them to your pubspec dependencies or installing them globally. goo.gle/4igoXFE
5
22
5
195
19,176
What type of content would you like to see more of here on our X account?
1
5
1,994
#KubeCon + #CloudNativeCon NA kicks off today! 🚀 Strong communities build resilient open source. 🤝✨ Looking to connect, find mentors, or meet community groups? Check out gatherings at the Community Hub. Full schedule in reply 👇
2
2
5
1,787
Explore the community gatherings schedule and add sessions to your calendar with @CloudNativeFdn: goo.gle/kubecon-na-schedule
2
317