Google Open Source

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

[Google Open Source @GoogleOSS]

Open source is free as in puppies, not free as in beer. 🐶📦

While prior research focuses on upstream maintenance, @sophia_iv studied the effort required to USE and update open source packages at Google.

👇 Link in first reply

[Google Open Source @GoogleOSS]

@CarlessVictor @naomibrockwell We're working to make Google more open source every day.

[Daniela Petruzalek @danicat83]

If AI brings us the promise of the future, why are we still stuck in the past with chatbots in Markdown?

Most AI applications today still work like a terminal from decades ago: you type a prompt, the agent runs a tool, and it dumps a wall of text back at you. Modern agentic interfaces should not only use richer UI elements, they should also adapt to what the user needs instead of forcing everything through a static, one-size-fits-all screen.

That changes how we think about frontend development. Instead of designing hardcoded application screens upfront, we design a shared catalog of UI building blocks and let the agent assemble the interface on demand.

Google's open Agent to User Interface (A2UI) protocol was built for this. Your frontend keeps full control over styling and native components, while the agent decides what to display based on context across three primitives:

• Catalog: The menu of UI components (charts, gauges, tables, 3D views, buttons) shared between the backend agent and the frontend client.
• Surfaces: The live UI canvas that updates in real time as the agent streams structured JSON messages.
• Actions: Two-way events that turn a click on a table row, graph node, or button directly into the next turn of the conversation without typing a follow-up prompt.

In my latest article, I used A2UI and Genkit to give my diagnostic agent, AIDA, the modern glowup she deserves, with dynamic UI elements rendered on-demand and human in the loop tool approvals for the most critical operations.

Read the full article on https://t.co/lXXOZ3aOjn:
https://t.co/tPS7CFzINn

Companion repository:
https://t.co/zH49U1m4Ib

[API Evangelist @apievangelist]

What 467 Conformant A2A Agent Cards Are Actually Offering

467 providers serve a conformant A2A agent card. Nearly half came from a docs platform default with one skill: read our docs. Most of the rest are agent-native companies selling by the call.

https://t.co/zSFsGRjisx

[Richard Seroter @rseroter]

No more frameworks on top of languages? Use AI to generate whatever code an app needs to reach a goal?

In some cases, sure. But I like using certain frameworks that do smart things for me.

@GenkitFramework is definitely that for agentic @golang apps.

https://t.co/yOkf3ARz53

[Genkit @GenkitFramework]

Genkit Dart 1.0.1 is here! 🚀

What's new:
• Primitive input schemas for tools across all providers
• genkit_firebase_ai agentPlatform & custom httpClient
• Session state & interrupt resume fixes

Thanks to garvitkaushik-123! 🎉

👇

[Genkit @GenkitFramework]

Check out the full release notes and migration details:
https://t.co/GWUPjAx6uO

[Google Open Source @GoogleOSS]

@meteatamel Read the full guide and explore the A2A v1.0 CLI samples: https://t.co/MZsnPHBrRU

[Google Open Source @GoogleOSS]

Testing and serving Agent2Agent (A2A) v1.0 agents from the terminal just got simpler. 🛠️

@meteatamel walks through the official `a2a` CLI: inspecting capability cards, streaming JSONL tasks, and `--echo`/`--exec` servers.
👇 Link in first reply

[Google Open Source @GoogleOSS]

Explore the schedule: https://t.co/8UiyRvO5YA and 🎟️ Register today!

[Google Open Source @GoogleOSS]

🌐Join us at #OSSummit Japan + #AutoLinux + #EmbeddedLinux Conference Asia, December 7-9 in Tokyo! We’re proud to support the #OpenSource community as a Diamond sponsor of this can’t-miss event. Explore the schedule👇

[Google Open Source @GoogleOSS]

Explore everything happening at the Community Hub, including peer group mentoring, workshops, and networking with @CloudNativeFdn and @linuxfoundation:

https://t.co/Ai00KKuR5J

[Google Open Source @GoogleOSS]

Heading to #KubeCon + #CloudNativeCon NA? Make sure to stop by the Community Hub! 🤝✨

A welcoming space to connect with peers, join interactive workshops, and celebrate community in cloud native.

Details 👇

[Google Open Source @GoogleOSS]

Read all 9 highlights (including google/skill-reach, AI copyright & AT Protocol PDS metrics) + the Oct–Dec 2026 event calendar (@AgenticAIFdn AGNTCon + MCPCon, @github Universe & @CloudNativeFdn KubeCon NA): https://t.co/EmDw3vykNs #OpenSource #FOSS

[Google Open Source @GoogleOSS]

This Week in Open Source #20 is here! 🌐
🏛️ OSI at UNGA81 & “AI-DDoS” maintainer research (-18.18% merge rate)
🔍 RedMonk’s 24K commit audit & OSS-Fuzz-inspired security
🎙️ Gemma 4 4B DiarizationLM & ESP32 voice AI
👇 Link in first reply

[Google Open Source @GoogleOSS]

https://t.co/9norl48Xba

[Google Open Source @GoogleOSS]

⏰ Welcome to Day 2 of #AGNTCon + #MCPCon!

Another full day of AI agents, open protocols, and production tooling. Catch today’s keynotes, breakout sessions, and Solutions Showcase!

Schedule & info ⬇️

[Karl Weinmeister @kweinmeister]

Instead of going token by token, DiffusionGemma refines a 256-token block at once.

I tested it and Gemma 4 on 10 tasks:
• Did better on copy editing & upfront counts
• Missed on valid JSON & multi-step reasoning
• Slower on my Mac, much faster on a GPU https://t.co/nhm5T3ZC6i

[Google Open Source @GoogleOSS]

Read the announcement on the Google Open Source Blog and explore the Apache 2.0 Elephant-Goldfish Model plugin repository on GitHub: https://t.co/KoCVgaDMaq

[Google Open Source @GoogleOSS]

AI coding speed without structured design leads to auto-scaled mistakes.

Introducing the open source Elephant-Goldfish Model plugin—co-design specs & validate architecture with stateless AI reviewers.

👇 Link in first reply

[Google Open Source @GoogleOSS]

@GoogleQuantumAI @kaggle Could we get more out of a quantum computer simply by asking it to do less? Read the announcement & Apache 2.0 benchmark details: https://t.co/HG60ESyiKg

Competition: https://t.co/vG5rkYN8Bl

[Google Open Source @GoogleOSS]

@GoogleQuantumAI is launching a @Kaggle competition: use AI to compile smaller, more efficient quantum circuits for fusion, optimization, and chemistry—powered by kickmix, our new open source library in Qualtran. ⚛️🤖

👇 Link in first reply