by Stephanie Taylor, Google Summer of Code Program Lead
Last weekend, we hosted our annual Google Summer of Code Mentor Summit in Munich, Germany, welcoming 185 attendees representing 133 open source projects across 36 countries. The past year has presented unique challenges for open source communities, particularly with the surge of AI-generated pull requests placing additional strain on maintainers.
Attendees at the GSoC 2026 Mentor Summit in Munich, Germany gathering for the annual group photo.
The 183 open source organizations participating in GSoC 2026 were at the forefront of these shifts. Each organization establishes its own AI guidelines and governance, and many continue to refine their policies as the technology evolves. Throughout the year, we have actively engaged with our communities to better understand these challenges and support them in identifying effective solutions.
Unsurprisingly, the Mentor Summit’s unconference format (where attendees create the schedule) featured numerous sessions on managing AI within both GSoC and the broader open source ecosystem. While many mentors arrived weary after a demanding season, the event fostered a strong sense of shared purpose. Discussions moved beyond common obstacles to highlight notable successes, with several organizations reporting their most impactful GSoC season yet, driven by dedicated contributors who are now transitioning into active maintainer roles.
Organizations shared a variety of strategies to address AI-assisted contributions, specifically targeting automated pull requests and supporting contributors in understanding underlying system architectures alongside AI-generated code.
Resilience emerged as a central theme of this year's summit. Open source communities have navigated dynamic shifts over the past two decades. The continued vitality and importance of open source in 2026 reflects a forward-looking mindset: one grounded in experimentation, adaptation, and continuous improvement.
For our GSoC team, the Mentor Summit underscored the remarkable collaboration across open source. Observing mentors from diverse projects exchange practical tools, workflows, and insights highlights the value of gathering community leaders to strengthen the broader ecosystem.
A huge thanks to all of the GSoC Mentors and Organization Administrators who joined us in Munich, as well as the more than 1,800 mentors worldwide who guided our 2026 contributors. Google Summer of Code wrapped up its 22nd year this week thanks to the dedication and leadership of this global community.
by Mofi Rahman, Developer Relations Engineer, and Brandon Royal, Product Manager, Google Cloud
When Google introduced Kubernetes twelve years ago, the Pod became the default unit of compute for the cloud. The design assumed a straightforward lifecycle: a program boots, binds to a port, serves traffic, and runs until you stop it.
AI agents work fundamentally differently.
An AI agent that’s running on a machine executes untrusted code, commands, and tool calls, and must be isolated from its host and its network resources. At the same time, these agents, whether they’re an interactive coding assistant, an SRE troubleshooter, or a multi-turn evaluation loop, do not run continuously. They execute a step, call a language model, and wait. Most of the time, the agent does nothing at all while it waits for a model to stream tokens or a human to review a pull request. In the same way that Kubernetes was created to address the needs of containerized applications, we introduced Agent Substrate as a new orchestrator for the emerging agentic applications. Secure by default, highly efficient and built for multi-million agent scale.
Agent Substrate was designed from the ground up for AI agent security and density and brings the best parts of Kubernetes together with a new control and data plane for fully-isolated, ultra-low-latency sandboxed compute and storage. The growing Agent Substrate community has been rapidly building core capabilities and stabilizing APIs, culminating in the latest 1.0 release.
This week at KubeCon North America in Salt Lake City, Google Cloud is donating Agent Substrate and gVisor to the Cloud Native Computing Foundation (CNCF) in Sandbox status, alongside OpenRL.
This post explains why traditional approaches struggle with stateful agents, how Agent Substrate solves the problem, how it connects to enterprise governance on Google Kubernetes Engine (GKE), and where to see it in action at KubeCon NA 2026.
Challenges of running AI agents at scale
AI agents introduce two infrastructure bottlenecks that standard Kubernetes was never built to handle: idle resource waste and control plane overload.
1. The idle resource problem
During a typical interactive session, an agent can spend 90% to 95% of its time waiting for model inference, external APIs, or human reviews.
During that wait, the agent consumes zero CPU. However, you cannot simply kill the container. The agent holds critical state in memory and on disk: active language runtimes, local Git repositories, uncommitted file edits, and tool sessions.
With standard Kubernetes, platform engineers face two poor options:
Keep a dedicated Pod running all day: The agent responds quickly because it is already warm, but you pay for idle compute around the clock. At ten thousand active users, cluster costs become unsustainable.
Scale to zero with serverless containers: You save compute costs while the agent sits idle, but resuming from cold storage forces the container to clone repositories, reinstall packages, and rebuild memory. Wakeup latency climbs past 20 seconds, breaking the interactive experience.
Platform teams need agents that wake up in under 500 milliseconds with full memory and disk state intact—without paying for running Pods while the agent sits idle.
2. Control plane overload
The opposite issue occurs when training or evaluating agents. Reinforcement learning loops and benchmark suites need to run untrusted agent code across thousands of tasks at the exact same time.
If every sandbox requires a full Kubernetes Pod, the control plane quickly grinds to a halt. Scheduling, creating, and destroying thousands of short-lived Pods every minute overwhelms etcd, saturates kube-apiserver, and exhausts subnet IP addresses. Kubernetes was built to manage stable application topologies, not to act as a microsecond sandbox allocator.
How Agent Substrate solves the problem
Agent Substrate adds an actor-based control and data plane built on top of Kubernetes. It solves the idle resource problem by decoupling Actors from Workers:
Actors hold the state: An Actor represents an individual agent session, complete with its own filesystem, environment variables, memory state, and network rules.
Workers provide the compute: A Worker is a pre-warmed Kubernetes Pod running in a shared pool (WorkerPool), equipped with container runtimes and sandbox boundaries.
The lifecycle of an agent turn
Request parking: An incoming call hits the atenet-router L7 proxy. If the target Actor is suspended, the proxy holds the request in memory and alerts the Substrate controller.
Fast placement: The controller assigns the Actor to an available slot in the warm Worker pool.
Sub-second restore: The Actor’s memory and disk state restore from tiered storage, the process wakes up, and the proxy delivers the request. The agent resumes in under 500 milliseconds right where it paused.
Suspension on idle: When the turn finishes, Substrate snapshots the Actor’s state to tiered storage and frees the Worker slot for other workloads.
Because Substrate manages Actor state directly, thousands of agents can activate, suspend, and resume every second without creating new Pod objects or stressing etcd.
Safe execution with gVisor and microVMs
Agents frequently execute model-generated, untrusted code. Standard Linux containers share the host kernel, which presents an unacceptable security risk on multi-tenant nodes.
Agent Substrate pairs with gVisor (runsc), Google's open source application kernel written in Go. gVisor intercepts system calls in user space, isolating untrusted code from the host kernel. Because gVisor virtualizes its own memory and process tables, Substrate can freeze a running environment, write memory pages to storage, and restore them on another node in under 500 milliseconds.
For workloads requiring hardware virtualization or custom kernel modules, Substrate also supports Cloud Hypervisor microVMs under the same unified API.
Why Google donated Agent Substrate to the CNCF
Over the past two years, engineering teams across the industry have run into the same agent scaling boundaries. Without a shared open standard, every organization ends up building custom schedulers on top of VMs or working around Kubernetes.
Donating Agent Substrate and gVisor to the CNCF gives the cloud-native community a neutral, vendor-independent foundation that works across any conformant Kubernetes cluster.
Together with community projects, they form an open stack for agentic infrastructure:
Layer 1: Secure Isolation (gVisor): Sandboxes untrusted code execution with user-space kernel isolation and fast memory checkpointing.
by Oleg Shaldybin and Monica Song, Google TPU Software
Today we are excited to announce the launch of XLA Serving Engine (XSE), a general-purpose model export and inference runtime for the OpenXLA ecosystem, open-sourced under the Apache 2.0 license.
We created XLA Serving Engine to enable a “Train in Python, Serve in C++” workflow. Most ML engineers develop models using Python (with frameworks like JAX and PyTorch); however, some production serving environments do not allow Python due to operational and deployment complexity. XLA Serving Engine offers a solution by providing an API for engineers to export their ML models to an “XLA SavedModel” containing the StableHLO representation of their model’s function, which can then be loaded and executed by the XLA Serving Runtime.
This approach allows users to serve models (typically non-LLMs) directly in a C++ runtime, which enables high throughput at any scale and optimizes for low latency by removing the overhead of the Python interpreter. In our studies, we found that for short requests, XSE delivers 2–5x the throughput of the best Python server we could build, and 8–36x that of a straightforward one, while using 30–70% less CPU time per prediction and a quarter of the memory.
Last year, we began deploying core components of XLA Serving Engine as part of Google’s internal load-bearing, production serving infrastructure. Today, XSE serves over 6 million queries per second and powers some of Google's highest-volume online inference pipelines, including YouTube Shorts, Google Ads, Search, and parts of Gemini Omni, with plans to onboard more customers internally.
Having successfully scaled XSE to support the high-throughput workloads of our internal teams, we are now ready to bring this core piece of infrastructure to the wider ecosystem of hardware partners and Cloud TPU users. We designed XLA Serving Engine to be self-contained and modular, so users can deploy it as a standalone inference engine or plug it into their more complex serving solutions.
Core production-ready features
Python-Free Serving: XSE enables a “Train in Python, Serve in C++” workflow via the intermediary compiled XLA SavedModel serialization format. In our studies, we found that XSE delivers 2–5x the throughput of a pure Python server.
Low-latency execution: The XLA Serving Runtime wraps PJRT directly and optimizes the model using the XLA compiler. This eliminates Python runtime overhead and the global interpreter lock. XSE is C++ native, which means one process is capable of using all eight cores.
Figure 1: Throughput under latency target
Orbax Checkpointing Integration: XLA SavedModel stores JAX model computations as StableHLO and uses Orbax checkpoint to store the weights.
Flexible Scale: Supports both single-host and multi-host serving by leveraging PJRT.
Performance Parity: XSE preserves the performance of the model’s forward function (including kernel optimizations). Hand off your model to XLA Serving Engine with confidence that performance will be identical to its training environment.
Docker Packages: We provide a suite of Docker images that are tested on CPU, GPU, and TPU and expose simple gRPC APIs for calling your model.
Customer testimonials and partner use cases
Our Private Preview partners have been actively leveraging XLA Serving Engine to bring their deep learning stacks to production. Here is a selection of private preview customers (alphabetized) to showcase the diversity of uses:
Arm: Performance Parity on Arm Neoverse platforms
Our work with Arm has proven the production capabilities of JAX and XLA for training and serving on Arm-based CPUs:
“As AI inference scales, software optimization is critical to delivering performance and efficiency across diverse workloads. Through our collaboration with Google and contributions to OpenXLA, we’ve optimized the XLA CPU backend and serving runtime for vision, embedding and recommendation workloads on Arm Neoverse-based platforms, including Google Cloud Axion.”
— Ian Bratt, Vice President of ML Technology and Fellow, Arm
Intel: Native JAX on Xeon server CPUs
We have also been partnering with Intel to showcase the power of Intel’s optimized compute:
“Our partnership with Google showcases the XLA Serving Engine (XSE) on Intel® Xeon® processors, delivering native, high-performance JAX inference accelerated by Intel® AMX. With XSE, enterprises get out-of-the-box performance across Vision, Recommendation, and LLM workloads — and the flexibility to serve them efficiently wherever their business runs, from cloud to edge to on-prem.”
— Bill Pearson, Vice President of Data Center and AI software, Intel
Moloco: Scaling advertising and recommenders
“To support our demand-side platform (DSP) processing over 650 billion daily inferences and peaking at 10.5 million QPS, we needed an inference runtime that could match our massive scale. XLA Serving Engine is the high-performance, C++-based JAX serving solution we’ve been waiting for that keeps us framework agnostic and complements our Cloud TPU Trillium investments.”
Torch Support: We plan to integrate with TorchTPU, allowing PyTorch users to export to an XLA SavedModel to serve their models on TPUs in C++.
Ahead-of-Time (AOT) Compilation: We are working on rolling out AOT compilation to our internal users before bringing this feature to external users. AOT compilation will reduce cold-start time by offering users the ability to compile their models offline (on CPU) for their target serving hardware.
C++ Data Preprocessing: We are offering a C++ based preprocessor for users who want to combine their data processing with their model’s graph in C++.
Limitations
XSE is a lightweight, one-shot inference engine. For large autoregressive language models with long context lengths, we recommend popular projects such as vLLM or SGLang, which benefit from LLM-specific memory management techniques.
Get started today
Ready to streamline your JAX production pipeline?
Check out the XLA Serving Engine quick-start instructions at github.com/openxla/xla-serving. You can find the list of available Docker images here.
If you're an ML engineer or researcher, the GPUs and TPUs you need increasingly live on Kubernetes, but Kubernetes should be an invisible layer. While your platform team sets up the GKE cluster and manages accelerator capacity and reservations, your job is to train and ship models—not to manage Kubernetes infrastructure.
Today we're introducing GKE Workspaces, a self-hosted AI/ML development platform for Google Kubernetes Engine, built on the open source Kubeflow Workspaces project. Deployed on your team's GKE cluster, it provides a dedicated web dashboard where you can pick an IDE and a machine, and start coding on cloud accelerators in minutes.
Why Kubeflow Workspaces
Kubeflow Workspaces gives every researcher a personal, persistent development environment on shared cluster hardware.
Self-service, in a few clicks. Sign in to your team's Workspaces web UI, choose JupyterLab or in-browser VS Code, pick an environment, and select a hardware size such as "Small CPU", "GPU", or "TPU". Researchers don't need Google Cloud Console access, YAML, or command-line setup.
Your work persists. Each workspace mounts its own home directory on a dedicated, encrypted Persistent Disk in your Google Cloud project—isolated to that workspace, so your files, notebooks, and installed packages stay safe across restarts, pauses, and machine changes.
Guardrails set by your platform team. Admins curate the approved images and hardware options, and decide which machines, reservations, and capacity types (such as on-demand or Spot) back them. Researchers only see choices that work.
Open source and portable. Workspaces is part of the Kubeflow community, empowering you with the freedom and flexibility of a fully open source notebook ecosystem.
Upstream integration and open licensing. GKE Workspaces builds directly on Kubeflow Workspaces with key enhancements for GKE environments—including standalone deployment without Istio dependencies, integrated authentication token flows, and stateful GPU/CPU pause and resume capabilities. Both GKE Workspaces and upstream Kubeflow Workspaces are released under the Apache 2.0 open source license, and we are actively working with the Kubeflow community to upstream these features.
What's unique about GKE Workspaces
GKE Workspaces brings Kubeflow Workspaces together with capabilities that are only available on Google Cloud.
1. Pause and resume—without losing your state
You've spent an hour loading a large model into GPU memory and building up results across notebook cells. Now it's time to go home. Normally you have two bad options: shut down and lose everything, or leave it running and pay for an idle GPU overnight.
With GKE Workspaces, you just pause by clicking a button from the Workspaces Dashboard. Built on Pod Snapshots, GKE takes a snapshot of your running notebook—including variables in memory and models loaded on the GPU—saves it to Cloud Storage, and releases the hardware so you stop paying. When you resume, you pick up exactly where you left off. Workspaces can also be paused automatically after a period of inactivity.
2. An easy on-ramp to Cloud TPUs
Moving from a quick experiment to large-scale training on TPUs used to mean switching tools at every step. GKE Workspaces keeps you in the same notebook the whole way, with both JAX and PyTorch:
Prototype on a single TPU chip, just like you would on a GPU.
Scale to multiple chips on one machine by choosing a bigger hardware option.
Scale out to multi-host TPU slices by sending the same training function from your notebook to a larger slice. Logs stream back to your notebook, and the TPUs are released as soon as the job finishes.
3. Use your local coding agents with cloud accelerators
Prefer using your own AI coding agents and editors locally? You can easily bridge them to cloud compute. In the Workspaces web UI, generate a short-lived, revocable link to run your notebooks and agent-generated code on a remote GPU or TPU. Traffic is secured over HTTPS via Google Identity-Aware Proxy and the cluster's access proxy, checking permissions on every request—so no SSH keys, local tunnels, or cluster credentials ever sit on your laptop.
4. A launchpad for distributed work
Your workspace doesn't have to be big to drive big jobs. From a small, inexpensive CPU notebook you can process data with Spark, launch multi-host training, serve your model, and scale Ray clusters—all from Python. Built-in Cloud Storage access, with no keys to manage, makes it easy to pass data between each step.
5. A portal to safe elastic sandboxes for coding subagents
Running a fleet of coding agents directly on your local machine can drain your battery, expose sensitive files or credentials to agent-executed code, and halt progress as soon as you close your laptop. From a standard GKE Workspace, you can safely task your favorite coding agent to hand off parallel work—such as resolving failing tests across a repository—to background subagents. Crucially, while your primary workspace operates as a full-featured environment, these subagents are isolated within lightweight, in-cluster sandboxes using providers like Agent Substrate, requiring zero local setup. Each subagent executes independently in its own sandbox on shared cluster nodes, continues running even when you log off, snapshots to Cloud Storage at zero compute cost while idle, and outputs a clean patch for your review and merge.
6. Lightweight and secure by default
GKE Workspaces runs on its own on GKE Standard or Autopilot—you don't need to install the full Kubeflow platform. Users sign in with their Google accounts, traffic is encrypted with Google-managed certificates, and each team's work is isolated from the others.
Get started
GKE Workspaces lets ML teams focus on models, not infrastructure: one-click development environments, stateful pause and resume, a smooth path to Cloud TPUs, and the power of GKE behind every notebook.
Ask your platform team to deploy it using the GKE deployment guide, and learn more about the upstream project at Kubeflow Workspaces. We welcome feedback and contributions for both Apache 2.0-licensed projects.
Want to add AI-generated 3D or XR to your web or mobile app? Unsure how to get started? That’s why we created the open source cross-platform project Three-Live—written entirely by Google Antigravity, in partnership with one former software engineer turned product manager. (Three-Live is not an officially supported Google product.)
You can think of Three-Live as a simple, unofficial companion to the more capable and feature-rich XR Blocks, Android XR’s web-based cross-platform framework for XR. The goal of Three-Live is to make it easy to get started with 3D and XR that’s AI-authored in Three.js. Three-Live combines the familiar developer experience of <model-viewer> and native mobile libraries with a sandboxed compiler for on-device or server-side compilation of untrusted LLM-generated JavaScript. Large language models (LLMs) have gotten remarkably good at writing software, and that includes generating 3D code in procedural Three.js JavaScript that describes objects, materials, lighting, and animations. But runtime AI-generated code is not trusted, and requires security sandboxing for safe execution. Also, mobile devices provide native support for 3D and XR rendering, and forcing them to run a webview limits the performance.
Today, we are excited to release Three-Live (v1.0.0, Developer Preview) under the Apache 2.0 license. It’s a lightweight, cross-platform open source library and on-device compiler that bridges the gap between AI-generated Three.js code and high-performance 3D, AR, and WebXR rendering across web, Android, and iOS.
How Three-Live works
Rather than use Three.js as a browser-only rendering loop, Three-Live treats a declarative subset of the Three.js API as a universal, portable scene description language:
Sandboxed execution with strict resource quotas: On the web, untrusted scripts execute inside an isolated Web Worker (with an optional opaque-origin sandboxed <iframe> tier) where network primitives (fetch, XMLHttpRequest, WebSocket, dynamic import()) are stripped before a single line of user code runs. On Android, iOS, macOS, and the command line, scripts execute inside an embedded Rust compiler powered by the Boa ECMAScript engine. Both runtimes enforce hard execution timeouts and geometry quotas (capping vertices, triangles, meshes, and procedural textures) so runaway scripts cannot freeze the host application.
Fast on-device compilation to GLB and USDZ: As the script constructs geometries, physically based rendering (PBR) materials, procedural canvas textures, hierarchy groups, and keyframe animations, Three-Live serializes the resulting scene graph directly into standard binary glTF 2.0 (.glb) and uncompressed Apple USDZ (.usdz) byte buffers in memory—typically in two to five milliseconds.
Platform-optimized rendering: Because the output is a standard binary 3D asset coupled with declarative animation bindings, each platform renders it using its built-in graphics pipeline:
Web (three-live and xr-live): A custom zero-dependency WebGL2 PBR renderer and <three-live> web component—inspired by <model-viewer>—complete with procedural Studio HDRI reflections, screen-space ambient occlusion (SSAO), bloom, shadow mapping, and a single-click AR handoff to Android Scene Viewer, Apple AR Quick Look, or immersive WebXR. At launch, the standard bundle with sandboxed scripting and AR is about 45 KB gzipped, and the xr-live spatial computing suite is about 55 KB.
Android (com.google.threelive:compiler): A Kotlin SDK and multi-architecture JNI library (arm64-v8a, armeabi-v7a, x86_64) that compiles Three.js strings on a background coroutine straight into ByteBuffer assets for Google FilamentModelViewer and Scene Viewer AR.
iOS and macOS (ThreeLiveCompiler): A Swift Package and five-slice XCFramework that compiles Three.js scripts on-device for RealityKit ModelEntity and ARQuickLookPreviewController.
Declarative 3D and spatial WebXR in HTML
On the web, adding an interactive, AI-generated 3D model with built-in orbit controls and mobile AR support requires just a single HTML custom element:
<script src="https://cdn.jsdelivr.net/npm/three-live@1.0.0/dist/three-live.min.js"></script>
<three-live allow-client-scripts camera-controls ar shadow-intensity="1" environment-image="studio">
<script type="text/three-live">
const knot = new THREE.Mesh(
new THREE.TorusKnotGeometry(0.6, 0.2, 128, 32),
new THREE.MeshStandardMaterial({
color: 0x4285f4,
metalness: 0.85,
roughness: 0.15
})
);
scene.add(knot);
// Baked into a looping glTF animation (4 seconds by default).
scene.userData.animate = (t) => {
knot.rotation.y = t * (Math.PI / 2);
};
</script>
</three-live>
The script runs in the sandbox described above. Because inline scripts are client-modifiable, they require the explicit allow-client-scripts opt-in; server-generated scripts can be loaded with script-src instead.
Only need to display existing models? The viewer-only build drops the script compiler and AR handoff and weighs about 23 KB gzipped at launch—tiny compared to the full Three.js library. The Three-Live viewer focuses on fast-loading PBR display, orbit controls, and glTF animation:
For spatial computing headsets and mobile AR browsers, the companion xr-live package adds <xr-live> and <xr-scene>, enabling six-degrees-of-freedom (6DoF) hand-tracking gestures (pinch, grab, two-hand scale), spatial DOM projection, and real-time voice- or text-to-3D object synthesis inside an active WebXR session.
Try it today
We built Three-Live to make 3D and spatial computing accessible to web, mobile, and AI developer communities without requiring complex 3D modeling pipelines. You can explore the project and try it right in your browser:
Interactive browser playgrounds: Launch the live Three-Live Studio and XR-Live Playgrounds to edit procedural Three.js code and inspect real-time GLB compilation in your browser.
Source code and prebuilt binaries: Visit the google/three-live GitHub repository and the v1.0.0 GitHub Release for prebuilt Linux, macOS, and Windows CLI binaries (three-live-cli), Android .aar archives, and Apple .xcframework bundles.
Package managers: Install three-live and xr-live from npm, three-live-compiler from Crates.io, or add https://github.com/google/three-live.git in Xcode via Swift Package Manager.
Three-Live is currently in Developer Preview, and we welcome feedback, bug reports, and pull requests from open source developer communities on GitHub.
