As I mentioned in my Friday video, there were some rumblings kicking into gear starting late Thursday, July 23rd, and through Friday, July 24th. It seems like Google kicked some of its search ranking servers, and as a result, we are seeing an unconfirmed Google search ranking update, with a lot of volatility throughout last week, but it kicked into high gear on Friday.

Distributed AI training is notoriously fragile because losing a single machine typically crashes the entire multi-node job, forcing a time-consuming, full-workload infrastructure restart. To address this, Google’s JAX ecosystem utilizes elastic training via Pathways, which converts a hardware failure into a catchable Python exception so the running process can survive. When an unplanned failure occurs, the system automatically replaces only the broken worker, restores the last viable checkpoint from Cloud Storage, and resumes training in place—minimizing total downtime to under two minutes without ever restarting the main controller process.

Tunix is Google’s new JAX-native post-training library designed to eliminate TPU idling bottlenecks when training multi-turn, tool-using LLM reasoning agents. It maximizes hardware throughput by combining highly concurrent, asynchronous rollouts with a decoupled producer-consumer pipeline, ensuring the trainer is constantly fed even while agents wait on network I/O or environment steps. Additionally, Tunix provides plug-and-play abstractions and continuous macro-level profiling, allowing developers to easily integrate custom open-source environments and optimize complex distributed workflows without massive code rewrites.

On May 23, 2026, fresh off the stage at Google I/O, our Google Developer Experts (GDEs) converged on...

The open-source Genkit framework has introduced the Agents API, a full-stack tool designed to simplify the complex plumbing of conversational AI by packaging message history, tool loops, and streaming into a single interface. The API supports flexible, server- or client-managed state persistence—allowing for advanced workflows like history branching, long-running detached tasks, and multi-agent coordination—while seamlessly connecting backends to frontends via a unified wire protocol. Currently available in preview for TypeScript and Go, it also integrates with the Genkit Developer UI to allow developers to easily test, debug, and inspect agent snapshots without writing client code.

Answering the questions of "why we built ADK 2.0". This explains the rationale, some of the features, and why a developer should consider upgrading. This will be published the day after ADK go 2.0 launches.

The Google Cloud Workbench Notebooks extension for VS Code has officially launched, allowing developers to connect their local IDE to scalable, cloud-based Jupyter environments. This integration streamlines the machine learning lifecycle by eliminating context switching and providing direct access to high-performance Google Cloud infrastructure. To support transparency and community-driven innovation, the newly released extension is fully open-sourced and available on GitHub and the VS Code Marketplace.

Celebrating the first anniversary of the Agent-to-Agent (A2A) protocol, this blog post highlights how the framework enables autonomous AI agents to securely collaborate and hand off tasks without the rigidity of traditional APIs. By delegating complex workflows to specialized peer agents, A2A prevents context pollution, ensures data privacy, and simplifies application design through modularity. To demonstrate this ecosystem in action, the post spotlights FoldRun—an agentic interface for life sciences that orchestrates complex protein structure predictions—alongside diverse A2A use cases spanning commerce, data streaming, DevOps, and telecommunications.

How a Python agent and a Go agent collaborate on contract compliance using the Agent2Agent protocolY...

To serve the 397B-parameter Qwen 3.5 Mixture-of-Experts (MoE) model on Ironwood TPUs, engineers developed a modular JAX/Pallas optimization stack that achieved up to a 4.7x inference speedup for prefill-heavy workloads. The team bypassed severe hardware sharding constraints by deploying a hybrid Data Parallelism and Expert Parallelism (DP+EP) topology, paired with custom low-level communication fusions like a hierarchical reduce-scatter to optimize cross-device token routing. Finally, by executing hardware-aware custom kernels—such as Batched Ragged Page Attention and a fully-fused Gated DeltaNet (GDN) block—they successfully saturated HBM bandwidth and TensorCore MXUs to push system throughput near its theoretical roofline limits.

Conductor has evolved from a Gemini CLI extension into a portable plugin, bringing conversational Spec-Driven Development (SDD) to ecosystems like Antigravity CLI and Claude. Rather than relying on strict command sequences, developers can now chat naturally with their AI assistant while it dynamically manages persistent markdown artifacts (like spec.md and plan.md) in the background. This update eliminates workflow friction while ensuring your repository remains a version-controlled, single source of truth for your project's architecture and state across different AI tools.

AI coding agents are rapidly shifting from reactive assistants that complete tasks when prompted to ...

To resolve the scaling bottlenecks and runtime errors caused by monolithic system prompts, engineering teams should treat prompts as build artifacts by modularizing instructions into reusable templates. By running these modular "skill files" through a transpiler, developers can enforce static validation, catch missing dependencies at build time, and integrate prompt generation directly into their CI/CD pipelines. This deterministic approach prevents code drift and ultimately establishes a safe framework where agents can propose updates to their own logic via standard pull requests.

Ray 2.55 introduces official, first-class support for Google Cloud TPUs, enabling developers to run distributed Python workloads on Google's accelerators using the familiar Ray task-and-actor APIs. To handle the strict networking requirement of keeping multi-host TPU "slices" together over their Inter-Chip Interconnect (ICI), the KubeRay Operator on GKE automatically provisions and labels the underlying hardware layout. Ray Core utilizes these labels via its slice_placement_group() primitive to atomically reserve complete slices, allowing developers to deploy jobs through KubeRay, Ray Train, or Ray Serve simply by declaring a hardware topology (like "4x4") without writing custom placement code.

We're excited to introduce LiteRT.js, the newest member of the LiteRT family! LiteRT.js is our powerful solution for running machine learning models directly in the browser, extending Google's cross-platform edge AI runtime to the web. Built for JavaScript developers, LiteRT.js delivers state-of-the-art ML model inference performance on WebGPU and upcoming WebNN, with a fallback to WebAssembly for CPU. This post provides a quick tour of LiteRT.js and gives web developers everything they need to get started.

Google Cloud has partnered with Parallel Web Systems to natively integrate Parallel's search infrastructure as a web grounding provider on the Gemini Enterprise Agent Platform. This integration enables developers to anchor their AI agents in verifiable, real-time web results, significantly improving factual accuracy for complex enterprise workflows. Additionally, the partnership offers expanded architectural flexibility, allowing users to programmatically extract, permanently cache, and process web data alongside other large language models.

Earlier this month, we announced our new model, Muse Spark 1.1, which powers the Meta AI app and meta.ai. The model is built to plan, work with your apps, and follow through from start to finish, and those capabilities are what make the new features possible in Meta AI. With it driving the experience, Meta AI is starting to take action on your behalf. 

This is our next step toward personal superintelligence: an AI that knows your context, is there for you whenever you need it, and handles things so you don’t have to.

An AI Assistant That Plans and Follows Through

You’ve been using Meta AI to find answers, think through ideas, and create content. Now it can go further and actually get things done for you. Meta AI can now make plans, follow through on next steps, and keep you on track without needing to be reminded or re-prompted. 

Tell Meta AI you’re renovating your kitchen and it will learn your style, scout Marketplace for furniture and fixtures that fit your budget, and send you a mood board so you can see how it all comes together. Ask it to help you build a running plan for your first half marathon and it will create a week-by-week training schedule, adjust for your availability, and share your plan for the week every Monday morning. Or ask it to plan a birthday dinner and it will find restaurants, check your calendar for a night that works, and suggest options.

Download video Download image

Meta AI can now give you a daily briefing, pulling what’s on your calendar, finding relevant updates like realizing when you’re double booked or when plans have changed, and giving you a quick summary at your preferred time. You only have to set up tasks with Meta AI once, then you can leave Meta AI to handle them, whether that’s a weekly meal plan, a heads-up on the latest sneaker drops, or an afternoon update on trends you care about. Tell Meta AI what you want and when, and it’ll keep delivering.

 

When you want to dig into a topic, Meta AI can do the research in minutes, synthesizing information from across the web, from research papers to what creators and communities share on our apps, so you can understand a subject or weigh your options without the manual work. And if you want to turn what you’ve learned into something visual, ask Meta AI to generate slides. 

You don’t have to wait for Meta AI to finish before you weigh in. While it’s putting together a report, a presentation, or a plan for you, you can steer it in real time. Tell it to shift focus, change the tone, or cut a section, and it’ll course-correct on the spot so the end result is actually what you wanted. And everything Meta AI creates for you, from training schedules to slide decks to mood boards, now lives in one place, so you can revisit it, build on it, and share it with friends.

https://about.fb.com/wp-content/uploads/2026/07/03_Research.mp4 How to Get Started

These features will start rolling out today in select markets in the Meta AI app and meta.ai. We’ll bring them to more countries and surfaces, including WhatsApp, in the coming weeks. 

It’s your choice how you use Meta AI in the app and meta.ai: find quick answers; get thoughtful responses that reflect your world; or shop with real inspiration from posts, Reels, and products that people are sharing across our apps. And when you want a fully private conversation, Incognito chats are always there. 

This is just the start of what we’re envisioning for Meta AI, and we’ll keep building to make AI more fun, relevant, and useful for you. 

The post Meta AI Doesn’t Just Think, It Acts appeared first on Meta Newsroom.