How Google Uses engineering

HeyGen ported their 18B+ parameter Avatar IV video generation model to Google Cloud's Trillium (v6e) TPUs via torchax and XLA, utilizing FSDP and Ulysses sequence parallelism across an eight-chip mesh. To achieve a 1.86x speedup for real-time streaming, the engineering team pipelined exposed all-to-all collectives, aligned sparse attention block sizes to eliminate mask padding, and bypassed softmax serial dependencies using a precomputed Cauchy-Schwarz upper bound. These custom Pallas kernel and compiler optimizations were deployed only after passing rigorous two-tier quality gates to guarantee byte-identical or mathematically equivalent pixel outputs.

As AI coding assistants shift the developer's primary role from writing boilerplate to reviewing and maintaining systems, language choice becomes critical for long-term architectural integrity. Go directly addresses this new paradigm by utilizing its strict compiler, integrated toolchain, and uncompromising readability to provide deterministic guardrails that help AI models self-correct and generate highly standardized code. By enforcing ecosystem-wide consistency and strict backward compatibility, the Go platform empowers engineering teams to efficiently verify, optimize, and maintain high-velocity, AI-generated output in production environments.

To prevent context window bloat and reduce token consumption, Genkit Go introduces Agent Skills based on a progressive disclosure architecture. Developers can package specialized instructions, scripts, and references into modular SKILL.md bundles where only the frontmatter metadata is initially exposed to the agent's system prompt. When a task matches the skill's description, Genkit's middleware dynamically loads the full instruction body and associated assets, ensuring the model accesses precise workflows exactly when needed.

Google's open-source TPU microbenchmark suite provides developers with granular performance metrics across Network, Compute, HBM, Host Transfer, and Attention components to validate real-world hardware capabilities. By leveraging these benchmarks to establish a Roofline model, engineers can accurately diagnose whether their machine learning workloads are compute-, memory-, or network-bound. This empirical baseline directly guides targeted software optimizations—such as kernel tuning, mesh sharding, and rematerialization—to maximize hardware utilization for large-scale model deployments.

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.

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.

Building AI agents often leaves developers uncertain if prompt tweaks to fix single errors will accidentally cause widespread regressions in production. To bridge this gap, Google has introduced a new developer skill for coding agents that automates a five-stage evaluation flywheel: preparing data, running inference, grading with adaptive AutoRaters, analyzing failure clusters, and executing targeted optimizations. Running continuously against production traffic or on-demand via synthetic scenarios, this tool allows developers to describe testing goals in plain language while an independent evaluation service safely validates and counts actual performance improvements.

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

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

An open specification for finding and verifying tools, skills, and agents across the web.Agents are ...

This post introduces three architectural patterns designed to integrate Model Context Protocol (MCP) Apps and Agent-to-User Interface (A2UI) to solve the tradeoff between highly custom iframe environments and native, declarative rendering. By combining these approaches, developers can serve native-feeling UIs directly over MCP servers, embed complex and stateful iframe apps securely inside declarative views, or inject generative UI components into legacy systems. Ultimately, these hybrid frameworks empower engineering teams to deliver secure, performant, and brand-consistent agentic user experiences tailored to their specific project constraints.

The Google AI Edge Gallery app has expanded its on-device AI capabilities by introducing experimental support for the open-source Model Context Protocol (MCP) on Android, allowing Gemma 4 to coordinate complex tasks across external data sources like Google Workspace and Google Maps. To enable more proactive and persistent user interactions, the update adds a "Schedule Notification" skill for automating routines and a persistent chat history feature that restores long session contexts nearly instantly. Driven by an open-source toolkit, the platform encourages community developers to build and share custom utility-focused workflows, prompt configurations, and tool integrations via its GitHub repository.

The Google Cloud and NVIDIA developer community is celebrating its first anniversary with 100,000 members and a renewed focus on providing builders with advanced AI infrastructure and resources. To accelerate development, the community offers curated learning pathways for mastering LLM optimization, GPU-accelerated data analytics, and monthly expert-led webinars. Moving into its second year, the initiative will expand to include hands-on labs, engineering events, and specialized content focused on the growth of agentic AI.

Researchers at UCSD have successfully implemented DFlash, a block-diffusion speculative decoding method, on Google TPUs to bypass the sequential bottlenecks of traditional autoregressive drafting. By "painting" entire blocks of candidate tokens in a single forward pass rather than predicting them one-by-one, the system achieved average speedups of 3.13x, with peak performance nearly doubling that of existing methods like EAGLE-3. This open-source integration into the vLLM ecosystem optimizes TPU hardware by leveraging "free" parallel verification and high-quality draft predictions for complex reasoning tasks.

The blog post outlines the transition of a brittle sales research prototype into a robust production agent using Google’s Agent Development Kit (ADK). By replacing monolithic scripts with orchestrated sub-agents and structured Pydantic outputs, the developers eliminated silent failures and fragile parsing. Additionally, the post highlights the necessity of dynamic RAG pipelines and OpenTelemetry observability to ensure AI agents are scalable, cost-effective, and transparent in real-world applications.

The Google Cloud AI Agent Bake-Off highlights a shift from simple prompt engineering to rigorous agentic engineering, emphasizing that production-ready AI requires a modular, multi-agent architecture. The post outlines five key developer tips, including decomposing complex tasks into specialized sub-agents and using deterministic code for execution to prevent probabilistic errors. Furthermore, it advises developers to prioritize multimodality and open-source protocols like MCP to ensure agents are scalable, integrated, and future-proof against rapidly evolving model capabilities.

TorchTPU is a new engineering stack designed to provide a native, high-performance experience for running PyTorch workloads on Google’s TPU infrastructure with minimal code changes. It features an "Eager First" approach with multiple execution modes and utilizes the XLA compiler to optimize distributed training across massive clusters. Moving into 2026, the project aims to further reduce compilation overhead and expand support for dynamic shapes and custom kernels to ensure seamless scalability for the next generation of AI.

The launch of Agent Development Kit (ADK) for Go 1.0 marks a significant shift from experimental AI scripts to production-ready services by prioritizing observability, security, and extensibility. Key updates include native OpenTelemetry integration for deep tracing, a new plugin system for self-healing logic, and "Human-in-the-Loop" confirmations to ensure safety during sensitive operations. Additionally, the release introduces YAML-based configurations for rapid iteration and refined Agent2Agent (A2A) protocols to support seamless communication across different programming languages. This framework empowers developers to build complex, reliable multi-agent systems using the high-performance engineering standards of Golang.

Google has released version 1.0.0 of the Agent Development Kit (ADK) for Java, introducing powerful new features like Google Maps grounding, built-in URL fetching, and a standardized Agent2Agent protocol for cross-framework collaboration. The update enhances agent control through a new "App" and "Plugin" architecture, which allows for global logging, automated context window management via event compaction, and "Human-in-the-Loop" workflows for action confirmations. Additionally, the release provides robust session and memory services using Google Cloud integrations like Firestore and Vertex AI to manage long-term state and large data artifacts.

To bridge the gap between static model knowledge and rapidly evolving software practices, Google DeepMind developed a "Gemini API developer skill" that provides agents with live documentation and SDK guidance. Evaluation results show a massive performance boost, with the gemini-3.1-pro-preview model jumping from a 28.2% to a 96.6% success rate when equipped with the skill. This lightweight approach demonstrates how giving models strong reasoning capabilities and access to a "source of truth" can effectively eliminate outdated coding patterns.

Google has introduced Finish Changes and Outlines for Gemini Code Assist in IntelliJ and VS Code to reduce developer friction and eliminate the need for long, manual prompting. Finish Changes acts as an AI pair programmer that completes code, implements pseudocode, and applies refactoring patterns by observing your current edits and context. Meanwhile, Outlines improves code comprehension by generating interactive, high-level English summaries interleaved directly within the source code to help engineers navigate and understand complex files.

While keynotes are available online, Google Cloud Next '26 in Las Vegas offers an irreplaceable in-person experience centered on networking, hands-on problem solving, and the transition to agentic AI. The event features specialized technical tracks covering everything from Gemini multimodal breakthroughs to zero-trust security on Cloud Run, providing developers with the tools to balance individual speed with organizational stability. Beyond formal sessions, the "in-person advantage" lies in over 20 developer meetups and collaborative whiteboard sessions designed to foster serendipitous breakthroughs. Ultimately, the conference serves as a high-energy hub for engineers to move beyond the hype and master the modern building blocks of software architecture together.

Google has introduced FunctionGemma, a specialized 270M parameter model designed to bring efficient, action-oriented AI experiences directly to mobile devices through on-device function calling. By leveraging Google AI Edge and LiteRT-LM, the model enables complex tasks—such as managing calendars, controlling device hardware, or executing specific game logic in the "Tiny Garden" demo—to be performed entirely offline with high speed and low latency. Available for testing in the Google AI Edge Gallery app on both Android and iOS, FunctionGemma allows developers to move beyond simple text generation toward building responsive, "agentic" applications that interact seamlessly with the physical and digital world without relying on cloud processing.