This blog post introduces a workflow for extracting high-quality data from complex, unstructured documents by combining LlamaParse with Gemini 3.1 models. It demonstrates an event-driven architecture that uses Gemini 3.1 Pro for agentic parsing of dense financial tables and Gemini 3.1 Flash for cost-effective summarization. By following the provided tutorial, developers can build a personal finance assistant capable of transforming messy brokerage statements into structured, human-readable insights.
Hi, everyone! We've just released Chrome 146 (146.0.76380.164) for Android. It'll become available on Google Play over the next few days.
This release includes stability and performance improvements. You can see a full list of the changes in the Git log. If you find a new issue, please let us know by filing a bug.
Android releases contain the same security fixes as their corresponding Desktop releases (Windows & Mac: 146.0.7680.164/.165, Linux: 146.0.7680.164) unless otherwise noted.
The Stable channel has been updated to 146.0.7680.164/165 for Windows/Mac and 146.0.7680.164 for Linux, which will roll out over the coming days/weeks. A full list of changes in this build is available in the Log
Interested in switching release channels? Find out how here. If you find a new issue, please let us know by filing a bug. The community help forum is also a great place to reach out for help or learn about common issues.
Advanced TPU optimization with XProf: Continuous profiling, utilization insights, and LLO bundles
In our previous post, we introduced the updated XProf and the Cloud Diagnostics XProf library, which are designed to help developers identify model bottlenecks and optimize memory usage. As machine learning workloads on TPUs continue to grow in complexity—spanning both massive training runs and large-scale inference—developers require even deeper visibility into how their code interacts with the underlying hardware.
Today, we are exploring three advanced capabilities designed to provide "flight recorder" visibility and instruction-level insights: Continuous Profiling Snapshots, the Utilization Viewer, and LLO Bundle Visualization.
Continuous Profiling Snapshots: The "Flight Recorder" for ML
Standard profiling often relies on "sampling mode," where users manually trigger high-fidelity traces for short, predefined durations. While effective for general optimization, this traditional approach can miss transient anomalies, intermittent stragglers, or unexpected performance regressions that occur during long-running training jobs.
To address this visibility gap, XProf is introducing Continuous Profiling Snapshots. This feature functions as an "always-on" flight recorder for your TPU workloads. How it works: Continuous profiling snapshots (Google Colab) operates quietly in the background with minimal system overhead (approximately 7µs per packet CPU overhead). It utilizes a host-side circular buffer of roughly 2GB to seamlessly retain the last ~90 seconds of performance data. This architecture allows developers to snapshot performance data programmatically precisely when an anomaly occurs, bypassing the overhead and unpredictability of traditional one-shot profiling.
Figure 1: Traditional trace capturing without Continuous Profiling Snapshots. Figure 2: Comprehensive context captured via Continuous Profiling Snapshots.
Key technical features include:
Circular Buffer Management: Continuously holds recent trace data to ensure you can capture the exact moments leading up to an anomaly or regression.
Out-of-band State Tracking: A lightweight service polls hardware registers for P-state (voltage and frequency) and trace-drop counters, ensuring the snapshot contains the necessary environmental context for accurate analysis.
Context Reconstruction: The system safely decouples state capture from the trace stream. This ensures that any arbitrary snapshot retains the ground truth required for precise, actionable debugging.
Visualizing Hardware Efficiency with the Utilization Viewer
Raw performance counters are powerful, but interpreting thousands of raw hardware metrics can be a daunting, time-consuming process. The new Utilization Viewer bridges the gap between raw data streams and actionable optimization strategies.
This tool translates raw performance counter values into easily understandable utilization percentages for specific hardware components, such as the TensorCore (TC), SparseCore (SC), and High Bandwidth Memory (HBM).
Figure: Raw Performance Counter
Figure 3: Deriving actionable insights from raw performance counters.
From Counters to Insights: Instead of requiring developers to manually analyze a raw list of event counts, the Utilization Viewer automatically derives high-level metrics. For example, it can translate raw bus activity into a clear utilization percentage (e.g., displaying an average MXU bus utilization of 7.3%). This immediate clarity allows you to determine at a glance whether your model is compute-bound or memory-bound.
Figure 4: Perf Counters Visualization in Utilization Viewer
Inspecting the Metal: Low-Level Operations (LLO) Bundles
For advanced users and kernel developers utilizing Pallas, we are now exposing Low-Level Operations (LLO) bundle data. LLO bundles represent the specific machine instructions issued to the TPU's functional units during every clock cycle.
This feature is critical for "Instruction Scheduling" verification—ensuring that the compiler is honoring your programming intentions and correctly re-ordering instructions to maximize hardware performance.
New Visualizations via Trace View Integration: You can now visualize LLO bundles directly within the trace viewer. Through dynamic instrumentation, XProf inserts traces exactly when a bundle executes. This provides exact execution times and block utilization metrics, rather than relying on static compiler estimates.
Why it matters: Accessing this level of granularity enables hyper-specific bottleneck analysis. For instance, developers can now identify idle cycles within the Matrix Multiplication Unit (MXU) pipeline, making it easier to spot and resolve latency between vmatmul and vpop instructions.
Conclusion
Whether you are trying to capture a fleeting performance regression with Continuous Profiling, verifying kernel efficiency with LLO Bundles, or assessing overall hardware saturation with the Utilization Viewer, these new features bring internal-grade Google tooling directly to the open-source community. These tools are engineered to provide the absolute transparency required to optimize high-scale ML workloads.
Get started by checking out the updated resources:
Our team will be on this server regularly to interact with the Display & Video 360 community, answering questions and responding to feedback. We’ll also let you know what’s new with our products in the #announce-display-video-360 channel.
Second, we have made a significant update to the Display & Video 360 API and Structured Data Files documentation. The updated site is designed to make it easier to learn how to perform common tasks with Task guides, browse important conceptual information with Concept guides, and review reference material with updated navigation that highlights currently-supported versions and features of the API and Structured Data Files.
If you have feedback on these changes or the content of any of the guides, please click the Send Feedback button at the bottom of any page in the documentation.
Easily find and set time zones in Google Calendar by searching for city or country
We’re introducing an improved time zone picker for Google Calendar on the web. Instead of manually scrolling through the list of options, you can now simply search for and select a specific city or country, making it easier to coordinate and schedule with others in different time zones. | Learn more about how to easily find and set time zones in Google Calendar by searching for city or country.
Workspace audit logs: New functionality and expanded event fields
Introducing the new Meetings section in Google Chat
We’re introducing a dedicated Meetings section in Google Chat—a new way to organize your conversation list and keep your meeting conversations in one place. | Learn more about the new Meetings section in Google Chat.
New ways to customize and interact with your content in NotebookLM
We’re introducing several updates to NotebookLM that give users more flexibility when uploading sources, new ways to create and edit visuals from NotebookLM and improvements to the interactive study tools. | Learn more about new ways to customize and interact with your content in NotebookLM.
The announcements above were published on the Workspace Updates blog over the last week. Please refer to the original blog posts for complete details.
We’re introducing several updates to NotebookLM that give users more flexibility when uploading sources, new ways to create and edit visuals from NotebookLM and improvements to the interactive study tools.
More functions and new visual communication tools
Slide revisions: Users can now revise presentation content, ask for changes or fix any issues on both desktop and mobile. Users can submit stylistic or factual feedback for each slide and the slides are regenerated quickly as a new deck in the Studio panel.
Cinematic Video Overviews: Cinematic Video Overviews are immersive, deep-dive videos with fluid animations and rich, detailed visuals to help you learn and engage with the topics you care about. Gemini models make hundreds of structural and stylistic decisions to best tell the story with your sources. The videos are ideal for visualizing complex narratives and academic research.
New infographic styles: Users can now manually choose from ten new predefined infographic styles to visually summarize their sources. Options include: Sketch Note, Kawaii, Professional, Scientific, Anime, Clay, Editorial, Instructional, Bento Grid, and Bricks. By default, NotebookLM will continue to auto-select the best style for your sources.
Improved flashcards & quizzes: Progress is now saved and loaded across sessions when using Flashcards and Quizzes in NotebookLM. Users can also now mark flashcards as "Got it" or "Missed it," shuffle the deck, and use a new results screen to rerun any cards they missed. Additionally, specific flashcards or quiz questions can now be deleted. These improvements are available on both web and mobile.
Improvements to flashcards and quizzes are available to all users. Slide revisions, Cinematic Video Overviews, and new infographic styles are available to users over the age of 18.
Support for more file types
EPUB: All NotebookLM users can now upload EPUB files, the standard, open-source file format for eBooks, as a supported source type in NotebookLM, making it easier to research and interact with long-form digital books.
PPTX export: Users can now export generated slide decks as PPTX (in addition to PDF).
Chat improvements
Saved and secure conversation history: To support long-term projects, your conversations will be automatically saved and kept private to you. You can close a session and resume it later without losing your conversation history. You can delete chat history at any time, and in shared notebooks, your chat is visible only to you.
Artifact creation in chat: You can now instantly transform your chat conversations into stunning Audio Overviews, Video Overviews, tailored reports, and more. Just ask Notebook to create the artifact of your choosing — all without leaving the chat.
Slide revisions, new infographic styles, improved flashcards & quizzes, EPUB as a source, PPTX export, saved and secure conversation history, artifact creation in chat
Available to all Google Workspace customers and users with personal Google accounts with access to NotebookLM
Cinematic Video Overviews (English only)
Business: Business Standard and Plus
Enterprise: Enterprise Standard and Plus
Consumer: Google AI Pro and Ultra
AI Add-ons: AI Ultra Access, AI Expanded Access, AI Pro for Education
Visit the Help Center to learn more about NotebookLM features and access by Workspace edition.
- Trevor Mulchay, Display & Video 360 API Team
Learn more about the evolving threat landscape based on Mandiant's new M-Trends 2026 cybersecurity report.
An overview of how Gemini models bring unmatched value to Google Marketing Platform.
For World Water Day we’re releasing our 2026 Water Stewardship Project Portfolio summary report and sharing details on new partnerships.
