AI Invades Whiteboards, Canvases, and Phones: 5 Key Nov 2026 Updates

01 Claude Adds Whiteboards and Sheets, Data Goes GA

Anthropic on November 10 released Claude Whiteboards and Claude Sheets in beta, and turned the previous Claude Dashboards into the generally available Claude Data. All three are available to every paid plan. Whiteboards let teams collaborate with Claude on a shared canvas — sticky notes, timelines, assignees, dates — and then convert the output into documents or presentations. Sheets generates spreadsheets with worksheets, formulas, formatting, and dropdowns, highlighting cell changes; files export to .xlsx or CSV. Claude Data builds visualizations and analyses from connected data sources, adding version tracking and cache settings; authenticated queries and live sharing remain in beta. Enterprise admins control feature rollout: Whiteboards and Sheets default off, Data defaults on November 19. The update moves the user experience from receiving a single suggestion to co-editing persistent, shareable artifacts.

Editor's diagram showing the three workspace types. Whiteboards and Sheets are beta; parts of Data are still in testing.

02 Adobe Expands Firefly: Generation and Editing on One Canvas

At its MAX conference on November 10, Adobe broadened Firefly by placing Creative Agent, multiple partner models, and image, video, and design tools into a single work environment. Firefly Design beta unifies illustration, image editing, animation, 3D, and print design on one canvas. Image tools gain text editing, relighting, and retouching controls; video tools add clip arrangement, captions, transitions, and insert shots. Firefly Graph lets users compose reusable pipelines that chain models and editing tools. Firefly Builder generates interactive prototypes and is available via private preview and waitlist. Adobe states that Edit Text, Retouch, Relight, and Quick Actions are generally available; other new capabilities remain in beta. The accompanying image shows brand visuals across posters, apparel, and other media — not the software UI — illustrating the intended expressive range.

Adobe's official release image showing creative outcomes across media, not the software interface.

03 Pixel Feature Drop: Voice Input, Multi-Step Tasks, Custom Widgets

Google began a new Pixel Drop on November 10, rolling out over the following weeks. Pixel 10 series devices start receiving Gemini Intelligence capabilities; feature availability depends on model, language, region, and settings. Rambler organizes spoken input — removing filler words and structuring the utterance into clean text — with a live preview before sending. Multi-step task handling expands to Japan, the UK, and India, chaining actions for shopping, ride-hailing, food delivery, and restaurant reservations; this regional rollout is separate from the global push. Users can now create personalized home-screen widgets by describing them, previewing the result, then adding it; icon style customization is also extended. Pixel Search adds cross-device content lookup for Pixel 6 and newer. Actual usability depends on the specific update each device receives and Google's stated conditions.

Google's official UI demo: left side shows voice input, right side shows the organized text ready to send.

Official demo showing the continuous steps from text description to widget preview to home-screen placement.

Different icon styles previewed in settings; device shown is Pixel 11 Pro.

04 Google Launches MKA Benchmark for Visual Knowledge Work

Google partnered with Artificial Analysis to introduce the Multimodal Knowledge Agent (MKA) benchmark, testing how agents handle knowledge tasks that include images, layouts, and mixed file types. The suite contains 200 tasks drawn from 10 domains (architecture, engineering, finance, healthcare, etc.) and 83 occupations. Inputs may be architectural drawings, engineering schematics, maps, or data files; outputs must be verifiable artifacts. Each task has expert-written scoring rubrics weighted 75% factual accuracy, 25% visual design. All models run on the same open-source agent framework, tools, sandbox, and turn budget. The full 200-task set stays private with Artificial Analysis; 10 sample tasks are public for researchers. The benchmark shifts evaluation from pure text fluency to the ability to open, inspect, and produce files with diagrams, tables, and formatting.

Official overview of the ten domains covered, including architecture, engineering, finance, and healthcare.

Official samples for architecture, geospatial, and engineering tasks, showing input files, target outputs, and scoring checks.

05 AMD 6th-Gen EPYC Ships, Highlighting CPU Roles in Agent Workflows

AMD announced on November 10 that its 6th-generation EPYC SP7 server processors, based on the Zen 6 architecture, are now generally available; servers built on the new platform have started shipping from major vendors. Microsoft Azure and Oracle Cloud Infrastructure (OCI) instances are listed as "coming soon." AMD categorizes agent-related CPU workloads into three roles: general services (running applications and databases), GPU host nodes (supporting accelerated compute), and agent sandboxes (executing code and tools). A single multi-model workflow may simultaneously perform retrieval, file processing, and program execution, requiring the server to handle diverse compute demands concurrently. AMD's positioning reflects how such mixed workloads are influencing server product segmentation. Performance and efficiency claims use specific configurations and include internal estimates; real-world deployment results will depend on workload and system configuration.

Editor's diagram based on AMD's announcement; the three roles do not indicate performance proportions.

Observation: From Answering Questions to Continuous Artifact Editing

Today's software updates share a clear thread: whiteboards and sheets capture discussion, creative canvases capture assets, phone interfaces capture daily actions. Benchmarks and server platforms behind them aim to measure and support longer execution chains. For practitioners, the key lenses are "is it open?" and "can it finish real tasks?" — general availability, public beta, private preview, and phased rollouts each imply different readiness levels. This brief is based on official releases accessible as of the morning of November 11, 2026; primary announcements occurred November 10. Two editor-created diagrams illustrate functional relationships; other images are from official announcements and UI demos. Vendor-disclosed capabilities and availability are reported as stated.