SiliconANGLE AI reports that Google’s GKE Labs team donated OpenRL to the Cloud Native Computing Foundation. OpenRL debuted as a research preview in June 2026 with a self-hosted training API for a company’s own Kubernetes cluster. Google Cloud senior software engineer Sunil Arora said it separates the reinforcement-learning loop from compute infrastructure, allowing infrastructure engineers to handle scaling while researchers retain control of experiments. The GKE Labs team also built an AI agent for Kubernetes tasks and a benchmark for evaluating models on that work. Arora described the workload as combining training, sampling, sandboxed agent actions and evaluation. Google Cloud sponsored the interview segment but did not have editorial control, according to the disclosure.
Meta's Muse is among a new wave of personal AI agents designed to keep working after users close the app, monitoring prices, bills, and tasks from dedicated cloud VMs. TMTPost reports that Muse reached the top of the US App Store free chart after 10 days and that Meta disclosed more than 10 million sign-ups after six weeks, while also highlighting compute constraints, platform blocks, security risks, and related products from Google, OpenAI, Anthropic, and Chinese companies.
In a Google Developers Blog case study, developer Karl Weinmeister describes a workflow that compiles Go functions, identifies bounds-check panic traps left after optimization, and uses Gemini 4 Argon to translate their paths into Z3 constraints. Solver results become concrete Go reproducers that are verified with go test. The author reports that the workflow caught all three bugs he planted in test parsers and found three latent bugs in skill-lint; go test -fuzz missed two of those crashes. He notes that an LLM-written solver model is not a sound proof system, although each reported crash was reproduced directly by go test.
Google AI Studio announced that Managed Agents users on the Gemini API can now bring their own sandbox through self-hosted environments. The brief announcement confirms the capability but provides few details about setup, compatibility, or availability.
Google Home speakers running Gemini have brought back many classic Google Assistant sounds for special timers, according to 9to5Google. The restoration appears on some speaker generations, with a 2017 Google Home producing generic sounds while a 2026 Google Home Speaker offered a broader selection. Smart-home commands now use a new confirmation cue that the author says resembles an error sound rather than the former success “ding.”
Henry Cavill’s Mustache Protocol is now playable on the web through Googlebook, offering a three-level, 16-bit retro game on desktop and mobile. Android Authority reports that Cavill used Googlebook tools including Continue On, Rambler, BandLab, and Google Antigravity to polish the project and create its theme song. Google also published the third and final episode of its behind-the-scenes development series.
Google launched the Gemini Decisions API, powered by Gemini Decision 1, to evaluate up to 1,000 yes/no, multiple-choice, and rubric-scoring questions across text, images, audio, video, and PDFs in one pass, according to Google AI. The API is available in Google AI Studio, while Google reported 150–250 ms end-to-end latency for inputs up to 16K tokens; an independent developer test also found that it could force out-of-range inputs into existing categories when no fallback option was provided.
Android Authority reports that some Chrome users could not close the Google Lens side panel because its “x” button was missing. Users said closing tabs, shutting down Chrome, clearing caches and cookies, and restarting their computers did not help. One user offered a temporary workaround involving dragging the tab with Lens open into a new window and then back. Google said the issue was fixed, but some affected users reportedly encountered it after the latest update.
NVIDIA says its open-source AI Cluster Runtime (AICR) now includes a qualified AI bill of materials component that integrates Google's k8s-aibom to inventory recognized AI workloads on Kubernetes. The integration supports version-locked deployment bundles, provenance and security-default checks, and per-namespace opt-in inventory, while NVIDIA reports low controller overhead in a GKE test using k8s-aibom v1.6.1. Runtime observations are not signed by default, and linking them with verified provenance and deployment intent remains unfinished work.
Google announced a global rollout of the standalone Gemini app for Apple CarPlay, bringing Gemini Live hands-free conversations to supported iOS users. The app supports interruptions, topic changes and pausing, integrates with Gmail and Google Calendar for email summaries and schedule queries, and syncs CarPlay conversations with the Gemini iPhone app. Users need the latest Gemini app, microphone permissions and prior Gemini Live use; rollout is phased, with some personalized features subject to regional and age restrictions.
Google researcher Pietro Miotti said he used 48 AI agents to analyze NASA lidar data and identify a Maya site in Mexico that was absent from the register of Mexico’s National Institute of Anthropology and History (INAH). Archaeologists then showed that the site had already been mapped and published in Estrada-Belli et al. 2025, making the result a rediscovery rather than a new discovery. Miotti updated his account to credit the earlier researchers and acknowledged that checking a government database alone was insufficient to establish whether a site was previously known.
Google is donating gVisor, including its name and trademarks, to the Cloud Native Computing Foundation (CNCF), shifting the project to maintainer-led governance. Modal, Ant Group and Tines are joining the maintainers, while OpenAI, Tencent and NVIDIA continue contributing. Modal says gVisor powers more than 40 million containers daily and reports that its RDMA implementation reaches about 99% of native bus bandwidth in NCCL benchmarks.
Charles Humble argued at GOTO Copenhagen that sustainability should be a top-level concern in AI architecture, alongside cost, performance, security, compliance and reliability. He recommended using prebuilt or smaller models, training with renewable energy, and placing or shifting workloads when lower-carbon power is available. InfoQ also cited IEA estimates and company reports showing rising electricity use and emissions, including Microsoft’s and Google’s attribution of much of their growth to AI infrastructure.
Google is expanding access to two existing Gemini features across Gmail, Android Authority reported. AI Overview summarizes long email threads, while Help me write generates drafts and suggests improvements. Google says the rollout covers Workspace customers, Workspace Individual subscribers, and personal Google account users aged 18 or older with Gmail smart features enabled, and may take up to 15 days to complete.
Android Authority conducted hands-on testing of flagship smartphones powered by Qualcomm’s Snapdragon 8 Elite Extreme Gen 6, Google’s Tensor G6, MediaTek’s Dimensity 9600 Pro, and Apple’s A20 Pro. The comparison covers CPU, GPU, and AI technologies, but the supplied report excerpt does not provide the test results or detailed AI benchmarks. Samsung’s Exynos 2700 was excluded because it was expected to debut with the Galaxy S27 series.
InfoQ 中文 AI&大模型 reports that Manus replaced traditional programming tests with a four-hour task in which candidates use Manus to solve a real problem and explain their AI-assisted workflow. Manus co-founder and chief product officer Tao Zhang says product managers and designers on his team can now write code and submit repository changes, though the company’s job listings retain conventional engineering titles. A tech hiring analysis cited by Lenny Rachitsky adds that U.S. software-engineer hiring has fallen below half its 2022 peak, while FDE openings have increased sharply and AI-related keywords have become more prominent in product and design roles.
Android Authority reports that Google Photos will no longer create a separate copy for edits made with features such as Magic Eraser and Unblur, as well as Ultra HDR and C2PA images. On Android 11 and later, users can reopen saved photos and continue adjusting previous edits, with settings such as filters and brightness retained. Google has not specified a rollout schedule.
Google has begun rolling out Step-by-step Workouts to the Pixel Watch 3, Pixel Watch 4 and Pixel Watch 5, according to Android Authority. The feature provides wrist-based guidance for strength, mixed and Coach workouts, synchronizes routines between a phone and watch, and allows users to adjust weights and repetitions during a workout. Users can select a Coach-created plan, ask Google Health Coach to build a workout, or assemble a routine from Google’s exercise library in the Google Health app. Google says Coach-based workout creation requires a Google Health Premium subscription; its post does not attach the same requirement to routines built from the exercise library.
融中财经, republished by 36氪 and 钛媒体, argues that AI devices must do more than understand requests and context: they must judge whether to engage, how to intervene and when to stop. The report cites Thinking Machines Lab's Interaction Models research preview, Sesame's low-latency voice work and SYNAPUFF's human-machine co-existence research as examples of an emerging effort to treat interaction as a distinct capability requiring new evaluation methods.
Berkshire Hathaway has meaningful AI exposure through major holdings such as Google and the 2016 acquisition of PCC, according to a Jinduan analysis published by TMTPost. The article argues that its rise reflects a system combining low-cost insurance funding, wholly owned cash-generative assets and long-term capital-allocation authority, while Chinese peers such as Fosun International, Anfu Technology and Dongshan Precision face costlier debt, restricted control over assets and tighter institutional constraints. It also reports that the Philadelphia Semiconductor Index fell more than 15% in October while Berkshire reached a record share price on October 22.
嘉越月风投资创世号, a private fund managed by Wu Yuefeng, had a unit net asset value of 0.3242 yuan as of November 6, 2026, according to 私募排排网 data cited by 36氪. The report says the fund reached a peak annual return of 257% before its AI-chain holdings drove an 83.79% drawdown from above 2 yuan. The fund’s concentrated exposure to AI computing infrastructure, memory chips, semiconductor equipment and optical modules amplified both gains and losses, while its warning and stop-loss lines had previously been removed from the product terms.
36氪 reports that Anthropic is planning a Nasdaq IPO, with investor valuation expectations falling from $2 trillion to $1.5 trillion and then to discussion near $1 trillion as technology stocks adjusted, infrastructure financing costs rose, and investors reassessed its financial performance. The report also analyzes how Anthropic’s revenue growth, GAAP operating loss, adjusted profitability, and long-term compute commitments may affect valuation expectations for Chinese model developers such as DeepSeek, Moonshot AI, Zhipu AI, and MiniMax.
Anthropic’s updated IPO filing set a preliminary price range of $70 to $78 per share and disclosed substantial non-voting holdings for Amazon and Google. After the offering and private placement, Amazon is expected to hold 14.7% of Anthropic’s outstanding shares and Google 13.7%, according to a report on the filing. The filing also states that Amazon could hold 23.8% of the outstanding shares if its bonds were fully converted and the share count otherwise did not change. A separate report estimated that Amazon’s roughly 20% theoretical stake could be worth about $258 billion at the range midpoint, while New York Times reporting estimated values of more than $250 billion for Amazon’s $18 billion investment and $140 billion for Google’s $33 billion investment. These are theoretical estimates and depend on assumptions about conversion, the final share count and the IPO price.
Google’s Android Developers account published Android Bench results for GPT-6.1 Terra, Kimi K3.1 Pro and DeepSeek V4.1 Flash, according to IT之家. With kimi-code, Kimi K3.1 Pro recorded a 24.0% pass rate and 82.3% average completion across 30 long-horizon tasks, taking 27.4 hours and costing US$317.4 per run. DeepSeek V4.1 Flash, evaluated with opencode, recorded a 17.3% pass rate and 71.2% average completion, with a 12.0-hour average completion time and US$9.9 run cost.
IT之家 AI reports that Anthropic has begun an IPO roadshow and plans to issue 825 million shares in a public offering that could raise up to $64.35 billion. The report says a separate $26 billion private placement could bring total financing to $90.4 billion at a valuation of $1 trillion to $1.4 trillion, with Google and Nvidia reportedly participating.
Anthropic began its IPO roadshow and said in an amended SEC prospectus that it plans to list on Nasdaq under the ticker “ANT,” with a proposed price range of $70–$78 per share. The report says the maximum valuation is about $1.4 trillion on a fully diluted basis, while the valuation excluding potential shares such as options and restricted stock is about $1 trillion. The offering would comprise 825 million Class A shares, including about 801 million newly issued shares and 23.65 million shares offered by existing holders. Underwriters could buy up to about 124 million additional shares from Anthropic within 30 days. The prospectus also proposes a $26 billion private placement of Class N common stock by companies affiliated with Google, Nvidia, AMD, Broadcom, Micron, Microsoft and TSMC, contingent on completion of the IPO and other agreed conditions.
Google released A2UI 1.0, the first stable version of its open generative UI protocol, allowing agents to describe forms, charts and confirmation cards that applications render with their own components. The release adds bidirectional function calls, broader accessibility and metadata support, web and Flutter renderers, and agent SDKs for Python, TypeScript and Dart. Google says Atlassian, Alation, Bloomberg and ByteDance are already building with the protocol, while Gemini Enterprise support is still being upgraded and Jetpack Compose support for version 1.0 remains forthcoming.
Google released Gemini Omni Flash 1.2 on November 5, according to Artificial Anlys. The model supports text-to-video, image-to-video and video editing with native audio, producing clips up to 15 seconds; Google says it will replace version 1.1 on the Gemini API on December 3. Artificial Anlys ranked the model second on AA-Video-Editing v1.1, while it rose from 12th to sixth on AA-Video-T2V v2.1 and from 17th to 14th on AA-Video-I2V v2.1 compared with version 1.1.
Gemini CLI v0.68.0 adds an extension-validation command and evaluation tests, along with support for recognizing dotted Gemini 3 model names in multimodal function responses. The release also fixes terminal-width history rendering, surrogate-pair display truncation, case-insensitive root checks on Windows, rate-limit cooldown messages, extension installation races, and selected-model persistence during fallback or rewind operations.
Simon Willison released llm-gemini 0.36, a plugin for accessing Google’s Gemini model family, and reported that Gemini 4 Carbon had expanded beyond the subset of cyber defenders who received it two weeks earlier. He listed launch prices of $2 per million input tokens and $10 per million output tokens, scheduled to double to $4 and $20 on December 1. Willison also tested Carbon in Google’s Antigravity IDE, prompting it to create a Three.js mech animation using OpenStreetMap data.
What problem it solves
Coding agents need both exact design values and the rationale for applying them consistently. Without a structured, persistent representation of a design system, agents may miss design context or interpret visual guidance inconsistently.
Key features
Structured tokens and design rationale — The format combines machine-readable design tokens with human-readable Markdown ratio…
What problem it solves
Pi is designed for developers who need to adapt an agent to their own workflows rather than work within a fixed feature set. It provides interactive, automated, RPC, and SDK interfaces while leaving optional capabilities such as sub-agents and plan mode to extensions or packages.
Key features
Extensible agent harness — Customize the harness with extensions, skills, prompt te…
What problem it solves
Managing multiple AI service sessions and translating them into compatible API interfaces can require repetitive account switching, custom routing, and protocol-specific integration. The project addresses these tasks with a desktop application that combines account administration, protocol conversion, request scheduling, and model routing.
Key features
Account dashboard and …
What problem it solves
App Store and Google Play rejections can leave teams searching for the cause of a brief review message, fixing it, and resubmitting. The project addresses this with documented rejection patterns, concrete fixes, pre-submission checks, and an automated guard that can run before an upload.
Key features
Pre-submission project scanning — The guard scans iOS and Android projects …
What problem it solves
Researchers often need to turn unconstrained ideas and experimental logs into complete research manuscripts with literature synthesis, structured sections, refinements, and figures. PaperOrchestra addresses this gap by coordinating specialized agents across the paper-writing workflow and producing LaTeX manuscripts that follow target conference templates.
Key features
Multi-…
What problem it solves
Building AI chat experiences in React requires assembling chat primitives, runtime integrations, generative UI, and production UX details. The project provides a composable library that supports multiple backends, custom styling, and strongly typed APIs.
Key features
Composable Chat Primitives — Build a chat interface from primitives including Thread, Message, Composer, Thre…
What problem it solves
Autonomous agents accumulate state, require strict isolation, call model APIs and tool servers, and can continue consuming resources without supervision. AX addresses these requirements through declarative primitives for sandboxed execution, prepared workspaces, model configuration, lifecycle control, and direct inspection.
Key features
Declarative task manifests — Tasks, wo…
What problem it solves
Long-form videos need to be turned into short, platform-ready clips, but metered usage, platform branding, and server-side control can constrain the workflow. SupoClip addresses these concerns with a self-hostable, open-source pipeline that offers no-watermark output and a hosted option.
Key features
AI clip selection and scoring — A supported language model selects clip-wor…
What problem it solves
Software design often depends on context accumulated in an earlier conversation, while reviewers can miss assumptions and blind spots when they share that same context. EGM separates context-rich design work from fresh-context review so findings can be surfaced before implementation.
Key features
Context-aware design workflow — Explore requirements and trade-offs, save the a…
What problem it solves
AI coding tools often expect particular provider protocols, making it difficult to use the models and providers a user prefers. A local gateway can translate between supported client APIs and provider APIs while preserving capabilities such as streaming, tool calls, reasoning, and images.
Key features
Provider protocol translation — The gateway translates Responses, Chat Com…
What problem it solves
Server operators need to detect malicious behavior across logs and HTTP requests, then apply remediation at the appropriate layers of their infrastructure. They also need a way to benefit from threat intelligence shared by the community.
Key features
Integrated Security Engine — The Security Engine combines IDS/IPS and WAF capabilities in one solution. It analyzes log source…
What problem it solves
AI clients change between sessions, making it difficult to preserve context, recover earlier decisions, and continue unfinished work. The project provides a local Markdown-based memory that records selected information and retrieves it with its source in later sessions without automatically converting the entire conversation history into knowledge.
Key features
Persistent Ma…
What problem it solves
The hidden system instructions that shape AI product behavior are difficult to inspect and change across models, products, and versions. This repository makes collected prompts available in a structured, product-oriented archive for study and comparison.
Key features
Broad product coverage — The archive covers prompts from major AI providers and products, including Anthropic…
What problem it solves
Building AI agents and workflows often requires combining visual orchestration with custom code while connecting them to existing systems. The platform addresses these needs with model flexibility, self-hosted or cloud deployment, workflow logic, human approvals, and observability.
Key features
Visual and code-based workflow building — Build workflows on a visual canvas and …
What problem it solves
LLMs and AI agents need structured access to Google Ads account data, API metadata, and management operations. The server provides this access through MCP while adding a two-step proposal and confirmation workflow for write operations.
Key features
Query Google Ads accounts — Use GAQL to retrieve information about a Google Ads account, list directly accessible customers, and…
What problem it solves
AI agents need a reliable way to perform browser tasks such as finding availability, selecting options, handling CAPTCHAs, and completing forms. Developers also need flexibility to run the agent locally or through hosted infrastructure while choosing their model, browser mode, tools, and output format.
Key features
Persistent Python agent runtime — The model writes Python in…
What problem it solves
Coding agents often default to legacy patterns because their training data contains substantial amounts of legacy code, which can produce bloated JavaScript where native web platform solutions are available. They may also know that a modern API exists without having enough real-world implementation patterns to apply it correctly in production-ready code.
Key features
Context…
What problem it solves
Go applications need client libraries for interacting with Google Cloud Platform APIs. The project provides packages for those services and supports common authentication flows for local development and Google Cloud Platform environments.
Key features
Google Cloud API client packages — The repository contains Go packages for Google Cloud Platform services, including storage,…
What problem it solves
Coding agents need a structured way to create, manage, and build from Google Stitch designs while integrating with the Stitch MCP server. The project organizes design, code generation, and supporting workflows into skills that can be installed with supported coding agents.
Key features
Design workflows — The design plugin covers converting frontend code into Stitch designs, …
What problem it solves
Market platforms can place broad access to prices, charts, company information, watchlists, and alerts behind paid subscriptions. OpenStock addresses this by providing a community-built platform that can be hosted publicly or self-hosted with every feature.
Key features
Multi-market dashboard — Explore eight markets through a dashboard with an index pulse, sector heatmap, ma…