Gemini Mac Computer Use, Liquid AI & VMware Zero-Day
Google launches Gemini computer use on Mac, Liquid AI debuts d1.2 models on Vercel, and VMware patches critical zero-day. Read the Nov 20 briefing.
Top Engineering Highlights
- • Google Gives Gemini Computer-Use Capabilities on macOS: The new desktop integration allows Gemini to parse screen coordinates, click UI elements, and automate multi-app desktop workflows directly on Mac hardware.
- • Liquid AI Deploys d1.2 and d1.2-Ultra Non-Transformer Models on Vercel: Liquid AI's proprietary dynamical system architecture models arrive on the Vercel AI Gateway, delivering sub-millisecond per-token latency with 256k context.
- • Broadcom Patches Critical VMware vCenter Flaw Exploited as Zero-Day by Chinese APT: Security teams confirm active exploitation since September of an unauthenticated remote execution flaw in VMware Directory Service, urging immediate ESXi isolation.
- • Google MaxText Adds Streaming DiLoCo for Multi-Metro LLM Pre-Training: Google Brain and Cloud researchers enable distributed LLM pre-training across geographically separated data centers with minimal inter-cluster sync bandwidth.
- • Amazon SageMaker AI Adds Native Latency Drift and Idle GPU Detection: AWS introduces default automated telemetry across vLLM inference endpoints to flag cluster drift and underutilized GPU instances through AWS Health.
Strategic Executive Synthesis
Client-Side Computer Use Bridges Frontier Models into Operating System Fabrics:
Google's macOS desktop agent release reflects the shift from pure chat interfaces to autonomous GUI actuators executing arbitrary system tasks directly through native OS accessibility trees.
Non-Transformer Architectures Achieve Commercial Parity in Edge Latency:
Liquid AI's d1.2 deployment demonstrates that continuous-time neural networks provide compelling cost and speed advantages for latency-critical agent loops without sacrificing context retention.
Enterprise Virtualization Infrastructure Faces Sustained Targeted Exploitation:
The zero-day exploitation of VMware vCenter highlights persistent nation-state targeting of core virtualization control planes to bypass perimeter boundary defenses.
Google Gives Gemini Computer-Use Capabilities on macOS
Google rolled out an experimental computer-use capability for Gemini on macOS, allowing the desktop application to interact directly with GUI elements across open applications. By tapping into native macOS Accessibility APIs and visual grounding models, the assistant can observe screen states, parse clickable interface coordinates, and execute mouse movements and keystrokes autonomously.
The rollout follows Anthropic's earlier computer-use release but integrates deeply into local desktop applications like Finder, terminal emulators, and local IDEs. Google enforces strict user confirmation boundaries on privileged system actions, requiring explicit authorization before the agent executes terminal commands, accesses credentials, or writes to external network sockets.
Liquid AI Launches d1.2 and d1.2-Ultra Non-Transformer Models on Vercel
Liquid AI announced the general availability of its flagship d1.2 and d1.2-Ultra foundation models on the Vercel AI Gateway. Built on continuous-time dynamical neural architectures rather than standard self-attention transformers, the models exhibit linear time complexity with respect to sequence length, unlocking sub-millisecond per-token latency during live streaming inference.
Featuring an expansive 256k token context window, the models match or exceed transformer baselines on code synthesis and financial document reasoning while requiring a fraction of the memory footprint. The architecture delivers exceptional cost efficiencies for developers building high-throughput agent swarms and real-time voice processing pipelines.
Advertisement
Broadcom Patches Critical VMware vCenter Flaw Exploited as Zero-Day by Chinese APT
Broadcom shipped emergency point patches for VMware vCenter Server addressing a critical authentication bypass vulnerability that state-sponsored Chinese espionage groups have exploited in zero-day attacks since September 2026. The vulnerability in the VMware Directory Service daemon allows unauthenticated remote adversaries to forge session tokens and achieve administrative privileges on virtualized data center clusters.
Cybersecurity researchers confirmed active in-the-wild exploitation targeting government, telecommunications, and high-tech virtualization fleets to execute lateral pivoting across ESXi hypervisors. Cloud infrastructure teams are strongly advised to apply the vendor updates immediately, rotate root service account credentials, and restrict vCenter management ports away from public-facing boundaries.
Google MaxText Adds Streaming DiLoCo for Multi-Metro LLM Pre-Training
Google researchers published an architectural breakthrough integrating Streaming DiLoCo (Distributed Low-Communication) into the open-source MaxText JAX framework. The technique enables distributed foundation model pre-training across distinct metropolitan data center clusters separated by hundreds of miles without requiring multi-terabit dedicated optical dark fiber.
By executing hundreds of inner optimization steps locally within each cluster before synchronizing outer momentum weight deltas over public WAN connections, Streaming DiLoCo reduces cross-cluster bandwidth requirements by over 500x. The advancement allows AI operators to pool geographically dispersed GPU/TPU capacity, overcoming municipal electrical power caps that restrict single-site data center expansion.
Amazon SageMaker AI Adds Native Latency Drift and Idle GPU Detection
AWS rolled out native operational telemetry enhancements for Amazon SageMaker AI, enabling automated detection of latency drift and underutilized accelerator hardware across managed generative AI inference clusters. By integrating runtime telemetry hooks into standard vLLM serving containers, SageMaker continuously correlates request concurrency with per-token response times.
When inference latency degrades past established baselines or GPU compute sits idle during demand valleys, the platform publishes actionable alerts directly to AWS Health dashboards and CloudWatch metrics. ML platform engineers can leverage these automated signals to dynamically re-scale cluster instance counts, adjust speculative decoding draft ratios, and curtail runaway cloud infrastructure expenditures.
Advertisement
AI Engineering Teams Move to Drop Bulky AGENTS.md and CLAUDE.md Files
A growing architectural consensus among AI engineers advocates deprecating exhaustive, static AGENTS.md and CLAUDE.md project instruction files. Engineering audits revealed that bloating system contexts with thousands of lines of defensive prompt rules dilutes attention heads, inflates session token bills, and increases coding hallucination rates.
Instead of monolithic static prompt files, leading teams are adopting just-in-time context injection powered by Model Context Protocol (MCP) servers and deterministic pre-commit linters. By offloading rule verification to external static analysis tools, coding agents retain focused context windows dedicated strictly to the immediate refactoring task.
PHP Releases Urgent Security Patches for Seven Critical RCE Vulnerabilities
The PHP development group shipped emergency point releases for PHP 8.1, 8.2, and 8.3 addressing seven critical vulnerabilities, including multiple remote code execution (RCE) flaws. The most severe flaw involves buffer overflow and heap corruption defects within the core multipart request parser and multibyte string processing extensions.
Security researchers confirmed active proof-of-concept exploit chains capable of achieving unauthenticated shell execution by transmitting maliciously crafted HTTP multipart payloads to vulnerable endpoints. Web engineering teams are urged to deploy the upstream patches immediately or enforce strict web application firewall (WAF) filtering on multipart boundary headers.
More Technical Stories Covered Today
- Dreamcoder: Coding Harness from the Author of Mastering Emacs
- The Fallout from Nvidia's Jensen Huang Market Controversy Continues
- iPhone Air 2 Dummy Unit Shows Off New Dual-Camera Layout
- An Experimental Web Browser Written from Scratch in Rust by an AI Agent
- Industry Reactions to Major Outages and the AI Patching Debate
- Harness Adapters: Adapting Your Favorite Model to Custom Testing Harnesses
- Still Tracking Passwords in Spreadsheets? Enterprise Credential Security Risks
- UN Court Cites Military Social Media Posts in Landmark Accountability Ruling
- Vendor Data Breach Exposes Chase, Visa, and Mastercard Customer Information
- Apple TV Unveils Return of Flagship Sci-Fi Series for New Season
- F5 Releases Emergency Patches for Critical BIG-IP Remote Code Execution Vulnerabilities
- European Commission Confirms AI-Assisted Cyberattack Disrupted Sysper and MyPMO Platforms
- Ruby Patches Ruby2Shell Heap Overflow Vulnerability (CVE-2026-132825) Leading to RCE
- Anthropic IPO Could Yield $156B for Charity via Long-Term Benefit Trust Allocation
- Windows 11 KB5132057 Update Released with 28 System Changes and Kernel Security Fixes
- Janax: Hand-Written x86/ARM64 Assembly OS Where an Autonomous Agent Self-Compiles
- Security Bite: Analysis of Two Months of Meta Muse AI Security and Data Privacy Flaws
- The U.S. National Cyber Director Steps Down Amid Federal Cybersecurity Realignment
- Veterans of the 2008 Crash Protest Anthropic's Record-Breaking IPO Outside Nasdaq
- Building Reliability Into the Cloud: Observability as the Distributed Control Plane
- WhatsApp Rolling Out Shared Custom Chat Themes on Android
- Improvements on the Pull Request Overview in Public Preview: GitHub Collaboration Upgrades
- AI Coding Agents on Amazon EC2 Mac: How Cognition Scales Devin for macOS Environments
- WhatsApp Raising Group Chat Limit to 1,500 Participants for Large Communities
- Who Are Anthropic's Seven Co-Founders, Now Holding Over $180B Following Market Debut?
- Fiona the Hippo Died After Eating a Hat: How Does a Hat Kill a Hippo?
- Anthropic Was Supposed to Anchor the AI Trade: Market Analysis and Post-IPO Volatility
- OpenAI Begins Testing Dots Visual UI with ChatGPT Plus Users and Business Tiers
- Using NVIDIA Run:ai GPU Fractioning on Amazon EKS for Multi-Tenant Workloads
- Microsoft Consolidating Task Ecosystem into Planner Across Desktop and Web
- AWS Releases a Free Local Emulator of AWS Built Entirely by Agents
- You Can Now Buy a Steam Machine Without Getting on a Waiting List: Valve Opens Direct Sales
- FCC Greenlights Rule Allowing Prisons to Jam Contraband Cellphone Signals
- Meta Now Sells and Rents 3D Movies on Quest VR Headsets via Digital Storefront
- iPadOS 27.3, watchOS 27.3 & More Get Public Beta 2: Full Changelog & Developer APIs
- macOS 27.3 Public Beta 2 Now Available: Full Disk Access Privacy Controls & Expanded Siri AI
- A Brief History of Nvidia's Recent Stock Fluctuations and AI Hardware Cycles
- Apple Developing Native Autonomous AI Agent to Control macOS System Operations
- nginx2root: Five New Nginx Memory Corruption Vulnerabilities and Origin Shielding
- How Country Road Group Built a Pattern-Driven Serverless Integration Platform on AWS
- Use Kimi Code CLI with Foundation Models on Amazon Bedrock
- Amazon SageMaker AI Inference: 2026 Generative AI Architectural Retrospective
- Google Introduces Tool Retrieval in the Gemini API for Semantic Function Calling
- FBI Contacts Claude User Following Flagged LLM Conversations: Legal & Privacy Analysis
- Transport for London Confirms 265,000 Oyster Card Accounts Compromised
- AI Bubble Protests Erupt Outside Nasdaq Following Anthropic Market Debut
- GPT-Live 1.1 Now Available on Vercel AI Gateway: Real-Time Audio & Streaming
- Apple Releases iOS 27.3 Public Beta 2 and Developer Beta 2 with Apple Intelligence APIs
- ChatGPT Health Feature Now Works with Mac App, Dots, and Connected Devices
- What's With All These Outages? Multi-Region Cloud Cascades & Resilience Engineering
- Amazon Bedrock AgentCore Adds Strands Harness Defaults Built-In for Multi-Agent Workflows
- Let Gemini Tackle Repetitive Chores on macOS: Desktop Automation Deep Dive
- Five Agent Chores: $2.58 on a Top Model vs. $0.11 on Small Models — Token Efficiency Benchmarks
- Muscle Memory: Repeated Claude Code Tasks Run with Zero Tokens via Deterministic Caching
- Elastic and HoneyHive Join Forces to Advance AI Observability and Evaluation Stacks
- Tech Outages Resurrect Y2K-Era Systemic Cascades in AI Agent Stacks
- ASUS Warns of Critical Hard-Coded Key Flaw in Control Center: Remote Exploit Analysis
- Amazon Polly Launches Neural Voice Cloning: Train Custom Brand Voices in Minutes
- PHP Security Updates Fix 18 Flaws Including Critical Remote Code Execution Bugs
- Next Year's iPhone Pro Redesign: Key Architectural & Hardware Leaks
- Debug a Failed Deploy with Claude: From Red Build to Clean Rollout in Five Steps
- Claude Chalk: A Terminal Mod that Lets Claude Code Color Its Output and Highlights
- Why Engineering Teams Need Better Open Source Protocols Instead of More Single-Purpose Apps
- Former Alibaba Exec's Everful Raises $48.5M to Automate Global Supply Chain Coordination
- Anthropic at $1.4T: Deep Discounted Cash Flow and Multiple Analysis Against Hyperscalers
- The Blue Economy: Ocean Tech Infrastructure and Autonomous Marine Data Gathering
- JPMorgan Chase Discloses Third-Party Data Breach Unrelated to Core Infrastructure
- Anthropic Prices IPO Shares at $83, Surpassing Initial Target Range on Nasdaq
- Google Launches Gemini 4 Neon with Carbon Quality at Low-Latency Flash Speeds
Tech Pulse Daily
Get tomorrow's pulse first
Join engineers who read Tech Pulse before stand-up. Free, weekday mornings.