The Informed Layer · Newsscan
Daily briefing

Agents drive product growth; Google pushes Gemini

← All issues

Anthropic / Claude ecosystem

Claude Code v2.1.139 Introduces Agent View and /goal Command

Anthropic's Claude Code v2.1.139 now features an 'Agent View' for unified session visibility, allowing developers to track multiple concurrent coding sessions from a single dashboard. The update also includes a new '/goal' command for autonomous multi-turn task completion with real-time cost tracking, enhancing developer productivity.

Thomson Reuters and Free Law Project Launch MCP Integrations with Claude

Legal research providers Thomson Reuters and Free Law Project have launched live MCP (Model Control Plane) integrations with Claude. This enables direct AI access to their respective legal databases—CoCounsel Legal and CourtListener—allowing AI to query and utilize legal data directly.

Frontier model providers

Google's Gemini Omni Leaked Ahead of Google I/O 2026, Pointing to New Video Generation Model

Reports indicate Google is preparing to launch 'Gemini Omni,' a new video generation model, possibly a successor to Veo 3.1 or a true multimodal architecture. The leak suggests a consumer release in the Gemini app ahead of Google I/O 2026.

Google Accelerates Gemini Integration Across Android, Chrome, and Android Auto Ahead of Apple's AI Reboot

Google is aggressively integrating Gemini as a core AI operating layer across its Android ecosystem, including multi-app task automation, Chrome updates, and a redesigned Android Auto. This strategic move aims to solidify Gemini's presence before Apple's anticipated AI Intelligence launch at WWDC.

xAI Ships Grok Connectors with 8 Integrations and BYO MCP Support for Enterprise

xAI has launched Grok Connectors, closing the enterprise integration gap for Grok Business by providing native integrations with Gmail, Workspace, SharePoint, Outlook, OneDrive, Notion, GitHub, and Linear. The release also includes bring-your-own MCP (Model Control Plane) support.

AI developer tooling & infrastructure

Signadot Launches New Skill for Claude Code, Codex, and Cursor to Validate Changes in Kubernetes Environments

Signadot has introduced a new '/signadot-validate' skill that enables AI coding agents like Claude Code, Codex, and Cursor to validate their code changes directly against production-like Kubernetes environments. This closes the 'agent loop' by allowing agents to perform real-time validation with actual dependencies before code is handed back to developers.

Coder Agents Enable Running AI Coding Workflows on Self-Hosted Infrastructure

Coder Agents introduces a new capability allowing organizations to run self-hosted AI coding workflows on their own infrastructure. This offers model flexibility, avoids vendor lock-in, and centralizes control over execution, policies, and observability for AI development.

Cloud & platform providers

Oracle AI Database@AWS Now Available in Switzerland

Oracle AI Database@AWS is now available in the AWS Europe (Zurich) Region. This expansion enables Swiss organizations to migrate their on-premises Oracle workloads to the cloud with zero-ETL integrations to other AWS services, providing local access to advanced AI database capabilities.

Amazon's 'Titus' Project Accelerates AI Data Center Deployment with Liquid Cooling

Amazon's 'Titus' project is focused on future-proofing its AI data centers through the rapid deployment of new liquid-cooling systems and modular designs. This initiative aims to support next-generation, power-hungry GPU systems required for advanced AI workloads.

Google Cloud Launches Cloud Storage Rapid for High-Performance AI Workloads

Google Cloud has released Cloud Storage Rapid, a new object storage solution designed to address AI storage bottlenecks. It delivers over 15+ TB/s throughput and sub-millisecond latency, aiming to keep GPUs and TPUs fully saturated during trillion-parameter model training with its Rapid Bucket and Rapid Cache features.

Amazon Connect Expands into Four Agentic AI Solutions for Enterprise Workflows

Amazon is expanding its Connect platform into four purpose-built agentic AI solutions: Amazon Connect Decisions, Amazon Connect Talent, Amazon Connect Customer, and Amazon Connect Health. These solutions are designed to integrate into existing workflows for supply chain, hiring, customer engagement, and healthcare, rather than requiring organizational restructuring.

AI policy, regulation & governance

U.S. Spy Agencies Seek More Power in Trump Administration Battle Over AI Model Evaluation

Internal divisions within the Trump administration are surfacing as U.S. intelligence agencies reportedly seek expanded authority in evaluating AI models. This signals a push for greater oversight and influence over AI development and deployment from national security entities.

Von der Leyen, Hillary Clinton Back New Youth AI Safety Institute to Childproof AI

A new multi-stakeholder institute, the Youth AI Safety Institute, has been launched with backing from EU leadership, including Ursula von der Leyen and Hillary Clinton. Its mission is to create and publish safety standards and ratings for AI applications targeting children, addressing growing concerns over tech harms to minors.

Canadian Digital Regulators Forum to Host Interactive Workshop on AI Impact

The Canadian Digital Regulators Forum, comprising CRTC, Competition Bureau, Privacy Commissioner, and Copyright Board, is convening stakeholders for an interactive workshop. The event will discuss AI's impact on competition, privacy, copyright, and telecommunications regulation in Canada.

Joint Cybersecurity Agencies Publish Guidance on Secure Adoption of Agentic AI

Major cyber agencies from five countries (Australia, US, Canada, New Zealand, UK) have jointly published guidance on the secure adoption of agentic AI. This coordinated effort treats agentic AI as an emerging operational security issue and provides best practices for mitigating risks.

Google Thwarts Hacker Group's AI-Enabled Mass Exploitation Attempt Using OpenClaw

Google's Threat Intelligence Group successfully detected and thwarted a coordinated hacker campaign that utilized AI models, including OpenClaw, to identify and exploit zero-day vulnerabilities at scale. This marked an attempt at an 'AI mass exploitation event.'

Industry & market moves

DeepSeek Prepares for China's Largest AI Funding Round at 50 Billion RMB

DeepSeek is undertaking China's largest-ever AI funding round, aiming to raise 50 billion RMB (approximately $7 billion USD). Founder Liang Wenfeng is reportedly contributing 40% of the capital, which would value the company at over 350 billion RMB.

OpenAI Forms $14 Billion Subsidiary to Help Businesses Deploy Custom AI Systems

OpenAI has launched a new enterprise-focused subsidiary with $4 billion in fresh capital and founding partners. This new venture aims to help businesses deploy custom AI systems and has acquired the consulting firm Tomoro to accelerate corporate AI adoption.

White Circle Raises $11 Million to Develop Guardrails for AI Models in the Workplace

Paris-based AI safety startup White Circle has secured an $11 million funding round with investment from senior staff at leading AI labs. The funding will be used to develop guardrails that prevent misuse and ensure safe deployment of AI models in enterprise environments.

AI product & feature launches

Nomagic and Brack.Alltron Expand Partnership to Deploy Vision-Language-Action Systems in Warehouses

Nomagic and Brack.Alltron are expanding their partnership to integrate Vision-Language-Action (VLA) systems into live warehouse operations. This deployment aims to enable autonomous shifts and enhance task autonomy within logistics.

Tether Launches Decentralized Local AI Platform Using Isaac Asimov's Psychohistory Concept

Tether has launched a decentralized local AI infrastructure, allocating stablecoin reserves and operating profits to edge models via its QVAC Psy platform and SDK. This initiative positions edge-native AI as a resilient alternative to centralized cloud-based frontier AI, drawing inspiration from Isaac Asimov's Psychohistory.

Thinking Machines Lab Unveils Highly Responsive AI Model for Real-Time Humanlike Interactions

Thinking Machines Lab has introduced a new dual-model architecture, featuring the TML-Interaction-Small model, which enables full-duplex AI communication with sub-0.4 second latency. This breakthrough allows AI to listen, see, and respond simultaneously in real time, mimicking natural human conversation.

AKOOL Unveils Breakthrough AI Video Inference Engine for Real-Time AI Video at Scale

AKOOL has launched a new AI Video Inference Engine, which promises 10-20x speed gains and enables real-time streaming with sub-30ms latency. This full-stack optimization reduces AI video generation from tens of seconds to 1-3 seconds, shifting AI video from post-production to a live interactive medium.

Nokia Launches Agentic AI for Home and Broadband Networks

Nokia has embedded autonomous AI agents across its broadband network platforms, including Altiplano, Corteca, and Broadband Easy. This initiative aims to enable self-optimizing infrastructure, with targets of a 50% lift in helpdesk resolution and a 50% reduction in field visit costs.

SAP and Cyberwave Deploy Fully Autonomous AI-Powered Robots in Live Logistics Warehouse

SAP has operationalized Physical AI in its own logistics warehouse at St. Leon-Rot, partnering with Cyberwave to deploy fully autonomous AI-powered robots. This demonstrates that robots can perform high-variability tasks like box folding and packaging in live enterprise environments, integrated with SAP Logistics Management.

New Emerson Industrial AI Platform 'AspenTech AVA' Delivers Enterprise-Scale AI

Emerson has launched AspenTech AVA, a new enterprise-scale industrial AI platform. The platform is designed to embed domain expertise and agentic workflows to accelerate AI adoption in manufacturing and operations.

Xero Delivers Claude Integration to Advance AI-Powered Financial Intelligence

Xero has launched a live integration with Claude, allowing small business users to access financial data directly within Claude conversations. This partnership aims to advance AI-powered financial intelligence by eliminating the need to switch between tools for financial queries.

Exterro Launches Industry's First Autonomous AI Engine to Eliminate Up to 95% of Manual Subpoena Work

Exterro has launched its first autonomous AI engine for subpoena management, branded as Exterro Subpoena Manager. The company claims this engine can reduce manual work by up to 95% and save enterprises over $500,000 annually by reclaiming 7,500 enterprise hours.

Research with immediate practical relevance

Google DeepMind's AI Controls Nuclear Fusion Reactor in Real Time

Google DeepMind has achieved a breakthrough in nuclear fusion research by successfully teaching AI to control a tokamak fusion reactor in real time. This marks the first application of reinforcement learning to the real-time magnetic control of a tokamak, demonstrating the ability to manage complex plasma configurations and sustained dual-plasma droplets.

Meta and Stanford Researchers Propose Fast Byte Latent Transformer to Reduce Inference Memory Bandwidth by Over 50%

Meta and Stanford University researchers have introduced three new inference acceleration techniques for byte-level language models, collectively called the Fast Byte Latent Transformer (BLT-D, BLT-S, BLT-DV). These methods reduce memory bandwidth by 50–92% compared to standard BLT, eliminating the need for tokenization while maintaining or improving generation speed.

AI Model from Korea Advanced Institute of Science and Technology Learns to Admit Uncertainty

Researchers at the Korea Advanced Institute of Science and Technology (KAIST) have developed a warm-up training method that teaches AI models to admit uncertainty and respond with 'I don't know' instead of fabricating confident, false answers. This breakthrough, published in Nature Machine Intelligence, mimics human brain processes for uncertainty recognition.