The Informed Layer · Newsscan
Daily briefing

Agents enter production; US preempts state AI

← All issues

Anthropic / Claude ecosystem

Claude Managed Agents add scheduled deployments and credential vaults

Anthropic's managed agents now support cron-based scheduling, environment variable vaults, and dynamic workflows in general availability. The beta features enable autonomous agent operation on a schedule with secure credential management.

Frontier model providers

OpenAI brings models and Codex to Oracle Cloud customers

OpenAI models and Codex are now available through Oracle Cloud, allowing Oracle customers to access them procurement-free via existing Oracle commitments. This expands OpenAI's footprint beyond its own cloud partnerships.

ChatGPT can now buy things for users after Visa payment network integration

Visa plugged its payment network into ChatGPT, enabling AI agents to autonomously shop and pay for users with built-in fraud monitoring and spending controls. The integration is branded Visa Intelligent Commerce.

AI developer tooling & infrastructure

Cursor's Bugbot gets 3x faster, 22% cheaper, and finds 10% more bugs

Cursor's AI code review tool, Bugbot, now reviews pre-push changes faster and more accurately, with support for partial PR recurrences. The update improves developer feedback loops and reduces CI costs.

Nous Research ships Hermes Agent Profile Builder for one-click agent assembly

Nous Research released a browser-based flow that lets users assemble Hermes Agent profiles by selecting identity, model, skills, and MCP servers all within a local dashboard. It simplifies the creation of task-specific agents without coding.

TiDB launches Agent State Stack for production-grade AI agent memory

At SuperAI Summit Singapore, TiDB introduced Agent State Stack, a distributed memory layer that gives AI agents durable, persistent state at scale. It targets production deployments where agents need to remember across sessions.

Cloud & platform providers

Microsoft Foundry adds Cohere Command A+, Chandra OCR 2, and GLM-OCR open-weight models

Microsoft Foundry now supports deploying three open-weight models via Hugging Face: Cohere Command A+, Datalab Chandra OCR 2, and Z.ai GLM-OCR. This shift highlights a cloud strategy favoring open-weights cost and speed for document and agentic workloads.

AWS Transform tool helps migrate AI workloads to Amazon Bedrock

AWS introduced Transform, a structured migration approach to move existing AI workloads to Amazon Bedrock. The tool provides discovery, assessment, and execution steps to simplify the transition from self-managed ML infrastructures.

AI policy, regulation & governance

White House and Senate revive push to block state AI laws through kids safety deal

Congress is negotiating a federal preemption of state AI laws by coupling the Kids Online Safety Act and NO FAKES Act. The move aims to create a uniform national AI framework and prevent a patchwork of state regulations.

Australia's National AI Plan released, focusing on inclusive economy and data-centre investment

The Australian government published its National AI Plan, prioritizing an inclusive AI economy, data-centre infrastructure, and workforce skilling, while avoiding broad high-risk AI guardrails. Critics question who will benefit from the investments.

Industry & market moves

Anthropic taps TCS to scale enterprise AI deployments across India

Anthropic partners with Tata Consultancy Services (TCS) to distribute Claude AI into financial services, healthcare, and other sectors. This is Anthropic's first major enterprise distribution deal through a global system integrator, aiming to accelerate adoption in regulated industries.

Meta partners with Reliance on India's largest AI data center, 168 MW Jamnagar facility

Meta and Reliance Industries closed a deal for a 168 MW AI data center in Jamnagar, Gujarat. The facility will be cooled by seawater and powered by renewables, marking Meta's first AI-enabled data center in India and deepening its local data localization efforts.

Meta commits $115M to craft labor training for AI data center construction

Meta is investing $115 million into 'America's Workforce Academy', a skilled-trades training program that guarantees jobs for graduates. The program aims to speed up AI data center buildout across the US by training electricians, welders, and other craftspeople.

Cursor opens London HQ, plans 200 EMEA hires amid SpaceX's $60B acquisition option

Cursor's parent Anysphere opened a new London headquarters and plans to hire 200 people across EMEA. Simultaneously, SpaceX holds an option to acquire the company for $60 billion, adding strategic uncertainty to its European expansion.

OpenAI acquires Ona to strengthen Codex cloud capabilities

OpenAI is acquiring Ona, a company that specializes in cloud infrastructure and tooling for AI development. The move aims to expand Codex's cloud hosting and enterprise deployment features.

DXC and Anthropic announce multi-year global alliance for mission-critical AI

DXC Technology will embed Claude across its OASIS platform and customer environments, with a dedicated Claude-certified workforce. The alliance targets mission-critical workloads in insurance, healthcare, and the public sector.

Reka and Moonvalley merge to advance physical AI models and infrastructure

AI research startup Reka and AI studio Moonvalley are joining forces to accelerate the development of models and infrastructure for the physical AI era. The combined entity aims to push the frontier of multimodal, embodied intelligence.

NEURA Robotics raises up to $1.4B series C from Amazon, NVIDIA, and Tether

European humanoid robotics company NEURA Robotics secured a series C round of up to $1.4 billion, backed by Amazon, NVIDIA, and Tether. The funding could make it Europe's best-funded humanoid maker.

Jeff Bezos' AI startup Prometheus closes $12 billion round at $41B valuation

Prometheus, an AI company focused on engineering, manufacturing, and drug design, closed a $12 billion funding round, valuing it at $41 billion. The massive raise signals confidence in application-specific AI beyond chatbots.

AI product & feature launches

Nvidia unveils Isaac GR00T open humanoid robot platform for research

Nvidia introduced Isaac GR00T Reference Humanoid Robot, an open platform combining hardware specifications and software to accelerate humanoid robotics research. It offers a standardized base for labs and startups.

Teradyne Robotics debuts production-ready physical AI applications at Automate 2026

Teradyne Robotics showcased a portfolio of AI-enabled automation solutions at Automate 2026, including collaborative robots and quality inspection systems powered by physical AI algorithms.

Recursive launches automated AI research system that optimizes training pipelines

Stealth startup Recursive, co-founded by Cong Lu, demonstrated an automated system that autonomously discovers and optimizes AI training workflows and GPU kernels. It achieved state-of-the-art benchmarks on nanoGPT and SOL-ExecBench and open-sourced its artifacts.

Coinbase debuts AI agent that can trade and pay for premium research

Coinbase launched an AI agent integrated with Coinbase Advanced that can autonomously execute trades and pay for premium research using the x402 protocol. The agent operates with defined spending limits and wallet controls.

Cresta launches Conductor, an agent for building and managing AI agents

Cresta introduced Conductor, a tool that acts as an AI agent to create, deploy, and monitor other AI customer conversation agents. It aims to reduce the complexity of agent lifecycle management in contact centers.

Research with immediate practical relevance

New AI model enables large-scale molecular screening with few GPUs

Imperial College researchers released DeCAF-Pearl, a flow map model that performs all-atom cofolding inference, enabling screening of up to one million molecules against a protein target in about 18 hours on just 64 GPUs. This makes practical large-scale drug discovery accessible.