The Informed Layer · Newsscan
Daily briefing

Custom silicon surges; robotics funding booms

← All issues

Anthropic / Claude ecosystem

Anthropic Launches $5M Research Grant Program for AI Wellbeing Evaluations

Anthropic has committed $5 million to fund independent, open-source evaluations examining how AI models influence human wellbeing. Applications for the grant program are open through September 21, 2026.

Anthropic Updates Claude Tag to Allow Unprompted Slack Agent Participation

Anthropic has rolled out an update to Claude Tag in Slack, allowing the assistant to read full conversation threads and autonomously determine when to contribute. The proactive context-reading mechanism reportedly boosts decision accuracy by roughly 30%.

Frontier model providers

OpenAI Details 'Jalapeño' Custom Inference Silicon Architecture

OpenAI has officially detailed Jalapeño, its first custom inference processor designed to power frontier model workloads. The proprietary silicon delivers measurable improvements in token throughput per kilowatt and lower inference latency compared to commercial off-the-shelf accelerators.

OpenAI Adds AI Sticker Maker to ChatGPT

OpenAI has rolled out a custom sticker maker in ChatGPT, allowing users to generate and export personalized sticker packs directly into messaging applications such as WhatsApp and iMessage.

AI developer tooling & infrastructure

Meta Unveils MTIA 300 Training Silicon with Integrated Network Interfaces

Meta has announced MTIA 300, its first custom training chip featuring built-in network interface cards (NICs) and hardware communication-offloading engines. Paired with Meta's HCCL library, the architecture delivers 3.9x faster communication in recommendation model training than equivalent GPU setups.

LangChain Releases Upgraded LangSmith Engine with 2x Issue Detection

LangChain has launched a major update to LangSmith Engine, claiming a 2x performance gain in agent issue detection benchmarks and a 25% improvement in automated fix quality. The engine is also made available for self-hosted enterprise deployments.

Cloud & platform providers

NVIDIA Unveils Vera Rubin LPX and CPX Inference Platforms

NVIDIA has introduced its Vera Rubin LPX and CPX inference platforms, purpose-built for agentic and long-context AI workloads. The platforms deliver up to 4x faster token generation speeds and lower token operating costs, with immediate cloud partner deployments underway.

AWS Open-Sources Voice-First Knowledge Accelerator on Bedrock

AWS has released an open deployment prototype for a voice-first knowledge retrieval accelerator utilizing Amazon Bedrock and OpenSearch Serverless. The system integrates conversational avatars for automated institutional knowledge access across industrial and public sector environments.

AI policy, regulation & governance

Australian Government Outlines Mandatory Standards for AI Data Centres

Australian Prime Minister Anthony Albanese has detailed upcoming legislation establishing national minimum standards for AI data centres, covering water consumption, net-positive renewable energy requirements, and security rules. The federal framework aims to harmonize rules across states ahead of National Cabinet discussions.

SEC Opens Investigation into AI Hedge Fund Situational Awareness

The US Securities and Exchange Commission has initiated a probe and issued subpoenas to financial institutions connected to AI hedge fund Situational Awareness. The regulatory action follows severe trading losses incurred by the firm's algorithmic strategies in July 2026.

Industry & market moves

TCS Acquires Porsche Subsidiary MHP for €320M in €1.25B AI Partnership

Tata Consultancy Services has agreed to acquire Porsche's IT consulting subsidiary MHP for €320 million, anchoring a five-year €1.25 billion automotive AI partnership. TCS plans to establish an AI Mobility Centre of Excellence to develop digital transformation software for the automotive sector.

Physical AI Startup General Intuition Raises Funding at $6B Valuation

General Intuition has secured a Series B funding round at a $6 billion valuation led by Valor and Point72. The startup is expanding its foundational robotics models for physical AI applications.

Infineon Acquires C2i Semiconductors to Expand AI Power Delivery Systems

Infineon Technologies has announced the acquisition of Bengaluru-based C2i Semiconductors to enhance its digital power management technology for AI data centres and accelerators. Infineon will establish an engineering center of excellence in India focused on AI power silicon.

Hugging Face Reportedly in Discussions for $13B Acquisition

Open-source AI platform Hugging Face is reportedly in preliminary acquisition discussions valuing the company at approximately $13 billion. If completed, the deal would represent one of the largest acquisitions in the AI developer ecosystem.

XPENG Robotics Secures Over $900M Series A at $6.3B Valuation

XPENG Robotics has raised more than $900 million in a Series A financing round valuing the company at over $6.3 billion. The round marks the largest single private financing for a Chinese embodied AI and humanoid robotics developer.

SpaceX and NVIDIA Plan Orbital AI Data Centres with Vera Rubin Hardware

SpaceX and NVIDIA have outlined plans to deploy space-optimized Vera Rubin NVL72 compute systems aboard Starmind AI satellites, targeting a first launch in late 2027 to extend hyperscale workloads into low Earth orbit.

AI product & feature launches

Thomson Reuters Develops Thomson-1 AI Model to Cut Frontier Lab Dependency

Thomson Reuters has introduced Thomson-1, a proprietary domain-specific AI model designed for legal workflows and fine-tuned on an open-source base. The initiative aims to reduce infrastructure costs and reliance on Anthropic's Claude.

Perplexity and NVIDIA Partner to Launch 'Portable Computer' Local AI Agent

Perplexity and NVIDIA have partnered to launch Portable Computer, a fully local AI agent platform running on RTX GPUs and DGX Spark hardware. The solution eliminates API token costs and ensures all prompt and document inference occurs locally on-device.

Google Cloud Launches Gemini Enterprise for Financial Services

Google Cloud has released Gemini Enterprise for Financial Services, a dedicated agentic platform featuring over 50 pre-built banking and capital markets skills. The offering includes native connectors to financial databases and automated research workflows.

Research with immediate practical relevance

Google Unveils AgentHands for Spatially Grounded Gestures in XR Agents

Google researchers have released AgentHands, an LLM-driven XR framework that automatically generates synchronized hand gestures for conversational agents. The system provides real-time spatially grounded visual guidance for users completing physical and procedural tasks.

Researchers Open-Source 4DAnyone Framework for Monocular Video 4D Reconstruction

A research collaboration including Zhejiang University and Ant Group has open-sourced 4DAnyone, a framework that generates high-fidelity 4D human reconstructions from standard single-camera video streams without studio capture rigs.