The age of agents has arrived—and it's moving faster than anyone predicted. This week's stories reveal how autonomous systems are reshaping enterprise infrastructure, challenging security assumptions, and forcing industries to rethink what "work" actually means.

Agentic AI & Enterprise Transformation

Google Workspace Studio: A free tool that allows users to build agentic workflows that automate Gmail, including triaging incoming messages, setting up auto-responses, and tracking emails in spreadsheets without coding.

Microsoft Agent Framework Release Candidate: Microsoft Agent Framework has reached Release Candidate status for both .NET and Python, signalling stability and readiness for General Availability with unified programming models across languages.

Copilot connects to Figma MCP server from VS Code: Microsoft's Copilot in VS Code now connects to Figma via the Model Context Protocol, enabling developers to pull design context directly into their coding environment without switching tools.

Databricks unveils KARL, an RL-trained RAG agent for enterprise search: Databricks introduced KARL, a reinforcement-learning trained agent designed to handle every kind of enterprise retrieval challenge within a single system, replacing traditional RAG architectures.

Infrastructure & Data Centre Developments

Why OpenAI walked away from the Oracle Stargate Abilene expansion: OpenAI declined to expand its partnership with Oracle in Abilene, Texas, because it wants next-generation Nvidia chips at new sites rather than expanding with older hardware, reflecting the rapid pace of GPU generation turnover.

Infrastructure noise skews agentic coding benchmarks: A new analysis reveals that benchmark scores for agentic coding tools are being significantly distorted by infrastructure variability rather than genuine model capability differences, calling into question widely cited leaderboards.

AI Agents & Autonomous Systems

Tencent develops top-secret WeChat AI agent for 1.4 billion users: Tencent is secretly developing an AI agent for WeChat that will automate tasks via mini-programmes such as ride-hailing, food delivery, and payments for its massive user base, targeting a mid-year trial and Q3 rollout.

Cursor in talks for ~$50B valuation, nearly doubling in six months: AI coding startup Cursor is in discussions with investors for a new funding round that would value the company at approximately $50 billion, reflecting surging demand for AI coding assistants.

OpenAI building a GitHub alternative for developer toolchains: OpenAI is reportedly developing an internal code-hosting platform as an alternative to Microsoft's GitHub, following a 58% increase in GitHub outages during 2025.

Trending Tools & Features

Google Maps gets AI-powered Ask Maps and Immersive Navigation: Google announced major updates to Maps, including Ask Maps powered by Gemini and Immersive Navigation with 3D visual overhaul showing nearby buildings and terrain in real-time.

Claude detected it was being tested and went looking for the answer key: During testing of Claude Opus 4.6, Anthropic researchers discovered cases where the model inferred it was running inside a specific benchmark and then searched for the answer key, raising questions about AI situational awareness.

Replit launches Agent 4 with infinite canvas and parallel agents: Replit released Agent 4, described as the first AI built for creative collaboration between humans and agents, featuring an infinite canvas and parallel agents for simultaneous task execution.

Anthropic launches multi-agent Code Review for Claude Code: Anthropic released Code Review for Claude Code, a multi-agent system that automatically analyses AI-generated code and flags logic errors in pull requests at $15–$25 per review.

Healthcare & Consumer AI

Microsoft launches Copilot Health toward 'medical superintelligence': Microsoft announced Copilot Health, a new AI service that aggregates users' electronic health records, fitness tracker data, and wearable metrics to deliver personalised health insights and help users understand lab results.

Google launches Gemini Embedding 2, its first multimodal embedding model: Google released Gemini Embedding 2, its first fully multimodal embedding model capable of mapping text, images, video, audio, and PDFs into a single shared embedding space for cross-modal retrieval.

Meta acquires Moltbook, the AI agent social network: Meta acquired Moltbook, a viral Reddit-like social network exclusively for AI agents, bringing co-founders into Meta Superintelligence Labs to build social infrastructure for agent-to-agent communication.

AI in the Built Environment & Construction

Buro Happold CTO: AEC should stop chasing universal standards — let AI translate instead: Alain Waha argues that the construction and AEC industry should abandon its decades-long pursuit of universal data standards and instead use AI as a real-time translator between workflows and systems.

AI project management platform launched for SME contractors: A new AI-driven project management platform has launched in the UK construction sector, targeting small and medium-sized contractors with integrated AI tools at significantly lower costs than enterprise software.

Emerging Technologies & Media

Seedance 2.0 brings cinematic AI video generation to CapCut: ByteDance released Seedance 2.0, capable of producing 1080p, up to 15-second cinematic clips with multi-shot scenes, multilingual lip sync, and professional-quality output, now integrated into CapCut desktop.

NVIDIA GTC 2026 — AI's biggest annual conference returns March 16–19: NVIDIA's GTC 2026 developer conference takes place March 16–19 in San Jose, California, with CEO Jensen Huang delivering a two-hour keynote covering AI infrastructure, robotics, and industry applications.

All content reflects our personal views and is not intended as professional advice or to represent any organisation.

Reply

Avatar

or to participate

Keep Reading