Global AI Weekly
Issue number: 140 | Tuesday, March 10, 2026
Highlights
Cursor Goes to War for AI Coding Dominance
Cursor, one of the fastest-growing AI coding tools, now faces a new kind of competition from powerful coding agents built directly into frontier AI models. As systems like Claude Code become capable of planning, writing, and managing software projects on their own, the role of the traditional code editor is being questioned. Cursor is racing to expand beyond assistance into a full AI development environment, setting up a high-stakes battle over the future interface of software creation.
forbes.com
OpenAI building GitHub alternative for developer toolchains
OpenAI is working on creating a new code-hosting platform designed to compete with Microsoft's GitHub. This initiative aims to offer developers an alternative option for hosting and managing their code, potentially becoming a key component in modern developer toolchains. Details about the platform remain limited, but it signifies OpenAI's interest in expanding its presence in the developer ecosystem.
developer-tech.comResearch
Evaluating Theory of Mind and Internal Beliefs in LLM-Based Multi-Agent Systems
The paper focuses on exploring the capabilities of Large Language Model (LLM)-based multi-agent systems in understanding Theory of Mind (ToM) and internal beliefs. It investigates how well these systems can represent and reason about the mental states of others, which is crucial for effective communication and collaboration. The study presents assessments and findings that contribute to understanding the strengths and limitations of LLMs in handling complex social and cognitive tasks.
arxiv.orgVideo
OWASP's Top 10 Ways to Attack LLMs: AI Vulnerabilities Exposed
Jeff Crume discusses the OWASP Top 10 vulnerabilities for large language models (LLMs), highlighting risks such as prompt injection and data leaks. He provides practical tips, including using firewalls and access controls, to protect AI systems from potential attacks. This overview emphasizes the importance of securing AI models against evolving threats and vulnerabilities.
youtube.com
LangChain & LangSmith Skills: Teach Your AI to Build Agents
Discover how to enhance your coding agent's expertise in agent development using LangChain and LangSmith skills. This video showcases Claude Code's journey through the entire agent development process, from building an agent to tracing and evaluating it effectively. Gain practical insights on leveraging these tools for creating smarter, more efficient AI systems.
youtube.comArticles
Copilot usage metrics now includes user-level GitHub Copilot CLI activity
GitHub has introduced user-level telemetry for GitHub Copilot CLI, following the recent release of enterprise-level CLI telemetry. This update allows users to track CLI-specific activity and usage totals, providing better insights into individual usage patterns. The feature aims to enhance understanding and optimization of Copilot's functionality at a more granular level.
github.blog
Databricks Unveils KARL, a RAG Agent Built for Enterprise Chaos
Databricks has introduced KARL, a new AI agent designed to tackle one of enterprise AI’s toughest problems: messy, fragmented internal data. Trained with reinforcement learning, the system can handle multiple enterprise search behaviors such as synthesizing reports across documents, navigating long technical files, and reconstructing insights from scattered records. Databricks says KARL rivals leading models while delivering lower cost and latency, positioning it as a step toward more general-purpose, enterprise-ready retrieval agents.
venturebeat.com
I used Nano Banana 2 to make perfect sketchnotes: 5 lessons learned (and how I broke Gemini)
The video explores the experience of using Nano Banana 2 for creating sketchnotes, highlighting both the successes and challenges faced along the way. It shares five practical tips to improve results and speed up the process, while also explaining how Gemini was unintentionally broken during the testing. The insights are a mix of intriguing outcomes, surprising fails, and helpful lessons learned.
zdnet.comUpcoming Events
Experience AI’s defining moment and join GTC online
The next generation of AI is here, and you’re invited to be part of it by joining GTC virtually from March 16–19. Get online access to the GTC Live and keynote livestreams, as well as live-broadcasted sessions, exclusive interviews, and demos. You can also participate in expert training, explore research posters, attend sponsored sessions, and view all GTC’s virtual and on-demand content. Register now for a truly unique experience, including: - The livestreamed keynote from NVIDIA CEO Jensen Huang - 700+ virtual and on-demand sessions, including panels on AI agents, agentic architectures, and robotics. Time is running out to grab your spot. Register today and be part of AI’s most transformative event.
nvidia.com
AgentCamp - Coming to a City Near You
AgentCamp continues to grow as a global series of hands-on gatherings dedicated to building and experimenting with AI agents. These community-driven events bring developers, founders, and AI enthusiasts together for practical sessions, collaborative building, and open exchange of ideas. Hosted in cities around the world, AgentCamp focuses on real-world experimentation, giving participants the space to prototype agent workflows, explore emerging tools, and learn directly from peers working at the edge of autonomous AI. Join the community to build, share, and help advance what AI agents can do in practice.
globalai.communityCode
Figma MCP server can now generate design layers from VS Code
GitHub Copilot now supports connecting to the Figma MCP server, enabling users to integrate design and code seamlessly. This update allows developers to pull design context directly into their code and send rendered UI back to Figma as editable design frames. It enhances collaboration between design and development, streamlining the workflow between Figma and VS Code.
github.blog
Release v1.0 of the official MCP C# SDK
The official release of the MCP C# SDK v1.0 brings exciting updates, including improved authorization capabilities, more detailed metadata options, and better support for tool calling. It also introduces enhancements to handle long-running requests efficiently. These updates aim to provide developers with a more streamlined and powerful experience when working with the MCP platform using C#.
devblogs.microsoft.com
Migrate your Semantic Kernel and AutoGen projects to Microsoft Agent Framework Release Candidate
Microsoft Agent Framework has achieved Release Candidate status for .NET and Python, signaling that its API is stable and v1.0 features are finalized. This is the ideal time to transition your Semantic Kernel and AutoGen projects to the new framework and provide feedback ahead of the General Availability release. Take advantage of this opportunity to explore the framework and ensure a smooth migration.
devblogs.microsoft.comPodcast
How I AI
How I AI, hosted by Claire Vo, is a podcast focused on practical ways to use AI to enhance work and life. Each 30-minute episode features guests sharing real-life examples, tips, tricks, and efficient workflows involving AI, often including live screen sharing. It's designed to help listeners understand and utilize AI effectively, providing actionable insights for navigating this rapidly evolving technology.
open.spotify.com