Global AI Weekly - Issue 152 - Global AI Community

In this issue

Highlights

2 stories

cybernews.com

Build 2026 Could Showcase Microsoft’s Next AI Leap

Microsoft is reportedly preparing to unveil a suite of homegrown AI models at Build 2026, including a new coding model designed to strengthen GitHub Copilot. The move signals a broader strategy to reduce reliance on external AI providers such as OpenAI, Anthropic, and Google. Alongside coding capabilities, Microsoft is said to be developing specialized models for reasoning, transcription, speech, and image tasks. If true, Build could mark a major shift as Microsoft doubles down on controlling more of its own AI stack.

anthropic.com

Introducing Claude Opus 4.8

Claude Opus 4.8 is here, bringing enhanced performance to the Opus class of models. It’s designed to excel in coding, agentic tasks, and professional applications, all while maintaining reliability for sustained work. This upgrade ensures both power and consistency for a wide range of demanding tasks.

In this issue

Research

2 stories

arxiv.org

SkillOpt: Executive Strategy for Self-Evolving Agent Skills

This paper introduces SkillOpt, a framework designed to enhance the development of self-evolving agent skills in challenging environments. It focuses on optimizing the selection and application of executive strategies that allow agents to adapt and improve over time. The approach is tested across various scenarios, demonstrating its effectiveness in creating more robust and capable agents.

techcrunch.com

RSI is the new AGI, and it's just as hard to pin down

A wave of AI labs is shifting focus to recursive self-improvement (RSI), an ambitious approach where AI systems iteratively improve themselves without direct human input. While the concept holds transformative potential, achieving it has been fraught with technical and philosophical challenges, similar to the hurdles faced in the pursuit of artificial general intelligence (AGI). Researchers are exploring innovative methods, but the path to meaningful progress in RSI remains uncertain and complex.

In this issue

Video

2 stories

youtube.com

Trying out the GitHub Copilot App

Exploring the GitHub Copilot App has been a game changer for managing daily tasks more efficiently. It’s helping reduce time spent at the terminal by streamlining workflows. The experience highlights just how much easier tackling projects can be with the app’s tools at your fingertips.

youtube.com

Multi AI Agent Systems: When One AI Brain Isn’t Enough

Bri Kopecki discusses the concept of AI hallucinations and highlights how multi-agent systems can enhance trust and verification in AI technologies. By leveraging multiple AI agents, practitioners can create safer and more reliable systems, especially in critical, high-stakes scenarios. The focus is on advancing AI's dependability to meet the challenges of real-world applications.

In this issue

Articles

2 stories

devblogs.microsoft.com

What's new in Microsoft Foundry | May 2026

This month’s Microsoft Foundry updates bring exciting enhancements, including new models, advanced agent skills, and expanded toolboxes to elevate your projects. Learn about the Foundry Local vision, updated evaluations, fresh SDK releases, and key platform improvements designed to streamline workflows and boost efficiency. Stay informed on all the latest developments shaping the future of innovation.

developer.microsoft.com

Improve your agentic developer tools by grounding in Microsoft Learn

Discover how grounding your developer tools in Microsoft Learn can enhance workflows across terminals, IDEs, background agents, and custom assistants. By aligning these tools to a shared and reliable knowledge base, you can ensure consistency and efficiency in your development process. Explore the benefits of integrating trusted resources into your toolkit for a streamlined and effective approach.

In this issue

Upcoming Events

1 story

globalai.community

Build //localhost

Missed Microsoft Build? Join a Build //localhost event and experience the highlights right in your local developer community. In this series of events, partially organized by the Global AI Community, these events bring together builders, engineers, and AI enthusiasts to explore the latest announcements, watch curated Build content, attend live technical sessions, and connect with peers. With events taking place across dozens of cities worldwide, Build //localhost is the perfect way to turn Build inspiration into real conversations, practical learning, and valuable local connections.

In this issue

Code

3 stories

platform.claude.com

Build an orchestration mode with Claude

Explore how to create an orchestration mode that allows session-level consent for multi-agent communication. This mode can be toggled on or off during conversations using system messages, enabling smoother collaboration between agents. Learn the steps to implement this functionality effectively for improved conversational workflows.

langchain.com

How Lyft Built a Self-Serve AI Agent Platform with LangGraph and LangSmith

Lyft leveraged LangGraph and LangSmith to create a self-serve AI agent platform designed to enhance customer support. This innovative approach significantly streamlined the process, reducing the time needed to develop AI agents from months to just weeks. By optimizing efficiency, Lyft has improved its ability to deliver quicker and more effective support solutions.

claude.com

Using LLMs to secure source code

In this issue

Podcast

1 story

open.spotify.com

The Cognitive Revolution

"The Cognitive Revolution" is a biweekly podcast hosted by Nathan Labenz and Erik Torenberg. They talk to leading AI builders and researchers about the breakthroughs in artificial intelligence and the transformative changes it’s expected to bring. Part of the Turpentine podcast network, the show offers insights into cutting-edge innovation and its future impact.