Global AI Weekly
Issue number: 149 | Tuesday, May 12, 2026
Highlights
GitHub Copilot is moving to usage-based billing
Starting June 1, GitHub Copilot will shift to a usage-based billing model, where Copilot usage will count against GitHub AI Credits. This new structure means organizations and developers will need to manage their AI Credits for continued access to Copilot’s features. It’s a move designed to align usage with costs while maintaining flexibility for users.
github.blog
OpenAI ends its exclusive partnership with Microsoft
OpenAI has revised its agreement with Microsoft, allowing its models to now operate on Amazon Bedrock. This move broadens OpenAI's reach by enabling its technology to be integrated with Amazon's suite of cloud services. The change signals a shift in strategy aimed at making OpenAI's tools more widely available across different platforms.
arstechnica.comResearch
Beyond Semantic Similarity: Rethinking Retrieval for Agentic Search via Direct Corpus Interaction
This paper explores a fresh approach to retrieval for agentic search by moving beyond traditional semantic similarity methods. It focuses on enabling more direct interaction with the corpus, offering new perspectives on improving search effectiveness. The discussion highlights the potential of this framework to enhance how information is accessed and utilized.
huggingface.co
ARIS: Autonomous Research via Adversarial Multi-Agent Collaboration
This paper introduces ARIS, a novel system for autonomous research that leverages adversarial multi-agent collaboration. The approach focuses on using competing AI agents to explore, test, and refine ideas, fostering innovative solutions in research tasks. The framework showcases potential in accelerating scientific discovery and developing more effective AI collaborations.
huggingface.coVideo
Will The Real Autonomous Agent Please Stand Up
This keynote dives into the complexities of creating and managing AI agents, emphasizing that the toughest challenges aren't in building the agents themselves but in determining their usefulness, organizing data, crafting proper instructions, and maintaining oversight. It also highlights the emerging challenge of designing systems and workflows specifically for AI agents. Featuring real-world demos, it provides insights and practical strategies to rethink how to build and integrate AI agents effectively.
youtube.com
Introducing Agent Toolkit for AWS
Agent Toolkit for AWS equips AI coding agents with the essential resources to seamlessly interact with AWS services. By offering access to accurate, up-to-date documentation, best practices, and proven workflows, this tool ensures higher efficiency and reduces trial-and-error. The video showcases an impressive live demo, featuring the deployment of a Python Lambda behind an API Gateway in under three minutes, highlighting the practicality and speed of the toolkit. It also covers core components, agent skill adaptations, and auditability using CloudTrail while providing rollout tips for teams.
youtube.comArticles
Gemini API File Search is now multimodal: build efficient, verifiable RAG
The Gemini API File Search tool now supports multimodal capabilities, allowing developers to build more efficient and verifiable retrieval-augmented generation (RAG) systems. These updates simplify the process of retrieving and managing diverse file types, ensuring seamless integration for advanced applications. This improvement enhances user experiences by making file retrieval faster, more accurate, and easier to implement.
blog.google
Focus areas for The Anthropic Institute
The Anthropic Institute (TAI) explores the real-world impact of AI by harnessing insights from a cutting-edge lab environment. They outline key questions shaping their research agenda, aiming to better understand AI's influence and share valuable findings with the public. Through this approach, TAI seeks to contribute to informed discussions about the role of AI in society.
anthropic.comUpcoming Events
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
microsoft/agent-lightning: The absolute trainer to light up AI agents.
Microsoft's Agent Lightning is a tool designed to enhance AI agents' performance. By participating in its development, contributors can collaborate on building a powerful trainer for AI systems. Check out this exciting project on GitHub to explore its features and contribute to its growth.
github.com
Claude Cowork Now Runs Any LLM. Test It Free.
Learn how to set up Cowork on 3P to integrate seamlessly with OpenAI, Gemma, Kimi K2, or local models using platforms like OpenRouter, Bedrock, Vertex, or Foundry. This guide walks you through the process step-by-step, highlighting its compatibility with free-tier options. A straightforward way to enhance your workflow with the LLM of your choice.
productcompass.pmPodcast
Eye on AI
Eye on AI is a podcast hosted by Craig S. Smith, a seasoned journalist from The New York Times, exploring the latest breakthroughs in artificial intelligence. Each episode features conversations with key figures in the AI field, offering insight into how these advancements fit into the bigger picture. With AI poised to transform the world, this podcast is a must-listen for staying informed.
open.spotify.com