/AI Weekly/Issue 113
Microsoft AI launches its first in-house models
Published Tuesday, September 2, 2025
Microsoft has introduced its first in-house AI models, signaling a major step in its push to compete directly in the AI space alongside OpenAI, the company it has heavily invested in. These new models reflect Microsoft's growing commitment to developing its own advanced AI technologies while expanding its influence in the rapidly evolving artificial intelligence landscape.
A study from MIT’s Project NANDA highlights a concerning trend: 95% of enterprise AI investments fail to deliver any return. The research points to fundamental learning limitations in artificial intelligence systems, which prevent organizations from achieving meaningful outcomes. This growing gap underscores the challenges enterprises face in effectively implementing and utilizing AI solutions.
The page explores InternVL3.5, a cutting-edge open-source multimodal model designed to enhance versatility, reasoning abilities, and efficiency in AI systems. It highlights improvements in handling diverse tasks, delivering smarter performance while maintaining cost-effectiveness. Dive into the details of how this model pushes the boundaries of multimodal AI capabilities.
This paper focuses on LiveMCP-101, a method designed to stress test and diagnose MCP-enabled agents using challenging queries. It explores how these agents perform under difficult conditions and analyzes their ability to handle complex tasks effectively. The discussion offers insights into improving the robustness and reliability of advanced AI systems.
This demo highlights how GitHub Copilot simplifies development by acting as more than just a coding assistant. From setting up environments and prioritizing tasks to generating code and debugging issues, Copilot streamlines the workflow, even tackling tests and dependencies. Its agent mode goes a step further, executing tasks like writing commits, creating pull requests, and reviewing its own work. It's designed to keep developers focused on building impactful software with minimal distractions.
A new standard is gaining traction in the open-source scene with AGENTS.md, a simple and collaborative format aimed at supporting AI coding agents in software development. With over 20,000 developers already on board, this initiative is driving consistency and ease in AI-assisted programming. The format is shaping up to be a key tool for fostering innovation and efficiency in the tech community.
Generative AI is now playing a role in assisting decisions on which Linux kernel patches should be backported to stable releases, including crucial LTS branches. This technology is being leveraged to streamline and enhance the process of maintaining stability and security across various kernel versions. By using AI, developers can improve efficiency and make more informed choices in patch management.
After an exciting set of AgentCon events before the summer holidays, AgentCon is happening in more locations during the second half of the year. The AI Agents World Tour is a series of one-day conferences for developers focused on autonomous AI agents. Spanning cities like San Francisco and Singapore, these events gather top engineers, researchers, and creators shaping the future of AI technologies. It’s an opportunity to connect, learn, and innovate in the fast-evolving world of autonomous AI. Check all events and register for an event near you!
Are you building with Azure AI Foundry? In today’s fast-paced dev world, you don’t want to miss this. The Azure AI Foundry Developer Community is your one-stop hub for: a vibrant forum for Q&A and code, with DevBlogs featuring the latest updates, plus real-time Discord chats, events, and learning. Connect with global peers and find everything you need in one vibrant space. Let’s build the future of AI - together.
Discover how to create a multi-agent system that efficiently plans, designs, and builds a comprehensive knowledge graph. This process integrates advanced collaboration techniques between agents to streamline the construction and organization of interconnected data. Perfect for anyone looking to enhance their understanding of AI-driven knowledge management.
Machine Learning Street Talk (MLST) brings insightful conversations with leading experts in AI, exploring critical topics in artificial intelligence, cognitive science, neuroscience, and the philosophy of mind. Hosted by Tim Scarfe, Ph.D., with regular contributions from MIT's Dr. Keith Duggar, the show emphasizes intellectual diversity and rigorous analysis while cutting through the hype surrounding AI. It's a must-listen for anyone interested in thoughtful discussions on the big ideas shaping the field.
The AI links worth your time, curated by the community and always free.
