Microsoft Build 2025 Satya Nadella Opening Keynote (AI column)

Last week, Satya Nadella, CEO of Microsoft, opener conference MS Build An interesting keynote report, which included guests in the form of Sam Altman, Elon Musk and others. Here are the key points that were in his speech.

1. GitHub Copilot as an Autonomous Agent GitHub Copilot has evolved into a full-fledged agent assistant capable of independently performing development tasks: from adding new functions and fixing errors to refactoring and generating documentation. It uses the Model Context Protocol. (MCP) to interact with external data and tools, and supports multimodal input data, including screenshots and layouts. 2. Azure AI Foundry: More 1900 models and multi-agent orchestration Azure AI Foundry Platform Expanded to Support More 1900 Models including integration with Grok 3 xAI. It provides tools for creating and managing multi-agent systems, allowing developers to build complex AI applications with a high degree of customization. 3. NLWeb: An Open Protocol for Agent Web NLWeb is an open protocol that allows you to implement AI interfaces on websites with minimal code. It supports connectivity to various LLMs and uses MCP to interact with site content, paving the way for the creation of an “agent internet.” 4. Windows AI Foundry: Local AI Development on Windows The new Windows AI Foundry platform provides tools for local development, optimization and deployment of AI models on Windows and macOS. It supports MCP integration and provides access to system functions such as file system and WSL. 5. Copilot Studio and Agent Setup Copilot Studio now allows you to create and configure AI agents using internal company data. This includes adjusting agents’ behavior, interacting with other systems, and supporting multi-agent scenarios. 6. Integration of xAI Grok models into Azure Microsoft Announces Partnership with xAI to Provide Access to Grok Models 3 grok 3 Mini via Azure AI Foundry. This expands LLM choice for developers and reduces reliance on OpenAI. It was on this news that Elon connected:) 7. Open source code for Windows Subsystem for Linux (WSL) Microsoft has opened source code for key WSL components, including wsl.exe and wslg.exe, under an MIT license. This is a step towards greater transparency and community involvement in the development of the tool. 8. MCP as an Interaction Standard MCP is becoming a key standard for the interaction of AI models with external data and tools. It provides a unified way to connect models to various data sources and systems, including Windows and Azure. 9. Microsoft Discovery: A Research Platform The Microsoft Discovery platform uses specialized AI agents to manage scientific research – from generating hypotheses to analyzing results. It is designed to accelerate scientific discoveries and increase the efficiency of research. Google and OpenAI already have such platforms. 10. Focus on multi-agent systems and their orchestration Microsoft focuses on developing multi-agent systems where agents can interact with each other to perform complex tasks. Copilot Studio and Azure AI Agents Service provide tools for creating such systems with the ability to delegate tasks between agents.

It is interesting that after this we were pleased with Google with its presentation of achievements. ~~economy~~ on Google I/O 2025

#AI #Software #Engineering