AI Dev News Digest - Nov 22nd, 2025

This week felt like the industry collectively decided to ship everything at once. Google dropped Gemini 3 with a coding platform called Antigravity, Microsoft used Ignite to roll out an entire suite of agent management tools, and then the headline everyone was talking about: Microsoft and NVIDIA teaming up to pour $15 billion into Anthropic. That's a lot of money moving around, and it says something about where the big players think this is all going. Claude is now the only major model available on all three major clouds, which is a significant shift.

On the developer tools front, it was mostly incremental but useful stuff. GitHub Copilot now lets enterprise teams bring their own API keys from Anthropic, OpenAI, or xAI. They also added linter integration to code review, which is the kind of practical feature that actually makes a difference in day-to-day work. xAI shipped Grok 4.1 with a 2M context window and some new agentic tooling, though the release got a bit overshadowed by some strange behavior in the consumer chatbot. And if you're curious about funding, Suno raised $250M for AI music, Function Health hit nearly $2.5B valuation, and Cursor (from last week) continues to be valued like it's inventing fire.

Model Releases & Platform Updates

• Google launches Gemini 3 with Antigravity coding platform. Google's newest model brings "generative interfaces" that create custom visual layouts on the fly. The new Gemini Agent handles multi-step tasks like organizing your inbox or booking car rentals. (Google Blog)

• xAI releases Grok 4.1 to all users. The update focuses on emotional intelligence and creative writing, with hallucinations reduced by roughly 3x. It's now available on grok.com, X, and mobile apps. (xAI)

• xAI launches Grok 4.1 Fast and Agent Tools API. A new tool-calling model with a 2M token context window. The Agent Tools API gives developers access to web search, code execution, and MCP integration—all handled server-side. (xAI)

• Gemini 3 Pro comes to Jules coding agent. Google's autonomous coding agent gets the new Gemini 3 model, improving agentic workflows and multi-step task reliability. (Google Developers Blog)

Microsoft Ignite 2025

• Microsoft unveils Agent 365 for managing enterprise AI agents. A new "control plane" that lets orgs deploy, monitor, and govern AI agents across their infrastructure with unified identity management through Entra ID. (Microsoft Blog)

• Work IQ, Fabric IQ, and Foundry IQ announced. Three new "IQ" products to help AI agents understand your work context, data semantics, and document retrieval needs. (Microsoft 365 Blog)

• Windows gets agent connectors and workspace preview. Native MCP support in Windows 11, plus a dedicated agent workspace where AI can interact with software in isolated environments. (Windows Blog)

• GitHub Copilot CLI adds GPT-5.1 and Gemini 3 Pro support. New models available alongside improved codebase search with bundled ripgrep. (GitHub Changelog)

Strategic Partnerships & Funding

• Microsoft, NVIDIA, and Anthropic announce $15B partnership. Microsoft invests $5B and NVIDIA invests $10B in Anthropic. In return, Anthropic commits to $30B in Azure compute. Claude becomes the only frontier model on all three major cloud providers. (Anthropic)

• Claude models launch in Microsoft Foundry. Sonnet 4.5, Opus 4.1, and Haiku 4.5 are now available in public preview on Azure, with integration into Microsoft 365 Copilot. (Anthropic)

• Sakana AI raises $135M Series B at $2.65B valuation. The Japanese foundation model lab expands its research into Japan-focused AI models. (SakanaAI)

Developer Tools

• GitHub Copilot Enterprise BYOK launches in public preview. Teams can now connect API keys from Anthropic, OpenAI, Microsoft Foundry, or xAI to use their own models in Copilot Chat. (GitHub Changelog)

• Copilot code review adds linter integration. ESLint, PMD, and CodeQL results now surface directly in Copilot code reviews. (GitHub Changelog)

• GitHub improves Next Edit Suggestions with custom model training. New data pipelines and reinforcement learning make the edit prediction faster and more accurate. (GitHub Blog)

Funding & Research

• Function Health raises $298M Series B. AI-powered health intelligence platform hits ~$2.5B valuation. (Tech Startups)

• Redrob raises $10M Series A for LLM infrastructure. Building out large-language-model platforms in Korea. (MorningStar)

• Flexion raises $50M for humanoid robot cognition. Working on the "brain" for humanoid and human-capable robots. (Flexion.ai)