OpenAI reportedly inks a $300B, multi‑year Oracle deal as Microsoft brings Claude to Office and Nvidia unveils Rubin CPX for long‑context inference.
Oracle may have secured a $300B, ~5-year OpenAI compute deal starting in 2027—one of the largest cloud contracts. It underscores OpenAI’s move beyond Microsoft to Oracle and Google and ties to the $500B Stargate Project.
Microsoft will add Anthropic AI to Office apps, moving beyond exclusive reliance on OpenAI as tensions rise. Leaders see Claude Sonnet 4 excelling in tasks like PowerPoint design, while Microsoft develops its own models and OpenAI pursues chips and a LinkedIn rival.
xAI cut 500 employees, a third of its data-annotation team, in a pivot from general AI tutors to specialist ones. The company plans to expand its specialist tutor team tenfold, hiring across fields like STEM, finance, medicine, and safety.
Meta’s push to recruit top AI talent with huge pay and resources has triggered defections, pay disputes, and tensions with existing staff. Its secretive TBD Lab and hiring freeze add pressure, while rivals like OpenAI and new startups continue poaching Meta’s recruits.
Nvidia unveiled the Rubin CPX GPU, built for context windows over 1M tokens and optimized for long-context tasks like video generation and coding. Part of its Rubin series, the chip arrives in late 2026 as Nvidia’s data center revenue hit $41.1B last quarter.
Microsoft and OpenAI struck a tentative deal to extend their partnership, easing tensions as OpenAI seeks approval for a $100B for-profit restructuring. Both sides will hold ~30% stakes, though rivals and regulators are challenging the plan, and Microsoft is diversifying with Anthropic and xAI.
Codex just got faster, more reliable, and better at real-time collaboration and tackling tasks independently anywhere you develop—whether via the terminal, IDE, web, or even your phone.
Seedream 4.0 adds batch input/output, prompt-based editing, versatile styles, and knowledge-driven generation. Benchmarks show top performance in prompt adherence, aesthetics, text rendering, and image editing.
The MCP Registry preview launched as an open-source hub for MCP servers, standardizing distribution and discovery while supporting public and private sub-registries. Built with broad community collaboration, it offers a moderated, extensible foundation for the MCP ecosystem.
The Qwen3-Next architecture uses a sparse Mixture-of-Experts (MoE) for extreme efficiency. This new model delivers top-tier performance with a fraction of the training cost and achieves over 10x faster inference speeds on complex, long-context tasks.
A MCP proxy that lets you dynamically aggregate MCP servers into a unified MCP server, and apply middlewares. MetaMCP itself is a MCP server so it can be easily plugged into ANY MCP clients.
Cursor improved its Tab model—which predicts developers’ next coding actions—by using online reinforcement learning instead of static datasets or filters. The new model makes 21% fewer suggestions but achieves a 28% higher accept rate, reducing noise and improving coding flow.
In developer mode, developers can create connectors and use them in chat for write actions (not just search/fetch).
Early screenshots show a new "background thinking" toggle that can be enabled.
Rolling out to Team & Enterprise plans while also introducing incognito chats for all users.
Using the web fetch tool, Claude will fetch and analyze content from any webpage URL—no additional infrastructure needed.
What coding agents have been doing for software engineering is soon expanding to all knowledge work - this is just the beginning.
The authors argue that small language models (SLMs) are more efficient, cost-effective, and better suited than LLMs for most repetitive, specialized tasks in agentic AI systems. They propose shifting from LLMs to SLMs, using heterogeneous systems when general conversation is needed, to reduce costs and optimize AI resource use.
Dexter Horthy, founder of Human Layer, shares what his team has learned about scaling coding agents in real-world software projects.
A curated weekly roundup of AI news, tools, and breakthroughs to stay ahead of the curve in a fast-moving industry.
Curated by devs, for devs. Helping 400+ developers stay ahead of the AI curve.