AI Pulse π¨
Here is a bite-sized roundup of what happened in AI last week
π SpaceX acquired xAI in a $1.25 trillion deal, the largest M&A transaction in history. The merger combines Musk's AI and space ambitions ahead of a potential SpaceX IPO.
π§ Anthropic released Claude Opus 4.6 with 1M token context, context compaction, and adaptive thinking. It leads on Terminal-Bench 2.0, Humanity's Last Exam, and GDPval-AA benchmarks.
π» OpenAI released GPT-5.3-Codex, a faster and more capable coding model that was instrumental in creating itself.
π₯ Claude Code launched agent teams in experimental preview. Coordinate multiple Claude Code instances working together, with a team lead assigning tasks and teammates communicating directly with each other.
π’ OpenAI launched Frontier, an enterprise platform for building, deploying, and managing AI agents. Many are calling it a "digital employee" platform.
π₯οΈ OpenAI launched the Codex app for macOS, a command center for managing multiple coding agents working in parallel.
π GitHub added Claude and Codex as AI coding agents. Pick your agent directly inside GitHub, GitHub Mobile, and VS Code.
π Anthropic launched Claude in PowerPoint in research preview. Claude reads your layouts, fonts, and slide masters to stay on brand when generating presentations.
π Apple's Xcode now supports the Claude Agent SDK, making it easier for iOS and macOS developers to integrate AI agents into their apps.
π Claude Code added the /insights command. It reads your message history from the past month and gives suggestions on how to improve your workflow.
π Claude Code now supports session sharing. Share your full conversation with team members or anyone with a link. Available on web, desktop, and mobile.
β‘ Claude Code added fast mode for quicker responses when you do not need deep reasoning.
π Vercel released Agent Browser v0.9 with iOS support. Run browser automation for AI agents on mobile devices.
π€ ByteDance released UI-TARS, an open source multimodal AI agent stack connecting cutting-edge models and agent infrastructure.
π¬ Anthropic published research on building a C compiler with a team of parallel Claudes, showing how agent teams can tackle complex engineering tasks.
π§ͺ Google launched Google CC as an experiment at Google Labs.
π§ Claude now has interactive work tools. Draft Slack messages, visualize ideas as Figma diagrams, or build Asana timelines directly in Claude.
βοΈ Cursor improved codebase indexing using Merkle trees, cutting time-to-first-query from hours to seconds on large codebases.
π€ Moltbook is seeing humans infiltrate the AI social network, posing as bots among the agent conversations.