OpenAI just expanded Codex beyond coding with six role-specific plugins and an interactive Sites preview. With weekly users topping 5 million—and non-developers growing 3x faster than coders—the platform now bundles 62 apps across analytics, sales, and design, letting teams turn ideas into shareable, live web tools via URL.
JetBrains just released Mellum2, an open-source 12B Mixture-of-Experts model under the Apache 2.0 license. By activating only 2.5B parameters per token, it cuts latency and runs more than 2x faster than similarly sized open models. It is built to handle high-frequency plumbing like routing, RAG compression, and sub-agents across software engineering stacks.
TwelveLabs just launched Rodeo, its first creator-facing tool that lets editors search and assemble raw footage using plain text prompts. Powered by its Marengo 3.0 and Pegasus 1.5 models, it grasps the narrative context across hours of archive footage so creators can build rough cuts in minutes. Why it matters: It shifts AI video from simple clip generation to solving the real pain of organizing messy raw takes.
OpenAI highlighted Hiroki Tomiyasu, a Hokkaido farmer with zero engineering background who built an automated system for his 100-hectare farm using ChatGPT and Codex. He developed 8 AI workflows linking ESP32 hardware and satellite mapping to track crop health. Why it matters: It shows technical coding skills are no longer the bottleneck for turning domain knowledge into real-world software.
Factory just launched Factory Router, an automated routing tool that cuts enterprise coding token spend by 20-25%. It sends routine tasks to efficient models and saves frontier models for heavy reasoning, retaining 99% of Claude Opus 4.7's pass rate on Terminal-Bench 2 while routing across providers for 99.9%+ reliability. It is currently in private research preview.
NVIDIA just unveiled Cosmos 3 at GTC Taipei, an open-world foundation model built to give robots and autonomous vehicles physical common sense. Using a mixture-of-transformers setup, it combines visual reasoning with native action prediction, spitting out joint angles and trajectory points directly. It is now open-sourced on Hugging Face to help embodied AI predict before it moves.
MiniMax just released M3, an open-weight model packing a 1M token window, native multimodality, and desktop computer use into a single package. Its custom sparse attention architecture cuts decoding compute dramatically, yielding a 15x speedup over its predecessor and top marks on SWE-Bench Pro. It brings closed-source agentic capabilities straight to open-weight developers.
NVIDIA just unveiled the DGX Station for Windows at GTC Taipei, cramming the GB300 Grace Blackwell Ultra Desktop Superchip into a deskside workstation. It packs up to 748GB of coherent memory and 20 petaflops of FP4 compute, enabling local inference for 1-trillion-parameter models and hundreds of simultaneous agents. Systems from Dell, HP, and ASUS roll out in Q4.
NVIDIA just unveiled Vera, its first CPU purpose-built for AI agents and modern data centers, and it is already in full production. Built with 88 custom Olympus cores, the processor delivers 1.8x faster task completion than traditional x86 chips. AI leaders like OpenAI, Anthropic, and ByteDance are already lining up to adopt systems arriving this fall.
NVIDIA and Microsoft just unveiled RTX Spark, a new superchip purpose-built for personal AI agents on Windows PCs. Packing up to 128GB unified memory and 1 petaflop of AI compute, it can run 120-billion-parameter models locally. RTX Spark laptops and desktops from Asus, Dell, and Lenovo will arrive this fall.
Codex just rolled out update 26.527, bringing Computer Use to Windows so the assistant can view screens, click, and type directly across desktop apps. Users can now also remotely control Windows devices via ChatGPT on iOS or Android, or from a Mac running Codex. A refreshed profile section also adds token activity and usage stats.
Anthropic plans to release its Claude Mythos cybersecurity AI model to all customers in the coming weeks, according to a Reuters report on May 28. Previously confined to a defensive testing group including Apple and Microsoft, Mythos can piece together complex exploit chains that once required elite human researchers. Despite Cloudflare warning that the model's safety refusals remain inconsistent, rising market competition is pushing Anthropic to democratize defense.
Microsoft just rolled out a major redesign for Microsoft 365 Copilot and its in-app experience across Office. The standalone app now loads over 50% faster, cuts complex chat response times by 10%, and adds Work IQ to pull context directly from emails, meetings, and files. In apps like Word and Excel, Copilot acts more like an active editing agent on the canvas than a passive chat box.
Figma just introduced local codebase integration for Figma Make, bridging the gap between visual design and software engineering. Designers can visually tweak UI elements or leave contextual notes while an in-canvas AI agent updates actual code, creates branches, and opens pull requests. It is rolling out in limited beta for Mac desktop users.
Perplexity's Computer agent just arrived inside Microsoft Word, Excel, PowerPoint, and Outlook via native add-ins. The integration eliminates the copy-paste shuffle for the 40% of users exporting to these formats, enabling teams to build financial models, draft documents, and design decks directly from a side panel with cited data from SharePoint and the web.
Anthropic just rolled out dynamic workflows in Claude Code, allowing the model to write orchestration scripts that spin up hundreds of parallel subagents in a single session. The tool tackles massive engineering lifts like full-codebase audits, and developer Jarred Sumner even used it to port 750,000 lines of Bun to Rust in 11 days. Why it matters: it turns Claude from a pair programmer into an entire dev team, though it burns through tokens fast.
Anthropic just released Claude Opus 4.8, upgrading coding and agentic capabilities while keeping pricing unchanged. The model is four times less likely to let code flaws slip through unnoticed and brings new effort controls to claude.ai. Claude Code also gains dynamic workflows that can run hundreds of parallel subagents.
CapCut just revamped its web creative suite with Design Studio 2.0, ditching rigid templates for an infinite AI canvas and a collaborative design agent. Creators can now circle-to-edit with brush tools, split layers automatically, and branch into new visual directions, backed by 20 free generations every day.
Researchers just introduced AutoScientists, a decentralized multi-agent system designed for long-running scientific experimentation. Instead of relying on a rigid central planner, agents self-organize into teams, vet ideas before burning compute, and share failures to avoid redundant runs. It optimized GPT training 1.9x faster than Autoresearch while topping previous agents across biomedical benchmarks.
Runway just launched an MCP integration that lets users generate images and videos directly inside Claude, ChatGPT, Cursor, and Replit. Creators and developers can now tap into flagship models like Gen-4.5 and Kling without switching tabs or leaving their existing workflows. It essentially gives text and coding agents native access to top-tier creative media generation.
