Alibaba Cloud officially released the mobile app for its AI Agent product JVS Claw on March 13, 2026, now available on Apple App Store and major Android app stores. JVS Claw integrates AI assistant Clawbot with cloud environment ClawSpace, enabling users without any programming background to create their own AI agent "Lobster" in just three steps. The product features zero-code deployment, cloud-isolated execution with 6-core/12GB Linux environment, dual cloud/local mode switching, and offers 8,000 free Credits during beta testing.

StepFun (阶跃星辰) has officially launched StepClaw, a cloud-based AI assistant built on the wildly popular open-source OpenClaw agent framework. Addressing the complexity that has kept "lobster farming" out of reach for average users, StepClaw offers one-click deployment through the StepFun AI app, eliminating the need for command-line expertise. To celebrate the launch, StepFun is offering 50,000 free trial spots that include a comprehensive package: 50 million model tokens, cloud server resources, and storage — all free for one month. Powered by StepFun's own Step 3.5 Flash model — the current #1 most-called model on OpenClaw globally — StepClaw brings enterprise-grade agent capabilities to everyday users .

iFlytek has officially launched AstronClaw, a powerful AI agent platform designed with enterprise-grade security at its core. Featuring sandbox isolation technology for data protection, one-click cloud deployment, and seamless integration with enterprise messaging tools like WeCom, DingTalk, and Feishu, AstronClaw empowers users to assign tasks through natural conversation. The platform provides access to 10,000+ specialized skills and supports free switching between leading models including iFlytek Spark X2, MiniMax-M2.5, Kimi-K2.5, and GLM-5 — making it a versatile solution for industry analysis, document processing, and beyond

Midjourney has officially released version 8.5, headlined by the groundbreaking Consistent Identity (CID) feature that enables true character consistency across generations. Unlike previous --cref approaches that required reference images each time, CID builds a persistent internal model of a character that can be summoned across sessions, scenes, artistic styles, and even aspect ratios — maintaining facial features, body proportions, clothing, and personality essence without degradation over time.

GitHub has officially launched the beta of Copilot Auto-Fix, a groundbreaking autonomous feature that automatically detects and repairs code vulnerabilities, bugs, and security issues directly within the development workflow. Integrated into pull requests and the IDE, Auto-Fix leverages Copilot's agentic architecture to analyze code, generate fixes, and even review its own patches before submission—dramatically reducing developer toil and accelerating secure code delivery.

“Live Align” is being mentioned as a new Cursor capability that keeps AI-generated edits continuously aligned with your codebase in real time. However, Cursor has not officially announced or documented a feature named “Live Align” in its public changelog/blog/docs at the time of writing. What Cursor has shipped recently is a fast-moving set of agent and workflow upgrades—such as MCP Apps, team plugin marketplaces, and Bugbot Autofix—that point to the same underlying direction: tighter human↔AI collaboration and better end-to-end execution inside the IDE.

OpenAI has officially launched GPT‑5.3‑Codex on February 5, 2026, describing it as its most capable agentic coding model to date and 25% faster than the prior generation. The model combines the frontier coding performance of GPT‑5.2‑Codex with GPT‑5.2’s reasoning and professional knowledge, enabling longer-running workflows that involve research, tool use, and complex execution. OpenAI says GPT‑5.3‑Codex can iteratively build highly functional complex games and apps from scratch over the course of days, running autonomously over millions of tokens while users steer and interact without losing context.

Anthropic has launched Claude Opus 4.6 on February 5, 2026, positioning it as its most capable model to date for enterprise agents and “professional work” — including financial analysis that can surface insights that would otherwise take days of manual compilation. Opus 4.6 adds major agentic upgrades (notably Agent Teams in research preview) and introduces up to a 1M-token context window (preview) for processing large collections of company filings, regulatory documents, spreadsheets, and research notes in a single workflow. The model is available to Claude paid tiers (Pro/Max/Team/Enterprise) and via API, as well as through Amazon Bedrock and other cloud platforms.

Kuaishou has rolled out a major upgrade to its AI video generation stack with Kling 3.0 (including an Omni variant), expanding generation to up to 15 seconds per clip while pushing harder into native audio and more consistent multi-shot, multi-character storytelling. Third‑party platform announcements and product pages claim improvements in lifelike physics, character consistency, and multi-shot storyboards, with support for multilingual speech and dialects/accents in the audio-enabled Omni model. The update strengthens Kling’s positioning against competitors like OpenAI Sora and Google Veo by moving from “silent video clips” to a more complete “video + sound + dialogue” creative pipeline.

Kunlun Tech (昆仑万维) has officially launched Skywork Desktop (天工 Skywork 桌面版) globally on February 4, 2026, positioning it as a privacy-forward, on-device productivity agent that can read and organize large volumes of files directly on a user’s computer—without uploading documents to the cloud. Skywork Desktop emphasizes “content understanding over file formats,” supporting unified semantic processing across documents, spreadsheets, PPT, images, video, and more, while also enabling multi-task parallel execution for workflow automation and content generation

A new generation of third‑party “AI Copilot for Unreal Engine 5” tools is pushing UE5 workflows toward agentic, voice‑like world building: creators can describe a scene (e.g., “18th‑century London street, destructible”) and the system helps scaffold assets, Blueprint logic, UI, and scene population inside the editor. Public listings and forum changelogs show capabilities such as project‑wide indexing, mass scene population, UMG UI scaffolding, Blueprint refactoring, and MCP integration to connect external AI clients like Cursor and Claude Desktop. However, there is currently no verifiable Microsoft official announcement of a “Copilot for Unreal” product with fully automated real‑time modeling + ray tracing + physics setup.

OpenAI did not "cut off the macOS version of Codex" - on the contrary, it released its official Codex app for macOS on February 2, 2026, positioning Codex as a "command center for agents" that focuses on parallel agents, long task collaboration, and secure execution. If your news topic is "Cutting off macOS version of Codex, fully intelligent", a more accurate way to write it is: OpenAI is cutting down/weakening the "traditional monolithic Codex form" (such as single tool/single interface/single workflow), pushing Codex into a cross CLI/IDE/cloud/mobile agentic platform, and iteratively strengthening the capabilities of "long-range tasks, refactoring migration, and collaboration" with models such as GPT-5/5.2-Code.