ByteDance just open-sourced Lark CLI under the MIT license, specifically designed for human developers and AI agents. It spans 18 enterprise domains—including Docs, Base, and Calendar—with over 200 curated commands. Why it matters: popular AI coding tools like Claude Code can now hook into and operate your entire Lark workspace with zero setup.
Google DeepMind just showed off Gemini 3.1 Flash-Lite generating entire websites in real time. Instead of serving pre-built pages, the demo browser builds each screen dynamically as you click and search. It is an impressive showcase of how fast and responsive lightweight models have become.
Anthropic just introduced Auto Mode for Claude Code, letting the AI agent execute tool calls autonomously without asking for approval on every step. A background classifier screens each action before it runs, blocking destructive moves like mass file deletion or data exfiltration while greenlighting routine tasks. It is now available in research preview for Team plan users.
ByteDance just open-sourced DeerFlow 2.0, an MIT-licensed agent orchestration harness designed for complex, multi-hour jobs. Instead of acting as a thin wrapper, a lead agent breaks tasks down and spins up isolated sub-agents inside Docker containers with dedicated filesystems and persistent memory. It works with local or cloud models like DeepSeek and GPT-4.
Anthropic just launched Channels for Claude Code, allowing developers to manage coding sessions directly through Telegram and Discord via MCP. You can now triage CI bugs and trigger fixes straight from your phone instead of being tethered to a terminal. It is currently in preview and requires a persistent local process.
OpenAI just launched a perk giving verified university students in the US and Canada $100 in free Codex credits, equivalent to 2,500 ChatGPT credits. Eligible students can claim the balance via SheerID using a school email, and the credits apply across all ChatGPT plans with a 12-month expiration. It’s an aggressive bid to hook the next generation of builders early.
GreenSock just rolled out gsap-skills, an official skill pack that teaches 40+ AI agents—including Cursor, Claude Code, and Copilot—how to write clean GSAP code. It covers timelines, ScrollTrigger, and React patterns, basically stopping your coding assistant from hallucinating outdated animation syntax.
OpenAI is reportedly working on a desktop “superapp” that unites ChatGPT, its Codex coding tool, and the Atlas browser into one client. An internal memo cited by The Wall Street Journal reveals that applications CEO Fidji Simo wants to curb fragmentation and drop “side quests” as competition heats up with rivals like Anthropic. The mobile app will remain separate.
Lovable just expanded beyond building full-stack apps by letting its AI agent run Python scripts in a secure environment. Users can now drop in spreadsheets, PDFs, or raw data to run deep analysis and generate branded pitch decks, reports, invoices, or launch videos directly in chat. Why it matters: It bridges the gap between writing code and running the actual business in a single thread.
OpenAI just announced it will acquire Astral, the team behind widely popular Python tools uv, Ruff, and ty. After closing, Astral's engineers will join OpenAI to embed their tooling directly into Codex, which now sees over 2 million weekly active users. The goal is to evolve Codex from a code-generating tool into an agent that handles the entire development lifecycle.
Cursor just launched Composer 2, its new frontier-level coding model now available directly in the editor. Trained via continued pretraining and reinforcement learning on long-horizon tasks, it hits 61.3 on CursorBench while cutting pricing to $0.50/M input and $2.50/M output tokens. A faster variant is also available and set as the default option.
Google Labs just transformed its Stitch tool into an AI-native design canvas designed for "vibe design." Users can now talk directly to the canvas to critique layouts, generate high-fidelity UIs from business goals, and instantly turn screens into interactive flows. It bridges the gap from raw sketches to production-ready code via an MCP server integration.
Google just completely overhauled the vibe coding experience inside Google AI Studio, supercharging it with the new Antigravity coding agent. It now natively integrates Firebase for instant database and authentication provisioning, supports Next.js, and builds real-time multiplayer apps straight from prompts. Basically, prototyping has officially turned into production-ready software development.
Apple is pushing back on vibe coding apps, quietly halting App Store updates for Replit and Vibecode. The company cites longstanding rule 2.5.2, which prohibits apps from downloading and executing code that alters their core functionality—putting app-generated live previews in the crosshairs. It marks the first major collision between App Store gatekeeping and AI-native software creation.
Xiaomi just officially launched MiMo-V2-Pro, its flagship agent model that previously dominated OpenRouter under the stealth codename "Hunter Alpha." It packs over 1T total parameters (42B active) and a 1M context window, claiming to outcode Claude 4.6 Sonnet on serious software tasks. Why it matters: Hardware giants are racing to build the central brains for autonomous AI agents.
CodeRabbit just launched Plan, a collaborative tool designed to turn vague ideas into structured prompts before coding agents touch your codebase. It pulls context directly from repositories, tickets, and Notion to generate editable phased plans, cutting down on rework and dev churn.
MiniMax just launched M2.7, a reasoning-focused model that autonomously handled 30% to 50% of its own reinforcement learning workflow, including debugging and metric analysis. It hit a 66.6% medal rate on MLE Bench Lite and scored 56.22% on SWE-Pro. Why it matters: it marks an early shift toward recursive self-improving agent architectures, available now via API.
Chrome 146 just rolled out experimental native support for WebMCP, turning web pages themselves directly into MCP servers. Instead of relying on clunky CDP setups or external backends, agents can now speak directly to frontend JavaScript to navigate, read cookies, and complete tasks with higher accuracy. It is currently locked behind a flag, but it paves the way for frictionless browser automation.
Mistral AI just unveiled Forge, a platform that lets enterprises build frontier-grade AI models using their own internal codebases, policies, and operational data. It supports pre-training, post-training, and RL across both dense and MoE architectures. Crucially, it is built agent-first, allowing coding agents like Mistral Vibe to autonomously handle hyperparameter tuning and synthetic data generation.
Google just open-sourced the Colab MCP Server, enabling any MCP-compatible AI agent to programmatically control Google Colab notebooks straight from your terminal. Tools like Claude Code or Gemini CLI can now create cells, run Python scripts, and render visualizations live in the cloud. It basically turns Colab into an automated sandbox so agents no longer run risky, resource-heavy code on your local machine.
