News & Updates
July 31, 2026
Gemini July review: Spark deeply binds Chrome, macOS natural language and Southeast Asia localization
Google delivered a batch of updates to Gemini in July: Deep integration of Spark and Chrome, new natural language operations on macOS, and the release of the "Gemini Southeast Asia 2026 Report" to demonstrate localization strategies; AI assistants are moving from independent applications to system-level entrances.
Source: Google official blog
Figma Config 2026 Inventory: Comprehensive evolution of Make, Motion, Weave and Design Agent
Figma Annual Conference Config 2026 released Figma Make, Figma Motion, Weave, shaders and extended design Agent; on July 31st, Make further launched the property panel and annotation functions to bridge the gap between design and code.
Source: Figma official blog
Doubao Professional Edition is connected to Seedance 2.5: Byte’s popular products and enterprise AI are heading in both directions
Byte released the Seedance 2.5 video generation model on July 31 and launched the Doubao Professional Edition. Doubao users can obtain high-quality AI video generation capabilities in conversation scenarios; behind this is Byte's adjustment of AI resource allocation from mass products to enterprise services.
Source: Doubao Official / IT Home
Cursor 2.0 review: Why should multi-agent, cloud Agent and "the most useful Tab" be the benchmark?
Cursor 2.0 is built on the VS Code core and is famous for its self-developed Composer 1, multi-file context-sensitive Tab completion, Background/Cloud Agents and MCP native support; as an AI programming benchmark, its strengths and weaknesses are worth dismantling.
Source: Cursor Official
Coze: How to publish AI agents to Feishu and WeChat without writing code
Coze is an AI Bot development platform launched by ByteDance, which allows users to build, publish and operate AI agents without programming. It can publish to Feishu, WeChat, Discord, Telegram and other channels with one click, and provides plug-ins, workflows, knowledge bases and multi-model access capabilities.
Source: Coze official
Codex settles in ChatGPT: One account manages the multi-agent programming command center
OpenAI Codex has settled in ChatGPT as the "agent programming command center", supporting multi-agent parallelism, long-term background tasks and skills, compressing the development cycle from weeks to days, and customers such as Harvey reported that iteration time has been shortened by 30-50%.
Source: OpenAI official
July 21, 2026
Gemini for macOS adds natural language capabilities: Desktop AI moves from "dialog box" to "system-level assistant"
Google launched new natural language capabilities for Gemini for macOS in July. Users can use direct voice/natural language commands to interact with desktop applications, marking the evolution of Gemini desktop from a "chat window" to an "operating system-level AI assistant", competing with OpenAI and Anthropic for desktop AI entrances.
Source: Google official blog
August 3, 2026
Google DeepMind releases Gemini Robotics 2: General-purpose AI "brain" adapts to new body within hours
Google DeepMind has released the Gemini Robotics 2 series of general-purpose AI "brain" smart kits. The three models are adapted to different hardware bodies across models. They can be adapted to a new body after a few hours of data training, and the robot's "smart layer" is positioned for development by hardware manufacturers.
Source: user feed
July 9, 2026
GPT-5.6 is connected to Figma Make: the last mile from design to code is accelerated
Figma announced on July 9 that OpenAI GPT-5.6 can be used in Figma Make, providing a starting point for development and construction with accurate first generation and rapid iteration; with MCP Server and Make, Figma is evolving into an AI infrastructure between design and development.
Source: Figma official blog
November 17, 2025
ElevenLabs releases Image & Video: Voice leader enters visual generation
ElevenLabs released ElevenLabs Image & Video on November 17, 2025, marking the voice AI company's official entry into the field of image and video generation, expanding from "voice single modality" to a "multi-modal creation platform" where users can complete image, video and voice collaborative creation on the same platform.
Source: ElevenLabs official blog
June 11, 2026
E2B completes US$21 million in Series A financing: AI sandbox execution volume exceeds 1 billion times, Fortune 100 customer coverage reaches 94%
E2B announced the completion of a $21 million Series A round of financing. During the same period, it disclosed core scale data such as a cumulative sandbox launch of 1 billion times, Fortune 100 customer coverage of 94%, and average monthly SDK downloads of 3.5 million times.
Source: E2B official
June 17, 2026
Doubao reaches 618, but it doesn’t want to help you place an order
On the eve of 618, Doubao launched the "Ask Before You Buy Doubao" on the first-level navigation bar of the homepage, running through the entire process of selecting products, receiving coupons, placing orders, and checking logistics. However, the final payment still needs to be completed manually by the user.
Source: Geek Park (authorized by 36Kr)
June 18, 2026
DeepSeek image recognition mode is officially launched: a key step towards the productization of multi-modal capabilities
DeepSeek's image recognition mode is officially launched on the web and App, using the Thinking with Visual Primitives framework, alongside "Quick Mode" and "Expert Mode", marking the transition of multi-modal capabilities from research and development to productization.
Source: IT Home
GPT-5.5 Instant’s health assessment capabilities have jumped, and doctors’ evaluations surpass those of their human counterparts
GPT-5.5 Instant reaches state-of-the-art performance on HealthBench and HealthBench Professional benchmarks, with answers better than those of human doctors in double-blind physician evaluations.
Source: OpenAI Official Blog
July 30, 2026
GitHub Copilot app launched Stacked Sessions and Stacked PRs: infrastructure for multi-agent parallel development
GitHub introduced the practice of Stacked Sessions and Stacked Pull Requests in Copilot app on July 30, and Stacked PRs entered public preview on the same day; it supports splitting a large change into multiple interdependent small PRs, and maintains multiple Agent sessions in parallel, improving the efficiency of collaboration in large code bases.
Source: GitHub official blog
June 2, 2026
GitHub Copilot app launches Agent native desktop experience: facing Cursor and Devin head-on
At the Microsoft Build conference in June 2026, GitHub launched the Agent native desktop experience of the Copilot app, upgrading Copilot from "in-IDE code completion" to an "independent Agent native development environment" to support Agents to independently perform multi-step development tasks.
Source: GitHub official blog
June 16, 2026
Copilot Cowork is officially available: Microsoft hands over "long-term multi-step tasks" to the AI execution layer
Microsoft announced on June 16 that Copilot Cowork is officially available, positioning it as the "execution layer" of Microsoft 365. Users can delegate long-term, multi-step, cross-application work tasks; officials call it the fastest-growing feature in the history of the Frontier plan, and user satisfaction ranks first among all Copilot/Agent experiences.
Source: Microsoft 365 official blog
July 14, 2026
Claude for Teachers is online: Anthropic adds a key piece to the education landscape
Anthropic released Claude for Teachers on July 14, which provides educators with teaching scenario solutions such as lesson preparation, classroom interaction and homework feedback, and competes head-on with the layout of Google and Microsoft in the education market.
Source: Anthropic Official News
June 9, 2026
Anthropic Releases Claude Fable 5: The First Mythos-Class Universal Model with Safety Guardrails
Anthropic releases the Mythos-class models Claude Fable 5 (universal version with safety guardrails) and Mythos 5 (cyber defense unrestricted version). Priced at $10/$50/million tokens, Stripe can complete two months’ worth of work in one day, and Evozyne’s drug design speeds up 10 times.
Source: Anthropic official announcement
July 23, 2026
OpenAI puts "health" into ChatGPT: AI assistant starts to take care of your body
On July 23, OpenAI announced the launch of the "Health" function in ChatGPT, extending the AI assistant capabilities to personal health management scenarios; together with Presence, ChatGPT for Academic Researchers, etc., it forms ChatGPT's multi-layered consumer product matrix.
Source: OpenAI official news