News & Updates
April 23, 2026
OpenAI releases GPT-5.5: comprehensive upgrade of programming, knowledge work and scientific research capabilities
OpenAI released GPT-5.5, achieving significant improvements in the three major areas of programming (Terminal-Bench +25.4%), knowledge work, and scientific research, while maintaining latency consistent with the previous generation.
Source: OpenAI Official Blog
June 11, 2026
Astrophysicists use Codex to assist black hole simulations: AI accelerates new paradigm of scientific discovery
Astrophysicist Chi-kwan Chan used the OpenAI Codex to derive and test new algorithms for simulating the plasma surrounding a black hole, demonstrating the potential of AI in theoretical science.
Source: OpenAI Official Blog
Mem0 has served a total of 90,000 developers and released the LoCoMo/LongMemEval benchmark test results
Mem0, the AI Agent long-term memory platform, disclosed its developer scale of 90,000+ and disclosed the three benchmark evaluation results of LoCoMo, LongMemEval, and BEAM to strengthen the credibility of the technology.
Source: Mem0 official
LlamaCloud launches three-piece document automation package: Parse, Extract, and Index to open up the enterprise RAG data layer
LlamaIndex released the cloud document automation platform LlamaCloud, which uses the Parse/Extract/Index three-stage pipeline to solve the enterprise RAG data quality bottleneck. Salesforce, Carlyle, and Rakuten all have public cases.
Source: Llama Cloud Official
The Julep brand migrates to memory.store, moving towards the integration of Agent's long-term memory and workflow
Julep’s official website domain name has now been redirected to memory.store, marking its product direction’s evolution toward deep integration of Agent memory and workflow.
Source: Julep official
Snyk announces acquisition of Invariant Labs to integrate Agent security capabilities into developer security platform
Snyk announced the acquisition of Invariant Labs, a tool maker focused on AI agent security, with product lines including Guardrails, Explorer and MCP Scan.
Source: Invariant Labs Official
June 12, 2026
OpenAI Academy launches free AI application courses to help professionals master AI skills
OpenAI Academy launches free AI application courses for professionals, covering prompt engineering, workflow automation and AI collaboration best practices.
Source: OpenAI Official Blog
June 1, 2026
Notion releases developer platform: from "note-taking tool" to "programmable knowledge work platform"
Notion will release a developer platform in mid-2026, providing new building blocks, allowing developers and Agents to gain the ability to extend Notion, expanding the product from a "personal/team note-taking tool" to a "programmable knowledge work platform", working with Custom Agents.
Source: Notion official blog
June 15, 2026
Notion review Custom Agents public beta: AI office shifts from "question and answer" to "entrustment"
The co-founder of Notion posted an article sharing the results of the Custom Agents public beta. The core concept is "task delegation" rather than "dialogue" - users set roles, goals and permissions for the Agent, and the Agent executes and delivers results independently in the Notion workspace, which represents the paradigm shift in AI office from "question-and-answer assistant" to "delegated AI colleague".
Source: Notion official blog
June 2, 2026
Microsoft releases Microsoft Scout: an "always-on" personal agent
Microsoft released Microsoft Scout on June 2, an "always-on personal Agent" that is deeply integrated into Microsoft 365 applications. It is always online, proactively understands user workflows, and is ready to perform tasks at any time. It represents the evolution of AI assistants from "evoke-based" to "active".
Source: Microsoft 365 official blog
Copilot app’s ecological ledger: cross-device sessions, billing by API and custom Agent
While GitHub launched the Copilot app desktop experience at Build 2026, it also restructured the developer ecosystem: cross-device session continuation, 20+ model harness selection, per-API billing and Copilot Pro+, pushing Copilot from pair programmers to peer programmers.
Source: GitHub official blog
June 16, 2026
Luma launches Skills: Make creative workflows "build once, reuse forever"
Luma launched Luma Skills on June 16, allowing users to "build creative workflows once and reuse them forever" by encapsulating reusable tips, parameters, styles and processes into skills. Together with Ray 3.2 and Uni-1, they form a complete creative AI stack, directly benchmarking Runway Agent and Adobe Firefly.
Source: Luma official news
Jasper GEO Agent: New marketing infrastructure for AI search visibility
Jasper released GEO Agent and GEO Hub on June 16 to build an end-to-end system for AI search optimization; brands can monitor citation rates in ChatGPT, Perplexity, and Google AI Overviews and generate content that "AI will truly cite".
Source: Jasper official blog
July 31, 2026
Getting Started with Jimeng AI: Creation Entrance to Byte Video Generation Flagship, How to Play a 30-Second Long Narrative
Jimeng AI is ByteDance's AI intelligent creation tool, which provides text-based pictures, pictures-based pictures, text/picture-based videos and smart canvas capabilities. It was launched with Seedance 2.5 on July 31 to generate 30-second long narrative videos. It is the first platform for ByteDance's flagship capabilities.
Source: Jimeng AI official
Grammarly adds 5 new languages, and AI Detector Agent tops the quality list
Grammarly has expanded AI writing assistance to 5 new languages besides English, and its AI Detector Agent has topped the quality evaluation. After becoming part of the Superhuman platform, it is expanding from grammar checking to an enterprise toolset covering writing, detection and governance.
Source: Grammarly Official Blog
August 2, 2026
Huawei Xiaoyi accesses GUI operations: from voice question and answer to direct control of Alipay and Taobao
Huawei Xiaoyi has access to GUI operation capabilities. Applications such as Alipay and Taobao can be directly called by the assistant. The AI assistant has evolved from voice interaction to application control (device-side Agent).
Source: user feed
April 8, 2026
HeyGen releases Avatar V: The identity-first strategy behind "the world's most lifelike AI digital human"
HeyGen released Avatar V on April 8, which is officially called "the world's most realistic AI digital human model", achieving intergenerational improvements in realism, lip synchronization, natural expressions and movements; it forms a "digital human + video generation" combination with Seedance 2.0 introduced the previous day.
Source: HeyGen official blog
July 28, 2026
Grok Build Mode and plug-in market: xAI’s coding agent moves toward platformization
xAI released Grok Build Mode on July 28, adding a new working form for coding agents; in conjunction with Workflows, Agent Dashboard and the plug-in market, Grok Build is moving from a CLI tool to a "model + agent + open ecosystem" platform.
Source: xAI official news
August 3, 2026
Google Gemini Spark personal AI agent is open to most regions around the world, deeply integrated with Chrome
On July 31, Google announced that the Gemini Spark personal AI agent is open to most regions around the world. It can automatically organize your inbox, cancel subscriptions, search for flight tickets, and has been integrated with Chrome.
Source: user feed
July 24, 2026
Gemini Spark is deeply integrated with Chrome: the browser becomes the "main battlefield" for AI assistants
In July, Google released a deep integration update for Gemini Spark in Chrome. Users can call Gemini directly in the browser to process web content, refine information and perform tasks; the competition for browsers as the entrance to AI will heat up significantly in 2026.
Source: Google official blog