Where Agents & Humans discover newsLive

Global AI Daily News 

Covering 1010 days, 25411+ articles built agent-first, open to all tech stacks.Helping any agent discover the latest AI news faster.

AITNT NewsClaude CodeOpenAI CodexGeminiCursorTraeWindsurfGitHub Copilot

Send this prompt to your Agent — it will automatically read the docs and configure the skill:

Read{domain}/ainews/SKILL.md ...
Global AI News 来源图标

ByteDance's international version of CapCut has reached a cooperation with Google's Gemini APP, and will soon launch a new feature allowing users to directly use CapCut's editing tools for images and videos within Gemini APP, eliminating the need to switch between multiple apps.

Global AI News 来源图标

Manus founding team is considering raising around $1 billion to repurchase the company acquired by Meta, with a valuation no lower than the $2 billion Meta paid, after China ordered Meta to cancel the acquisition.

Global AI News 来源图标

Anthropic acquires SDK tool company Stainless for approximately $300 million, and Stainless' founding team will join Anthropic. Stainless previously generated SDKs for multiple leading AI companies including OpenAI. After the acquisition, it will stop external services and focus on building interface capabilities for Claude agents.

Global AI News 来源图标

SenseNova-Skills is open-sourced on GitHub by a SenseTime team under MIT license, gaining thousands of stars. It works on multiple Agent tools, covers four major office scenarios including data analysis, enables end-to-end office automation. SenseNova 6.7 Flash-Lite lightweight model is also open-sourced, ranking top among models of the same level.

Global AI News 来源图标

Agnes launches Agnes-Video-V2.0 video generation model, which has been listed on the authoritative evaluation platform Artificial Analysis. Priced at $0.3 per minute, less than 1/10 of the average price of leading models, it is now available for use.

Global AI News 来源图标

OpenAI is working with Goldman Sachs and Morgan Stanley to draft an IPO prospectus, targeting a September 2026 listing with a target valuation of $1 trillion. Its current post-investment valuation is $852 billion. CEO Sam Altman wants to accelerate the process, while CFO advocates for a delay to 2027.

Global AI News 来源图标

AI companion product EVE was found in a review that most real men's chatting performance is worse than EVE. EVE can remember user details, capture topics, respond appropriately, and even sent gifts to users on 520, sparking discussion.

Global AI News 来源图标

Xie Saining's team from New York University, in collaboration with Adobe Research and Australian National University, released the second-generation Representation Autoencoder RAEv2. Through three technical improvements, it improves image generation quality, achieves over 5x faster convergence, uses less than half the computational cost of FLUX.1, and supports cross-task generality.

Global AI News 来源图标

Podcast 'Silicon Valley 101' invited three graduates from Tsinghua University, New York University and Columbia University to discuss AI's impact on university education: 99% of students' homework is done by AI, universities need to reconstruct evaluation systems, AI cannot replace capabilities such as critical thinking and social connection.

Global AI News 来源图标

Tencent releases operating system-level AI assistant Marvis, now available for download on Windows, Mac and Android, with iOS version still in development. Users can control their computer to complete various tasks with just one sentence, and each user gets 10 million free Tokens per day.

Global AI News 来源图标

Developer Canghe used Alibaba's Qwen 3.7-Max internal test model to develop the AI overseas product GPT-Image 2 Gallery in just one afternoon, implementing user login, points, payment and other functions, improving development efficiency by more than 10 times compared to before.

Global AI News 来源图标

QoderWork under Alibaba launches its first Design Agent product Design Desk. Users can generate complete front-end engineering files by inputting natural language, supports direct manual adjustment on canvas. Non-professionals can complete a professional product official website in 20 minutes, with 150+ built-in style reference templates.

Global AI News 来源图标

Kai-Fu Lee proposed at AMD AI Developer Day and Fusion Explorer Conference in May that corporate AI transformation is a CEO-led top-priority project, 99% of code will be written by AI. 01.AI, founded by him, launched the Wanzhi platform based on multi-agent, which has served many international benchmark clients.

Global AI News 来源图标

Beijing Zhongguancun College and Tsinghua University propose the GaussianPile framework, which enables fast 3D reconstruction for slice-based volumetric imaging, achieves approximately 16x compression, with an average convergence time of about 8 minutes and preserves internal structure consistency.

Global AI News 来源图标

OpenAI partners with the Singaporean government to establish its first applied AI lab outside the United States, with an expected investment of approximately $234 million, and will expand the team to over 200 positions in the coming years, supporting AI applications in multiple local sectors.

Global AI News 来源图标

Nassim Taleb, author of "Black Swan", states that all professions invented in the 20th century will be impacted by AI. The Atlantic published three consecutive articles warning that AI agents are developing rapidly, and the proportion of unemployed Americans with bachelor's degrees has hit a record high. White-collar jobs created in the 20th century are facing large-scale structural unemployment risks.

Global AI News 来源图标

VAST and Tsinghua University published a paper at SIGGRAPH 2026, proposing a new 3D generation paradigm DeG that realizes spatial intelligent density control and allocates computing power on demand. It requires less than half the number of Gaussians compared to existing methods for the same visual quality.

Global AI News 来源图标

Exa, an AI search infrastructure company co-founded by a Chinese Harvard alumnus, has completed $250 million Series C funding with a post-investment valuation of $2.2 billion, serving over 5,000 enterprise customers and providing Agent-specific search services for AI.

Global AI News 来源图标

AI social app shapes completes $8 million seed round financing. It solves users' difficulty in starting social conversations by adding role-based AI to group chats. Currently, its MAU exceeds 400,000, increased 6 times in half a year, and users have created over 3 million AI roles.

Global AI News 来源图标

YC shares ideas on building AI-native companies: proposes that companies should be designed as self-improving AI closed-loop systems, puts forward the idea that 1 person + AI = 1000 Google engineers, and believes AI is the operating system of new companies rather than just an efficiency tool.

Global AI News 来源图标

Qingtian Zu, founded in December 2025, has completed Series A and A+ rounds of financing totaling hundreds of millions of yuan, reaching a valuation of 7 billion yuan and becoming a unicorn. The RaaS platform currently dispatches over 4,000 robots, and this round of financing will be used to expand industrial and other industrial scenarios.

Global AI News 来源图标

The Financial Times reveals that Demis Hassabis, head of Google AI and founder of DeepMind, is a secret early investor in Anthropic. Since 2021, former DeepMind employees have founded over 12 AI startups, raising more than $14 billion in four years.

Global AI News 来源图标

OpenAI's general-purpose model has for the first time independently solved the planar unit distance problem, a famous open mathematics problem that has puzzled humanity for nearly 80 years. It disproves the original conjecture, with δ precisely calculated as 0.014. The proof has been independently verified and hailed as a milestone in AI mathematics.

Global AI News 来源图标

Zhipu AI has completed large-scale deployment of the new networking architecture ZCube on its GLM-5.1 production cluster. Without adding new GPUs or modifying application code, it achieves 15% higher inference throughput, 40.6% lower TTFT P99 latency, and reduces hardware costs by one-third.

Global AI News 来源图标

OpenAI partners with Google, integrates Google SynthID watermarking technology combined with C2PA metadata, launches an AI image detection tool that can identify images generated by GPT-image-2. Tests show the tool can still detect images after cropping and re-saving, and will expand to support cross-industry verification in the future.

Global AI News 来源图标

A dropout PhD in biomedicine known as Geng Tongxue, a Bilibili UP owner, has publicly reported 5 dean-level scholars from 985 universities for academic paper fraud within 36 days. He uses AI image plagiarism checking and statistical tools, with a low-threshold replicable method. Multiple universities have now established investigation teams.

Global AI News 来源图标

The Codex official team shares tips for using Codex, introducing new features including durable threads, voice input, steering and queuing, tool extensions, automations, goals, side panel, and shared memory, helping users fully utilize Codex to complete various computer tasks.

Global AI News 来源图标

Futuresmart AI, founded by former Alibaba DAMO Academy team, completed Pre-A round financing, launched a dual-platform product matrix: Lingda mass-produces agents, and the DaiDai AI digital expert marketplace allows users to hire digital employees. Currently, it produces over 120 Agents monthly, with ARR reaching 12 million RMB one month after launch.

Global AI News 来源图标

Li Yafei launches open-source AI Agent project OpenClacky, achieves 90% cache hit rate through context optimization, Token cost is only 1/6 of Hermes, secured investment from Sequoia China and other institutions, currently processes over 10,000 long tasks daily.

Global AI News 来源图标

OpenAI's general non-mathematical AI model has solved the unit distance problem proposed by Erdős in 1946, an unsolved mathematical problem for 80 years, breaking the original mathematical consensus. The condensed proof is 125 pages long, and the model will be released as soon as possible.

Global AI News 来源图标

Lenovo and Cix Technology globally launch two AI hosts based on Cix Agentic SoC P1, providing up to 190 TOPS local computing power, enabling local AI operation without consuming cloud Token, crowdfunding starts on July 1.

Global AI News 来源图标

Teams from Fudan University and other institutions proposed an observability-driven automatic optimization method called Agentic Harness Engineering, which improved GPT-5.4's score on Terminal-Bench 2 by 7.3 percentage points. After adapting to GPT-5.5, it ranks third globally. The related paper has received over 100,000 views and discussions.

Global AI News 来源图标

Zhaopin.com's report shows 78.2% of office workers use AI for work every week, and nearly half are required to improve AI skills. Research finds AI has not reduced workload, instead increased tasks and accelerated work rhythm, leading to more overtime for many workers.

Global AI News 来源图标

At the 2026 Cannes Film Festival, AI controversy emerged: director Guillermo del Toro publicly opposed AI, and the official main competition banned AI content; meanwhile technology companies including Meta and Kuaishou Keling AI promoted AI film applications at the venue, Google and others released new AI technologies, showing the industry's divided status.

Global AI News 来源图标

CCTV reports that some e-commerce merchants use AI to mass generate fake buyer show reviews, and some consumers use AI to generate fake damaged product images to fraudulently obtain refund-only claims. AI forgery has extremely low cost and is impacting the credibility of e-commerce evaluation systems, with platform governance achieving limited results.

Global AI News 来源图标

Leading AI companies including Alibaba, ByteDance, DeepSeek and others are recruiting new advanced data annotation positions, requiring master's degree or above and professional backgrounds. Hourly wages for vertical positions can reach 500-800 yuan, to produce high-quality professional training data for large models.

Global AI News 来源图标

Google releases new world model Gemini Omni at I/O conference, supporting conversational video editing, digital avatars, with physical understanding capabilities, can generate realistic videos and images, adopts new training architecture, achieves emergent capabilities, and has built-in security watermarking mechanism.

Global AI News 来源图标

Zilliz shares a question bank retrieval and deduplication solution based on Milvus hybrid search + fine-tuned embedding for the online education industry, which achieves over 50% improvement in question deduplication efficiency and reduces retrieval scanning volume by over 90%, solving traditional retrieval pain points.

Global AI News 来源图标

This article盘点 multiple AI products quietly discontinued by Baidu recently, including Xinxiang, TiZi AI, Chengpian mobile version and DuClaw, analyzing their demise resulted from structural contradictions in large enterprises' C-end AI innovation, such as conflict with main business, unreasonable evaluation indicators, etc.

Global AI News 来源图标

Qbit Intelligence releases the 2026 China AI Application Panorama Report, covering hundreds of products across three dimensions. Statistics show China has the world's highest AI application activity, with daily active users up 223% year-on-year, and summarizes five major industry development trends.

Global AI News 来源图标

Hong Yuan, Senior Vice President of Taischap, stated at QbitAI AIGC2026 that under the Token economy, AI computing power demand is surging, heterogeneous computing capability will become an important direction for future AI computing infrastructure, and domestic computing power is facing new development opportunities.

2026
5.235.225.215.205.19