Breaking News

The AI programming tool Cursor announced the launch of the Voice Chat with Agents on November 6. Developers can now engage in a two-way conversation with Cursor “ s coding smart body, such as orally describing the code to modify the requirements, asking for task progress, viewing differences and operational status, while allowing the smart body to continue to perform the coding work backstage. This feature supports its use in Cursor's Charters Windows, IDE, web-end and iOS applications. When the smart body is empty and the input box is empty, the user can..

ON 7 NOVEMBER, ACCORDING TO INFORMATION RECEIVED, MS. WU WAS REPORTED TO BE IN THE LEGAL SERVICE OF AN INSURANCE COMPANY AND, IN JULY 2024, IN ORDER TO OPTIMIZE WORK PROCESSES AND IMPROVE PPT PRODUCTION AND DATA ANALYSIS EFFICIENCY, SHE WAS ABLE TO ORGANIZE SELF-PROCESSED CORPORATE AUDIT DATA SHEETS AND UPLOAD THEM TO A THIRD-PARTY AI PLATFORM THAT HAD BEEN REPEATEDLY RECOMMENDED BY THE HEAD AND IMMEDIATE HEAD OF THE CORPORATE SECTOR FOR USE IN SUPPORT OF THE OFFICE. ONE MONTH LATER, SHE WAS DISMISSED BY THE COMPANY FOR “SERIOUS BREACHES OF RELEVANT PROVISIONS SUCH AS INFORMATION SECURITY”. MS. WU FILED AN APPLICATION FOR ARBITRATION WITH THE LABOUR PERSONNEL ARBITRATION COMMISSION OF THE ZHONGYANG DISTRICT OF BEIJING, WHERE THE COURT FOUND THAT MS. WU HAD NOT DISCLOSED THE COMPANY'S SECRET WITH SUBJECTIVE INTENT AND INSUFFICIENT..

On 7 November, the test found that the DeepSeek dialogue function was again inoperable and that the chat interface once showed “server is busy, please try again later”. According to the DeepSeek official status page, the DeepSeek website/API section has been cut off since 15:11 today and the official inquiry is being conducted to determine why. The services affected included Chat Service, DeepSeek V4 Pro API and DeepSeek V4.1 Flash API. At present, DeepSeek..

The Chief Finance Officer of OpenAI stated that, regardless of the outcome of the election, OpenAI would continue to advance the United States data centre expansion plan. (First financial account)

Google announced the launch of the AI Video Generation Model, Gemini Omni 1.2 Flash, to support the original generation of up to 15 seconds of video and to add additional dialogue editing capabilities. Compared to the previous generation model, the maximum length of video was raised from 10 to 15 seconds, with an extension of 50 per cent. The user can use one second to set the generation time between 3 and 15 seconds. Google says that the new model enhances compliance with reference content and allows for better role, object in multiple lenses..

OpenAI indicated on Thursday that it would publish its data centre water consumption, make public its construction programme and commission an independent noise study before the data centre became operational. The company also undertook not to require the signing of confidentiality agreements with public officials with whom it discussed the disputed facilities. Two days after voters in at least 38 jurisdictions passed the bill restricting or prohibiting the construction of data centres, OpenAI published its National Community Pact. “The New Wave”

On 5 November, local time, Google announced the official launch of the video generation model Gemini Omni 1.2 Flash, which is fully open in Gemini API and Gemini for Business. Google indicated that Gemini Omni 1.2 Flash had introduced original 15-second video creation, multi-cycle interactive editing and enhanced reference compliance. The model price was reduced by 30 per cent from Gemini Omni 1.1 and was valued at 0.07 per second at 720p resolution. (Interface News)

According to sources, OpenAI has completed some $11 billion in new financing, the first portion of the $1.25 trillion in financing for the current round. Funds were transferred to the company ' s account on Friday. The remainder of the current round — which could increase to a maximum of $20 billion — is expected to be completed in the coming weeks. The AI developer has been negotiating a large-scale financing round for more than a month following the postponement of the First Public Collection (IPO) project. OpenAI originally planned to finance $30 billion or more with a $1.4 trillion valuation, which was then scaled back against the backdrop of the collapse of the AI plate in the open market. Sin..

On 6 November, the five-hour limit was lifted on the personal version of the bean bag. Following this adjustment, individual users are no longer subject to a five-hour short-cycle limit, and use time can be arranged flexibly within a weekly limit and according to the actual working tempo. The weekly scale and other usage rules will continue to be based on the corresponding package. It is understood that previously, the soybean-packed personal version was similar to the dominant model on the market and was managed on a five-hour, seven-day double-cycle basis. Even if the week is still remaining, the user will have to wait for the short-cycle replacement to continue using the five-hour limit. Shell

On 6 November, Chief Executive Officer Zuckerberg of Meta published a string on a social platform announcing the official live voice dialogue and virtual image functions of the personal AI assistant, Muse, and spearheaded the delivery to United States users, covering mobile, web-based and Mac applications. By clicking on the top right telephone icon, a user can talk to Muse like a telephone and can interrupt at any time during their response. Muse's default image is a cartoon character nicknamed “Jolly”, which can also be replaced by a person, animal or other role. According to the information received, this update is mainly about Muse Realti..

Local time, on 5 November, World Labs, founded by Madonna Lee, launched a research preview of Demo Frameboy, moving the Atlas under the flag back to the world model to a real-time roaming interactive phase. This is also the first reply submitted by World Labs after he announced his accession to AMD. Frameboy was supported by the World Labs self-returned world model released about two months ago, which was able to process raw text, images, videos and 3D, and all inputs were combined into a shared space context, where the model was available..

ON 6 NOVEMBER, ACCORDING TO CHINESE BLUE NEWS, AT THE UTOWN SUMMIT OF THE 2026 WORLD INTERNET CONGRESS, THE FOUNDER, CHAIRMAN AND CHIEF EXECUTIVE OFFICER OF THE BOARD OF DIRECTORS OF THE SEARCH FOX SAID ZHANG ZHANG ZHONGYANG WAS “MISSING THE MOBILE INTERNET”. IN THE FACE OF THE WAVE OF LARGE-SCALE GOLD-SOLDLING MODELS FROM MAJOR BUSINESSES IN THE INDUSTRY, THERE WAS A VOICE IN THE OUTSIDE WORLD THAT THE FOX-SEARCHER'S ATTITUDE “COOL” ON AI, AND HE EXPLAINED, “NOT CALM, IT IS A SOBERING CHOICE, BECAUSE WE MISSED THE MOBILE INTERNET, SO OUR RESOURCES ARE NOT THAT LARGE, AND BIG MODELS ARE RESOURCE-INTENSIVE, AND WE CANNOT DO IT.” IN HIS VIEW, THE INDUSTRY DOESN'T HAVE TO BE HELD HOSTAGE BY THE A.I.'S HEAT

Data released last night by the DeepSeek Harness team on 6 November show that, as of 4 November alone, there were 1.54 million active users of DeepSeek Harness Day, which consumed about 92 trillion Token daily, using official API users. Data disclosure indicates that on 30 September, DeepSeek Harness Day active users (average of 7 days per day) stood at only 271,000, reaching 429,000 on 7 October, climbing to 1.54 million on 4 November, an increase of more than a million a month. (FIA)

On 5 November, Kimi K3.1 Pro of Moonshot AI officially landed at Microsoft Foundation. This is an open source weight model for intelligent missions, which supports the text understanding that the context window can be up to 1 million tokens. The reporter of the newspaper Kotseng learned that the model was the first open-source weight model of Microsoft Clouds and the second partner of the Global Head Cloud Platform that landed after Amazon AWS. "The Daily Journal of Science"

The official team of ants is today announcing the launch of the original multi-modular model Ling-3.2-flash, with 560B total parameters, 25B activated parameters, and 1M context window. The model integrates visual and linguistic abilities from the structural level, supports images, text and video input, and focuses on the four main capabilities of complex visual understanding, multimodular long-range Agent, 3D understanding and interaction, and medical multimodularity。

Under the byte-jump flag, an independent client working on the Agent-type AI products bag to the productivity scene has been officially installed with the AppGalery computer and can search for downloads. The master of the product, by upgrading its capacity from content generation to completion, can self-dismantling missions around user targets, calling tools and advancing complex workflows. The current online version of the Hongmong computer carries a double-model to support the transition between dialogue and working patterns, covering multiple core office scenarios, matching multiple use needs and supporting long-link Agent tasks under the working model. Up..

According to The Verge, Anthropic has introduced Claude's personal finance function, which allows users to connect bank and credit card accounts through Plaid to allow Claude to analyse expenditures, bills and savings. This function has only read-through privileges and is unable to transfer, manage or invest funds and does not provide investment advice. Claude has access to account balances, credit lines and a record of transactions for up to two years, but cannot obtain a bank password. Personal Financial Services This Week Pushes..

At the WIRED World Expo, held in Miami, United States, Executive Director Clement Delange, Executive Director of Open-source AI Platform Hughes Face, stated that he assessed the probability of artificial intelligence leading to human extinction (i.e. P (Doom)) as “near zero”. In his view, the more real risk than the great concern of “AI Destroying the World” was the clandestine development of frontier models by a small number of large companies in closed environments. According to Drago, the main risk for AI is not from open ecology, but from “a small number of closed car participants..

According to Business Insider, Meta no longer intends to name the AI model with fruit. According to an informed source, the social media giant has decided internally not to use fruit as an AI replacement number in the future. This adjustment is mainly due to brand positioning considerations. According to the source, the symbol named fruit is not commensurate with the ambitious AI development goals of Meta. The internal code for Meta's latest AI model Wonder Full is Watermelon. Meta says..

ON 5 NOVEMBER, THE DEMOCRATIC PARTY OF THE UNITED STATES ANNOUNCED THAT IT WOULD PROMOTE THE ANTI-AGENT SUPERINTELLIGENCE ACT, WHICH WOULD COMPLETELY BLOCK THE DEVELOPMENT OF SUPERAI AND EXPLICITLY PROHIBIT ANY INDIVIDUAL OR BUSINESS INSTITUTION FROM DEVELOPING AND DEPLOYING ARTIFICIAL SUPERINTELLIGENCE (ASI). ACCORDING TO THE DRAFT BILL, ASI IS A SYSTEM THAT HAS THE FOLLOWING CHARACTERISTICS: AN AI FRAMEWORK THAT REACHES OR EXCEEDS HUMAN COGNITIVE EXPRESSION AND CAPACITY IN A WIDE RANGE OF FIELDS AND TASKS, OR SYSTEMS THAT CAN BE EASILY MODIFIED TO DEMONSTRATE THAT CAPACITY; AND AN AI SYSTEM THAT IS CAPABLE OF PLANNING AND IMPLEMENTING A SYSTEM THAT WEAKENS HUMAN DOMINANCE AND EVEN OVERTHROWS OR DAMAGES THE UNITED STATES GOVERNMENT SYSTEM. THE BILL REQUIRES A COMPLETE MORATORIUM ON ALL ADVANCED..

Original link