【独家|网易有道推出全场景个人助理Agent"LobsterAI"】智通财经2月11日电，智通财经记者获悉，网 易有道今日正式推出桌面级Agent"LobsterAI"（中文名：有道龙虾），该产品定位为"7×24 小时帮你干 活的全场景个人助理Agent"，目前已在官网开放内测申请。据悉，LobsterAI采用了类似Claude Cowork 的直观GUI界面，无论是资讯获取、日程管理、还是深度数据分析，用户只需与其对话，LobsterAI便能 在获得授权后，自动在本地计算机中通过程序化方式执行复杂流程，并交付结果。在设备支持上， LobsterAI目前已打通移动端与PC端的连接，用户可通过手机端在钉钉、飞书等软件中进行远程交互。 （智通财经记者 付静） 转自：智通财经 ...

Group 1: Adobe Insights - The stock of Adobe is under pressure due to concerns about the disruptive impact of AI, and the company may need time to stabilize before accelerating growth [1] - The analyst notes that new users can start with free products, leading to a longer monetization period, with 70 million active monthly users currently [1] - Adobe's upcoming summit in April may focus on product data affected by AI rather than an "AI-first" approach, with new AI-driven factors expected to include shifts to third-party models and increased demand for higher resolution [1] Group 2: Oracle Developments - Oracle announced a capital plan for 2026, aiming to raise up to $50 billion through debt and equity financing, which includes an unexpected equity issuance [2] - The market's initial reaction suggests that the financing arrangement has somewhat eased market sentiment, despite the focus on credit discussions since last October [2] - Oracle remains confident in achieving its 2030 goals, with data center construction and delivery on track, and anticipates more orders from existing AI clients as the AI arms race continues [2]

Core Viewpoint - ByteDance officially launched the image generation model Seedream 5.0, positioning it as a practical AI creation engine aimed at disrupting the content creation industry, directly competing with Google's Nano Banana Pro [1][11]. Group 1: Technological Breakthroughs - The model's core technological advancements are evident in three areas: significant improvement in image quality, optimization of detail textures and lighting effects, enhancing the commercial usability of e-commerce posters and character portraits [3][13]. - Breakthroughs in intelligent interaction allow for real-time retrieval of raw images, addressing the traditional AI image generation's "information lag" issue, while accurately interpreting abstract commands and supporting localized brush editing [3][13]. - A deep ecological closed-loop is constructed, integrating tools like Jianying, CapCut, and Xiaoyunque, achieving full-link coverage from "generation to editing to distribution" [3][13]. Group 2: User Access and Experience - Domestic users can experience Seedream 5.0 through the image preview entry in Jianying and Xiaoyunque apps, while overseas users can access it via the CapCut integration [5][15]. - The platform offers a limited-time feature allowing 20 free generations per day, with members able to unlock unlimited generations and commercial licenses [5][15]. Group 3: Industry Impact and Future Outlook - Seedream 5.0 is expected to revolutionize creation efficiency, potentially increasing content usability for ordinary users to 90% and shortening the creation cycle by 10 times, thereby accelerating the industrialization process in the self-media and e-commerce sectors [10][18]. - The competitive logic of domestic models is shifting from "parameter competition" to "ecological integration," which may lead to increased legal risks related to copyright infringement and false content [10][18].

Core Insights - The article discusses a significant shift in the software industry, where software is transitioning from being viewed as a long-term asset to a disposable product, driven by changes in production costs, organizational structures, and business models [1][4][22]. Group 1: Changing Nature of Software - Software is increasingly being developed for short-term use, often created for specific tasks and discarded after completion, rather than being maintained as long-term systems [1][3][4]. - Examples of this trend include applications developed for single events or temporary needs, such as a birthday party app or a family news app, which are deleted after use [2]. Group 2: Structural Changes in Software Production - Four structural changes are occurring simultaneously in the software industry: 1. Software is shifting from system-based to task-based forms, focusing on completing specific tasks rather than long-term operation [5][6]. 2. Business departments are taking the lead in system development, utilizing low-code and no-code platforms to create temporary solutions [7]. 3. AI development tools are making it more cost-effective to rewrite software rather than maintain it, leading to frequent replacements of internal systems [8]. 4. Result-based payment models are emerging, allowing businesses to pay for software based on quantifiable outcomes rather than long-term usage [9]. Group 3: Impacts on the Software Industry - The traditional criteria for evaluating software quality are becoming obsolete, with a shift towards valuing speed of delivery and quantifiable results over long-term maintainability [11][12]. - The focus of development is moving from building long-lasting systems to creating reusable components and workflows that can be quickly adapted for various tasks [14]. - Pricing models are evolving from annual subscriptions to more flexible structures based on results or task completion, reflecting the transient nature of many software applications [15]. - Customer relationships are shifting from long-term partnerships to project-based collaborations, requiring vendors to continuously demonstrate efficiency and results to secure future contracts [16]. Group 4: Boundaries of Software Consumption - Not all software should adopt a disposable model; critical systems related to core business functions, security, and compliance must maintain long-term viability due to their high stakes [17][18]. - The article warns against blindly applying the disposable model in inappropriate contexts, as it may lead to technical debt and a lack of understanding of key processes [20]. Conclusion - The trend of software consumerization is a natural outcome of increased production efficiency in the AI era, leading to a proliferation of software with shorter lifecycles [22][24]. - Companies must develop the ability to distinguish between different software types, determining which should be disposable and which require long-term investment [21][25].

Group 1 - The China Software Service Index decreased by 0.14% as of February 11, 2026, with mixed performance among constituent stocks, led by Qihoo 360 and followed by Guanglian Da and Shenxinfeng, while Dameng Data experienced the largest decline [1] - ByteDance's platforms, including Jianying and the AI creation platform Xiaoyunque, launched the image generation model Seedream 5.0 Preview, which supports 2K and 4K resolution outputs, enhancing image generation capabilities and allowing users to make precise adjustments [1] - GF Securities noted that AI large models significantly impact general software with high standardization and low technical barriers, but companies with vertical data barriers and private deployment capabilities will maintain structural advantages in regulated sectors like finance and energy [1] Group 2 - As of January 30, 2026, the top ten weighted stocks in the China Software Service Index accounted for 60.27% of the index, including companies like iFLYTEK, Kingsoft Office, and Tonghuashun [2] - The software ETF (159852) tracks the China Software Service Index, providing a convenient tool for investing in the computer software industry [2] - Investors can also access AI software investment opportunities through the software ETF linked fund (012620) [3]

Core Viewpoint - Over 1,400 Salesforce employees have expressed their concerns regarding the company's potential business dealings with the U.S. Immigration and Customs Enforcement (ICE), urging CEO Marc Benioff to cease all engagements with the agency [1][2]. Group 1: Employee Concerns - Employees are troubled by reports indicating that Salesforce is pitching AI technology to ICE to assist in hiring 10,000 new agents and managing tip-line reports [2]. - The letter from employees demands the cancellation of all active pitches to ICE and calls for a public statement against the presence of masked agents in U.S. cities [2]. - The letter highlights concerns that Salesforce's products may enable ICE to enhance its recruitment and operational capabilities [4]. Group 2: Context and Reactions - The employee letter follows incidents where ICE agents were involved in the deaths of U.S. citizens, raising ethical concerns about the agency's use of technology from companies like Salesforce [3]. - CEO Marc Benioff's recent comments about ICE at an employee gathering have drawn criticism from staff, further fueling discontent within the company [3]. - The letter also requests transparency regarding the services Salesforce provides to ICE and calls for a pause on any infrastructure or AI systems that could support ICE's operational scale-up [4]. Group 3: Company Performance and Market Reaction - Salesforce is facing challenges as its stock has declined approximately 27% in 2026, amid investor concerns about the impact of AI models on software company growth [5]. - Despite these challenges, Salesforce projected a growth rate of 9% to 10% for the current fiscal year, indicating a slight increase from previous periods [5].

Core Viewpoint - The article discusses the emergence of the TI-DPO framework, which addresses the limitations of the Direct Preference Optimization (DPO) method in fine-tuning large language models, particularly in identifying critical tokens that influence model performance [2][24]. Research Background and Significance - Mainstream methods face two core challenges: the binary classification trap at the sequence level, which oversimplifies data into good and bad categories, and the "pseudo" importance tied to biases in token evaluation, leading to a lack of nuanced semantic control [5][7]. TI-DPO Core Mechanism - TI-DPO introduces a hybrid weighting mechanism and triplet loss to enhance the identification of key tokens while suppressing noise, resulting in more accurate alignment compared to traditional DPO [9][10]. - The hybrid weighting mechanism combines data-driven and prior structural approaches to calculate token weights, while the triplet loss framework structures the optimization as a geometric problem, encouraging the model to generate responses closer to preferred answers [9][10]. Experimental Results - TI-DPO was tested on models like Llama-3 and Mistral-7B, outperforming over 10 alignment algorithms, including DPO and GRPO, with an average score of 62.3 on Llama-3.1-8B-Instruct [13][14]. - In specific tasks such as instruction following, truthfulness, and code generation, TI-DPO significantly surpassed DPO, SimPO, and GRPO, demonstrating its effectiveness in fine detail handling [17][20]. Case Demonstration - A medical consultation case was presented to illustrate TI-DPO's ability to identify critical tokens, showing that the model effectively understood human values rather than merely memorizing responses [22][24]. Summary and Contribution - TI-DPO represents a significant shift from coarse sequence-level optimization to more precise token-level control, clarifying each token's contribution to value alignment. The framework's performance improvements in various tasks validate the effectiveness of enhancing data granularity for model capability [25].

Core Viewpoint - Software stocks are experiencing significant declines due to concerns that artificial intelligence (AI) may disrupt their business models, leading to reduced reliance on software solutions [1][2]. Group 1: Market Trends - Investors are selling shares of software companies in anticipation of future struggles as AI changes business operations [2][3]. - The iShares Expanded Tech-Software Sector ETF, which includes major software stocks like Salesforce and Adobe, has dropped approximately 20% this year, despite a recent bounce following comments from Nvidia's CEO [8]. Group 2: Nvidia's Perspective - Nvidia CEO Jensen Huang argues that it is "illogical" to believe AI will completely replace software, suggesting instead that AI will enhance existing tools and solutions [5]. - Nvidia has significantly contributed to AI growth, reporting nearly $100 billion in profits over the past 12 months, a stark increase from under $5 billion a few years ago [6]. Group 3: Investment Opportunities - Current market conditions may present attractive buying opportunities for software stocks, with Salesforce and Adobe trading near multi-year lows [9]. - Investors are advised to evaluate each software company individually, focusing on upcoming earnings reports and growth amidst AI developments to determine the viability of investments [11].

36氪制图 OpenClaw不依赖单一模型，而是在后台协调多个Agent分工协作。它像一个中控台，让AI从"回答者"变成"执行者"。尽管目前在多步骤规划上成功率仍有衰 减，但它证明了：在App之上，已经可以建立一层新的执行结构。 "透视图"是36氪新推出的轻量化数据图文栏目——以数据透视趋势，以图片呈现要点。"Talk is Cheap. Show me the data." 作者丨邱晓芬 GitHub上的开源项目OpenClaw近期迅速走红。在与Y Combinator的访谈中，创始人Peter Steinberger抛出一句话："未来的操作系统不再需要图标，只需要意 图。"他甚至激进地判断，80%的App将因为不再被主动打开而自然消亡。这并非危言耸听，而是三种交互结构正在重塑"入口"。 36氪制图 01. 调度层重构：OpenClaw的"蜂群"思路 02. 硬件层接管：豆包手机的"视觉暴力" 36氪制图 字节跳动选择了更直接的路径——不等待接口开放，利用视觉识别让 AI 自己"用App"。这验证了界面层接管的可能性，但在当前30 TOPS的端侧算力瓶颈 下，跨应用操作仍面临延迟（约3秒）与成功率（约50% ...

21世纪经济报道新质生产力研究院综合报道 早上好，新的一天又开始了。在过去的24小时内，科技行业发生了哪些有意思的事情？来跟21tech一起 看看吧。 【巨头风向标】 近期，字节跳动最新视频生成模型Seedance2.0在即梦、豆包、小云雀等产品开启内测，引发国内国外 广泛关注。值得注意的是，由于该模型在真实感上的突出表现，引发不少网友对可能混淆现实的担忧。 对于用户反馈，字节跳动很快做出反应。2月9日，即梦创作者社群中，平台运营人员发布消息 称："Seedance2.0在内测期间收获了远超预期的关注，感谢大家的使用反馈。为了保障创作环境的健康 可持续，我们正在针对反馈进行紧急优化，目前暂不支持输入真人图片或视频作为主体参考"，并表示 平台深知创意的边界是尊重，产品调整后会以更完善的面貌与大家正式见面。 追觅创始人俞浩隔空喊话余承东：在哪上班不是上？要不加入追觅吧 2月10日，追觅科技创始人在其个人微博发文@华为常务董事、产品投资评审委员会主任、终端BG董事 长，俞浩表示"在哪上班不是上？要不加入追觅吧！"在评论区，有网友调侃"俞总调戏余总"，俞浩回 复"真诚邀约"，并且俞浩还在评论区提到"你来追觅，我不会对 ...