![[44b5e654-7b70-4f57-9891-3a7766fb09a6.mp3]]

OpenAI推出Agent Mode，集成浏览器与云端文件高效生成智能报告

简报：

• OpenAI即将发布全新Agent Mode（代理模式），融合Operator的浏览器自动化与Deep Research的深度研究功能。
• 用户通过简单指令可自动完成网页任务、分析Google Drive、Dropbox等云端文件，并一键生成结构化、带引用的专业报告。
• Agent Mode适用于个人旅行规划、企业数据分析等多场景，支持市场主流云存储服务，无需额外API即可与网页交互。
• 技术基础为OpenAI的CUA和预期o3大模型，具备多重安全措施，支持多用户与企业集成，未来还将开放API和SDK供开发者自定义应用。
• 该功能预计将逐步向ChatGPT Plus、Team和Enterprise用户开放，标志着OpenAI在AI代理领域的技术突破和生态扩展。

相关链接：

• OpenAI重磅来袭！Agent Mode一键解锁浏览器+云端文件，智能报告秒生成
• OpenAI Agent Mode更多应用场景解析
• Agent Mode技术与安全详解

OpenAI Launches Agent Mode, Integrating Browser and Cloud Files for Efficient Smart Report Generation

Briefing:

• OpenAI is set to release a new Agent Mode, integrating Operator's browser automation with Deep Research's in-depth research capabilities. 🤖
• Users can automatically complete web tasks, analyze cloud files from Google Drive, Dropbox, and more with simple instructions, and generate structured, cited professional reports with a single click. 📄
• Agent Mode is suitable for various scenarios, including personal travel planning and enterprise data analysis. It supports mainstream cloud storage services and can interact with web pages without requiring additional APIs.
• The technical foundation is OpenAI's CUA and the anticipated o3 large language model, featuring multiple security measures. It supports multi-user and enterprise integration, with future plans to open up APIs and SDKs for developers to customize applications.
• This feature is expected to be gradually rolled out to ChatGPT Plus, Team, and Enterprise users, marking OpenAI's technological breakthrough and ecosystem expansion in the field of AI agents. 🎉

Related Links:

• OpenAI's Heavyweight Launch! Agent Mode Unlocks Browser + Cloud Files with One Click, Smart Reports Generated in Seconds
• More Application Scenarios for OpenAI Agent Mode Explained
• Agent Mode Technology and Security Deep Dive

黄仁勋称AI将迈向“物理AI”波澜新阶段

简报：

• 7月17日，在北京第三届中国国际供应链促进博览会（链博会）上，英伟达创始人兼CEO黄仁勋表示，人工智能的下一个浪潮是“物理AI”（Physic AI），即AI将从软件和生成式AI阶段迈入可应用于物理机器和机器人等领域的阶段。
• 黄仁勋指出，生成式AI已使AI具备多模态能力，并能解决前所未有的问题，而新一轮的“物理AI”将使AI进入真实世界的物理设备，实现更多实际应用。
• 此外，黄仁勋还宣布美国政府批准英伟达H20芯片恢复对中国大陆销售，并推出新款RTX Pro GPU，加速AI和智慧工厂、工业数字孪生等应用落地。

相关链接：

• IT之家报道
• 三立新闻：黄仁勋物理AI言论

Jensen Huang: AI Set for a New Wave of "Physical AI" 🚀

Briefing:

• On July 17th, at the 3rd China International Supply Chain Expo (CISCE) in Beijing, Nvidia founder and CEO Jensen Huang stated that the next wave of artificial intelligence will be "Physical AI." This means AI will transition from the software and generative AI stages to applications in physical machines and robotics.
• Huang pointed out that generative AI has already equipped AI with multi-modal capabilities and the ability to solve unprecedented problems. The new wave of "Physical AI" 🤖 will enable AI to integrate into real-world physical devices, leading to even more practical applications.
• Furthermore, Huang also announced that the U.S. government has approved the resumption of Nvidia H20 chip sales to mainland China. He also unveiled a new RTX Pro GPU, designed to accelerate the implementation of AI, smart factories 🏭, industrial digital twins, and other applications.

Related Links:

• IT Home Report
• Sanli News: Jensen Huang's Physical AI Remarks

NeuralOS发布，可实时模拟Windows界面引领AI人机交互新体验

简报：

• 华人团队研发的开源AI操作系统NeuralOS正式推出，首次实现通过神经网络实时预测与模拟Windows操作界面；
• 系统采用循环神经网络（RNN）追踪状态、扩散神经渲染器负责界面输出，能准确反映用户操作，并提供高度还原的Windows交互体验；
• NeuralOS已发布在线体验版，代码即将开源，预示未来操作系统可由AI动态生成界面，开启人机交互新纪元。

相关链接：

• AIbase报道：NeuralOS上线
• IT之家新闻：NeuralOS模拟Windows

NeuralOS Launches: Real-time Windows Interface Simulation Heralds New Era in AI Human-Computer Interaction

Briefing:

• The open-source AI operating system NeuralOS, developed by a Chinese team, has officially launched, marking the first time a neural network has achieved real-time prediction and simulation of the Windows operating interface. 🚀
• The system utilizes Recurrent Neural Networks (RNNs) for state tracking and a Diffusion Neural Renderer for interface output, accurately reflecting user actions and delivering a highly authentic Windows interactive experience. ✨
• NeuralOS has already released an online demo version, with its code soon to be open-sourced, hinting at a future where operating system interfaces can be dynamically generated by AI, thus ushering in a new era of human-computer interaction. 💻

Related Links:

• AIbase Report: NeuralOS Launched
• IT Home News: NeuralOS Simulates Windows

台积电维持2025年全年资本支出指引不变

简报：

• 台积电在7月17日法说会上重申，2025年全年资本支出维持在380亿至420亿美元不变。
• 公司预计第三季度美元营收将在318亿至330亿美元之间，平均值为324亿美元，环比增长约8%。
• 本次法说会受到市场高度关注，机构分析师密切留意其对全球半导体景气、AI需求及先进制程的最新展望。

相关链接：

• 台积电重申今年资本支出380亿-420亿美元维持不变
• 台積電財報／台積電賺爆！每股盈餘達15.36元　護國神山Q2賺3983億元
• 半導體寒風吹來了？台積電法說會前驚見的訊號　6大聚焦重點一次看

TSMC Maintains Full-Year 2025 Capital Expenditure Guidance Unchanged

Briefing:

• During its investor conference on July 17th, TSMC reiterated that its full-year capital expenditure for 2025 will remain unchanged at US$38 billion to US$42 billion. 💰
• The company projects its third-quarter dollar revenue to be between US$31.8 billion and US$33 billion, with an average of US$32.4 billion, representing an approximate 8% quarter-over-quarter growth. 📈
• This investor conference garnered significant market attention, with institutional analysts closely monitoring TSMC's latest outlook on the global semiconductor industry, AI demand, and advanced process technologies. 👀

Related Links:

• TSMC Reaffirms This Year's Capital Expenditure of US$38-42 Billion Unchanged
• TSMC Financial Report / TSMC Profits Soar! EPS Reaches NT$15.36, 'Silicon Shield' Earns NT$398.3 Billion in Q2
• Is the Semiconductor Winter Coming? Signals Seen Before TSMC's Investor Conference - 6 Key Focus Points at a Glance

谷歌AI摘要功能上线引出版业流量担忧

简报：

• 谷歌已在美国地区的移动新闻聚合平台Discover正式上线由AI生成的新闻摘要功能，用户可以直接在搜索应用内看到三行AI摘要，取代了传统媒体标题和品牌标识。
• AI摘要下方提示内容或有误，并显示来源，支持展开查看所有来源链接。
• 谷歌称功能专注体育、娱乐等生活方式类话题，方便用户筛选内容，但出版业担忧“Google Zero”加剧——即用户无需点击即可获得关键信息，导致媒体网站流量和广告收入下滑。
• 该模式引发媒体生态可持续性与内容回报机制的行业争议，目前多家知名媒体已因流量减少而倒闭或削减运营，进一步限制了谷歌AI可用的优质内容来源。

相关链接：

• 谷歌在 Discover 引入 AI 新闻摘要，出版业再遭“流量掠夺"
• 谷歌 AI 摘要上线引发出版业担忧

Google AI Summaries Launch, Sparking Publishing Industry Traffic Concerns

Briefing:

• Google has officially launched its AI-generated news summary feature on its mobile news aggregation platform, Discover, in the US. Users can now see three-line AI summaries directly within the search app, replacing traditional media headlines and brand logos. 🤖
• Below the AI summary, a disclaimer indicates the content might be inaccurate, and the source is displayed. Users can expand to view all source links. ℹ️
• Google states the feature focuses on lifestyle topics like sports and entertainment, helping users filter content. However, the publishing industry fears an intensification of "Google Zero" – where users obtain key information without needing to click, leading to declines in media website traffic and advertising revenue. 📉
• This model has triggered industry debate on media ecosystem sustainability and content monetization mechanisms. Several well-known media outlets have already shut down or scaled back operations due to reduced traffic, further limiting the availability of high-quality content for Google's AI.

Related Links:

• Google Introduces AI News Summaries in Discover, Publishing Industry Again Faces "Traffic Plunder"
• Google AI Summaries Launch Sparks Publishing Industry Concerns