昨天刷到了新的 Figma 远程 MCP 服务，进行了一大堆升级，然后又看到 GPT-5 Codex 有 API 了。 GPT-5 Codex 的话因为有了 API 所以常见的 AI IDE 比如 Cursor 等都加上了，选择后直接用就 行。 Figam MCP 这次升级最大的一个更新就是不再需要原来复杂的添加流程和本地 Figma 客户端了。 你不需要管之前咋装，我们直接看现在就行，这里我先按 Cursor 的操作路径介绍一下。 首先我们需要找到 Cursor 的设置，在设置里面找到 MCP 这个 TAB，然后点击"New MCP Serve r"将下面的 Json 代码复制进去保存就行不需要做任何改动。 { "mcpServers": { "figmaRemoteMcp": { "url": "https://mcp.figma.com/mcp" } } } 然后回到设置页面你就会看到多了一个 Figam 的 MCP，右边还有个"Connect"按钮，我们点击， 系统会询问是不是要打开外部网站，你选择打开就行。 于是就都研究了一下，没想到这么顶啊，这个美学表现直接拉满了，下面这是直接给 GPT-5 Co ...

以下文章来源于腾讯云开发者 ，作者余志臻 腾讯云开发者 . 腾讯云官方社区公众号，汇聚技术开发者群体，分享技术干货，打造技术影响力交流社区。 目录 1 引言：一个看似倒退的选择 2 理解状态的本质 3 无状态思想的历史脉络 4 无状态设计的优势 5 现实的权衡 6 AI时代的新思考 当AI编程助手都在比拼谁的索引更智能时，Claude Code选择了每次都实时搜索、不保留任何状态。这个反直觉的设计背后，是对Unix哲学的现代传 承，也是对"什么才是好工具 " 的重新定义。 01 引言：一个看似倒退的选择 最近，Claude Code的技术选择引发了不少讨论。 有观点认为，Claude Code与Gemini放弃代码索引是"一步烂棋 " 。Milvus的技术博客更是直言不讳："Claude Code的grep-only方式会烧掉太多tokens " 。 在HackerNews的讨论中，有开发者质疑："Claude用grep，Cursor用向量搜索——我们是在技术倒退吗？ " 当主流AI编程助手纷纷采用向量索引实现语义搜索时，Claude Code却选择了grep——这个诞生于1973年的命令行工具。它不建立持久 ...

Core Insights - The article discusses how entrepreneurs can leverage exponential forces and build network effects to create lasting value in the tech industry [3][4][5] Group 1: The Power of Networks and Network Effects - Many significant internet services are networks that become more valuable as more people use them, exemplified by email and social media platforms like Facebook and Instagram [5][6] - The tech industry benefits from powerful exponential forces, such as Moore's Law, which states that semiconductor performance doubles approximately every two years, leading to rapid advancements [6][7] - Entrepreneurs should focus on identifying these exponential forces, as they will dominate any tactical product work [6][10] Group 2: Strategies for Building Networks - Successful companies often start with a strong product that attracts users, then leverage existing networks to grow, as seen with Instagram and Substack [10][11] - The challenge lies in making networks useful from the beginning, as initial user bases can be small and unappealing [12] - The emergence of "narrow startups" that charge premium prices for specialized services indicates a shift towards more focused business models in the tech landscape [23] Group 3: The Role of Branding and Pricing - Brand power and consumer inertia are significant in the tech sector, as seen with ChatGPT's rapid rise to prominence despite lacking traditional network effects [15][21] - The increasing willingness of consumers to pay higher prices for software suggests a shift in spending priorities, with software potentially consuming a larger share of disposable income [14][21] Group 4: The Impact of AI and Open Source - The rise of AI tools has diminished the need for traditional web traffic, leading to a decline in SEO-driven traffic for many websites [20][21] - Open source software has played a crucial role in democratizing technology, allowing startups to thrive with minimal initial investment [35][36] - The future of open source AI remains uncertain, with potential for it to lag behind proprietary models, but it could provide affordable solutions for consumers [36][37]

Group 1 - The "Zhaocai Cup" ETF live competition series aims to enhance investors' asset allocation and risk management skills, promoting the healthy development of the ETF market [1] - The recent probability of a Federal Reserve rate cut has increased, which may lead to a more favorable liquidity environment for the Hong Kong stock market, particularly benefiting the Hang Seng Technology Index [1][19] - Investors with limited time for industry and company analysis are encouraged to consider index products, especially those interested in both technology and pharmaceutical sectors [1][28] Group 2 - The Hong Kong technology sector experienced two significant market rallies this year, driven by factors such as valuation attractiveness and the emergence of DeepSeek technology [2][3] - DeepSeek's R2 model has positioned China as a core competitor in the AI field, leading to a revaluation of Chinese tech assets and increased investor confidence [2] - The first rally in Q1 was primarily driven by valuation and the success of DeepSeek, while the second rally from late April to early June was fueled by liquidity conditions and significant inflows from southbound funds [3][21] Group 3 - The domestic AI industry has advantages in application, with a growing number of companies and a market size nearing 600 billion yuan, indicating a robust ecosystem for AI applications [5][6] - China has a unique market advantage due to its large population and demand for AI applications, which supports rapid innovation and commercialization [5] - The AI chip sector, while lagging, is seeing the emergence of strong domestic players, with companies like Alibaba preparing for future developments in AI chips [7][12] Group 4 - AI applications are entering a phase of accelerated commercialization, with companies like Kuaishou and Meitu reporting rapid revenue growth from AI products [8][9] - The government has set ambitious goals for AI integration across various sectors by 2035, indicating a strong push for AI adoption [8] - AI programming is expected to be one of the first core applications to achieve widespread adoption, driven by advancements in large models [9] Group 5 - The Hong Kong stock market's technology sector is currently undervalued compared to global peers, with the Hang Seng Technology Index trading at a PE-TTM of less than 22 times, indicating potential for upward movement [24][26] - The anticipated Federal Reserve rate cuts are expected to improve liquidity conditions, benefiting the technology sector in Hong Kong [25][26] - Southbound funds have significantly increased their investments in Hong Kong stocks, particularly in technology and high-dividend assets, reflecting strong demand from mainland investors [21][23] Group 6 - The performance of major companies like Alibaba and Tencent in the AI sector has exceeded market expectations, with significant growth in cloud services and AI-related revenues [16][17] - Alibaba's cloud business reported a 26% year-on-year revenue increase, while Tencent's AI investments have become a core driver of its business growth [16][17] - The overall positive performance of these companies is likely to enhance the valuation sentiment for Chinese assets in the global market [18]

Core Viewpoint - OpenAI has launched a new model, GPT-5-Codex, which is a fine-tuned variant of GPT-5 designed specifically for AI-assisted programming tools, demonstrating improved performance in coding tasks and dynamic thinking time [2][3][4]. Group 1: Model Features and Performance - GPT-5-Codex features enhanced code review capabilities that can identify potential critical errors before product release, helping developers mitigate risks [5]. - Unlike static analysis tools, Codex matches the intent of pull requests (PRs) with actual differences, reasoning through the entire codebase and its dependencies, thus filling the gap left by manual reviewers [6]. - The model can dynamically adjust its thinking time based on task complexity, showing strong capabilities in handling complex engineering tasks independently for over 7 hours [9][18]. Group 2: User Experience and Feedback - Users have reported that GPT-5-Codex can autonomously run tasks for extended periods, significantly improving efficiency compared to its predecessor, GPT-5 [21][24]. - The model supports seamless switching between local and web development environments, enhancing user experience [21]. - Feedback from users indicates that GPT-5-Codex is capable of solving bugs that previous versions could not, marking a significant upgrade in performance [22][24]. Group 3: Market Context and Competition - The AI coding tools market is becoming increasingly competitive, with significant investments flowing into companies like Anysphere and Anthropic, which are also developing AI coding products [26][27]. - Anysphere recently completed a $900 million funding round, achieving a valuation of $9.9 billion, while Anthropic raised $13 billion, becoming one of the most valuable startups globally [27][28]. - The rapid growth of AI coding tools is prompting discussions about the future of programming jobs, with some users expressing concerns about job displacement due to the efficiency of AI tools like GPT-5-Codex [24][25].

Core Viewpoint - OpenAI has launched the GPT-5-Codex model, which significantly enhances programming capabilities, allowing for independent continuous programming for up to 7 hours, and introduces a new "dynamic thinking" ability that adjusts computational resources in real-time during task execution [1][4][5]. Group 1: Model Enhancements - The new GPT-5-Codex model is specifically trained for complex engineering tasks, including building complete projects from scratch, adding features, testing, debugging, and executing large-scale refactoring [8]. - In testing, GPT-5-Codex demonstrated a nearly 20% improvement in success rates for code refactoring tasks compared to the original GPT-5 [9]. - For simple tasks, GPT-5-Codex reduced output token count by 93.7%, resulting in a 10-fold speed increase in response time [11]. Group 2: Dynamic Thinking Capability - GPT-5-Codex can spend double the time reasoning, editing, and testing code for complex tasks, leading to a 102.2% increase in output token volume [12]. - The model's dynamic thinking capability allows it to adjust its computational approach during task execution, enhancing its problem-solving efficiency [4]. Group 3: Code Review and Quality Improvement - GPT-5-Codex underwent specialized training for code review, reducing the error comment rate from 13.7% to 4.4% and increasing the proportion of high-impact comments from 39.4% to 52.4% [15]. - The model can understand the true intent of pull requests (PRs) and traverse entire codebases to validate behavior through testing [15][17]. Group 4: Ecosystem and Tool Integration - OpenAI has restructured the entire Codex product ecosystem, introducing features like image input support, allowing users to input screenshots and design drafts for implementation [18]. - The updated Codex CLI now tracks progress with to-do lists and integrates tools like web search and MCP for enhanced task management [19]. - New IDE extensions bring Codex directly into editors like VS Code and Cursor, enabling seamless cloud and local task management [23]. Group 5: Market Positioning - The timing of this upgrade coincides with a decline in user subscriptions for Claude Code due to quality issues, positioning OpenAI to capture market share in AI programming [25][26].

Core Viewpoint - The article discusses the evolution of software engineering in China, emphasizing the need for a localized methodology that integrates agile principles with the unique cultural and organizational context of Chinese enterprises [5][12][14]. Group 1: Historical Context and Evolution - Wu Qiong, a key figure in the software engineering field, introduced Rational Unified Process (RUP) to China, significantly impacting the development practices of many companies [5][6]. - After experiencing the agile development wave in the U.S., Wu Qiong recognized the cultural mismatch when applying Western agile methodologies in Chinese companies, leading to the realization that a tailored approach was necessary [6][7][12]. Group 2: Challenges and Adaptation - The article highlights the contradictions between Western agile practices, which promote self-organization and flexibility, and the more controlled, hierarchical nature of Chinese corporate culture [7][12]. - Wu Qiong's transition from merely importing methodologies to creating a localized framework, known as Adapt, reflects the need for a more suitable approach for Chinese enterprises [8][14]. Group 3: The Impact of AI - The introduction of AI into software engineering is seen as a transformative force, with the potential to disrupt traditional practices and create new challenges in productivity and management [9][21]. - The article discusses the dual perception of AI tools as both productivity enhancers for management and distractions for employees, highlighting the need for a balanced approach to AI integration [9][36]. Group 4: Future Directions - The future of software engineering is expected to involve a more specialized and differentiated approach to AI agents, moving away from a one-size-fits-all model to tailored solutions for specific tasks and industries [24][25]. - The concept of managing AI agents as team members is proposed, suggesting a shift in organizational structures to accommodate this new dynamic [35][38]. Group 5: Methodology and Tools - The Adapt methodology emphasizes the importance of aligning organizational structures, task management, and data flow to enhance efficiency and effectiveness in software development [30][32][49]. - The "Zhiwei" platform is introduced as a flexible management tool that can adapt to the unique needs of organizations, contrasting with rigid off-the-shelf software solutions [52][53].

Group 1 - The core viewpoint emphasizes that startups in the Agent field should avoid competing directly with major model capabilities, as they risk being outpaced by upgrades from larger model companies [1] - It is important for companies to clearly define the domain of their Agent and the tasks it can solve, as initial market significance may appear small but can expand exponentially if the product is executed well, creating barriers to entry and enhancing competitive advantages [3] - AI Coding is identified as a critical capability within the main track of model iteration, where companies developing general-purpose Coding Agents are significantly impacted by the advancements of mainstream large models [3] Group 2 - The competitive landscape in the AI coding field is described as relentless, with no company being secure, as evidenced by the rapid growth of products like Claude code and OpenAI context, highlighting the necessity for continuous improvement in capabilities [4]

Group 1 - Andrew Ng introduces the concept of "Agentic AI" to redefine the discourse around autonomy in AI, positioning it on a spectrum rather than a binary classification [1][5][6] - Ng criticizes the prevailing narrative of "bigger is better" in AI models, arguing that the focus should be on engineering practices, multi-modal model reconstruction, and the effective use of proprietary data [1][3][4] - The current bottleneck in AI development is identified as a lack of skilled personnel capable of systematic error analysis and correction, rather than computational power [1][7][10] Group 2 - The shift in product development timelines from weeks to days has led to a new scarcity in decision-making capabilities, emphasizing the need for product managers to possess empathy and intuition rather than relying solely on data [2][20] - Ng advocates for an organizational philosophy of "hiring AI instead of people," suggesting that small, skilled teams using AI tools can achieve greater efficiency and output than traditional larger teams [2][20] - The future of AI will hinge on transforming proprietary processes and compliance constraints into "learnable organizational memory," which will be crucial for competitive advantage [2][20] Group 3 - Ng emphasizes that the development of intelligent workflows and multi-modal models are critical dimensions of progress in AI, alongside breakthroughs in new technologies like diffusion models [3][4] - The concept of self-iteration in AI is highlighted, where models generate training data for the next generation, indicating a shift towards self-sustaining evolution in AI systems [2][17] - Ng warns that organizations still using outdated workflows from 2022 will be at a competitive disadvantage, as those embracing AI will rapidly outpace them [2][22] Group 4 - The discussion reveals that the ability to automate tasks within intelligent workflows is limited by the need for human engineers to gather external knowledge and contextual understanding [9][10] - Ng points out that while many tasks can be automated, the decision of which tasks to automate is crucial, as some require human judgment and contextual knowledge that AI currently lacks [42][44] - The legal industry is cited as an example of a sector undergoing significant transformation due to AI, with firms reconsidering their staffing and operational models in light of AI capabilities [35][36] Group 5 - Ng notes that the landscape of entrepreneurship is changing, with the speed of product development increasing and the focus shifting to product management as a bottleneck [20][21] - The importance of empathy in product management is emphasized, as successful product leaders must quickly understand user needs and make informed decisions [29][30] - The conversation highlights the need for founders to adapt to rapid technological changes and the importance of technical knowledge in leadership roles [24][32]

Investment Rating - The report maintains an "Outperform" rating for the AI industry [1] Core Insights - AI Agents represent a significant evolution in AI technology, moving beyond simple command execution to autonomous decision-making and task execution, achieving performance levels equivalent to 90% of skilled adults [3][10] - The AI infrastructure is undergoing a transformation, with major cloud providers like Microsoft, Google, and Amazon enhancing their AI/Agent platforms to capture new market opportunities [3][51] - The global AI IT spending is projected to grow at a CAGR of 22.3% from 2023 to 2028, with Generative AI (GenAI) expected to account for 73.5% of this growth [3] Summary by Sections 01 Agent Definition, Technology, and Development - AI Agents are defined as intelligent entities with autonomy, planning, and execution capabilities, surpassing traditional automation [10] - Key features include autonomous decision-making, dynamic learning, and cross-system collaboration [10] 02 Agent Development Platform Layout - Major players in the AI Agent development space include Microsoft, Google, Amazon, Alibaba, and Tencent, each with distinct strategies and market focuses [3][51] 03 Model Layer and Tokens Usage Analysis - The report highlights the rapid increase in token usage, with Google's Gemini model projected to reach 980 trillion tokens by July 2025, a 100-fold increase from the previous year [3] - Domestic models like Byte's Doubao are also seeing significant growth, with daily token usage expected to reach 16.4 trillion by May 2025, a 137-fold increase [3] 04 C-end and B-end Agent Progress - C-end applications are heavily reliant on model capabilities, with significant growth in image and programming-related products [3] - B-end applications, such as Microsoft's Copilot, have over 100 million monthly active users, but face challenges related to data security and cost [3] Agent Market Size and Development Expectations - The AI Agent market is expected to reach $103.6 billion by 2032, growing at a CAGR of 44.9% [3] - The report anticipates that by 2035, AI Agents will become mainstream as cognitive companions for humans [3]