System translated (Gemini)

2026-05-13 AI Daily | Thinking Machines Debuts Real-Time Interactive Model, Autonomous Agents Enter “Goal-Oriented” Era Link to heading

Today, the AI industry is focused on a leap in autonomous productivity. Thinking Machines, founded by a former OpenAI CTO, showcased a native real-time interactive model, challenging current interaction paradigms. OpenAI Codex introduced a “goal-oriented” mode, marking the transition of Agents from conversational interaction to closed-loop task execution. Additionally, industry analysis reveals that distribution advantages and vertical domain penetration are becoming the core moats for AI commercialization.

📖 Deep Dive into This Issue’s Watch List Link to heading

Today’s Watch List focuses on the paradigm shift in AI infrastructure and deep penetration into vertical domains.

First, AI infrastructure and distribution advantages are becoming decisive factors in the competitive landscape. a16z partner Jennifer Li explored in-depth how infrastructure is reshaping storage and computing, while AppLovin’s Axon.ai platform has shown powerful monetization capabilities in ad distribution. Stratechery further analyzed the strategic positioning battles among giants like SpaceX and xAI, highlighting the role of distribution channels as a key moat in the AI era.

Second, Codex is redefining engineering and business workflows. From NVIDIA engineers using GPT-5.5-based Codex for complex tasks to AutoScout24 scaling its AI workflows and finance teams building no-code business solutions, AI agents are shifting from “assistive tools” to “autonomous productivity.” Furthermore, experiments from Parameter Golf offer cutting-edge insights into coding agents and model quantization for AI-assisted research.

These updates indicate that AI is penetrating beyond the model layer into deeper engineering practices and business logic, warranting close attention from technology decision-makers.

🌐 AI Hot Topics on X Link to heading

Topic 1: Garry Tan Merges 29,000 Lines of AI-Generated Code in 72 Hours Link to heading

  • Category: AI · News
  • Overview: Trending for: 16 hours ago, Related posts: 324
  • What happened: Y Combinator CEO Garry Tan merged 29,000 lines of AI-generated code into his development project within 72 hours.
  • Why it matters: This event showcases the incredible efficiency of AI-assisted programming tools in handling large-scale engineering tasks, marking a paradigm shift in software development from manual coding to high-level architectural management.
  • Discussion summary: Discussions focus on the quality and long-term maintainability of AI-generated code, potential technical debt risks, and whether “lines of code” is still a valid metric for developer productivity in the age of AI.

Topic 2: Mira Murati’s Thinking Machines Lab Unveils Real-Time AI Interaction Models Link to heading

  • Category: AI · News
  • Overview: Trending for: 1 day ago, Related posts: 11,000
  • What happened: The new lab Thinking Machines, founded by former OpenAI CTO Mira Murati, publicly demonstrated its real-time AI interaction models for the first time.
  • Why it matters: This progress marks an evolution in AI interaction from the traditional “prompt-response” model to low-latency, native, real-time multimodal engagement. It also reflects the pursuit of next-generation AI paradigms as top talent migrates to startups.
  • Discussion summary: The conversation is focused on whether its real-time performance can surpass GPT-4o and if the lab, under Murati’s leadership, can challenge the market position of major industry players through a more efficient model architecture.

Topic 3: Obsidian Launches Community Site for Plugins and Themes Link to heading

  • Category: AI · News
  • Overview: Trending for: 3 hours ago, Related posts: 241
  • What happened: The note-taking application Obsidian has officially launched its community site, offering a centralized web interface to browse, search, and discover third-party plugins and themes.
  • Why it matters: Obsidian is a key tool in the Personal Knowledge Management (PKM) field. Its rich ecosystem of AI plugins (such as local LLM integrations and RAG workflows) is a vital part of building AI assistants. This launch will make AI tools more discoverable and promote the developer ecosystem.
  • Discussion summary: Discussions center on the significantly improved search experience, user-friendly mobile browsing support, and the community’s hope for future features like AI-powered plugin recommendations or review systems.

Topic 4: Google Unveils Gemini Intelligence and Googlebook Laptops at Android Show Link to heading

  • Category: AI · News
  • Overview: Trending for: 11 hours ago, Related posts: 7,800
  • What happened: Google launched the Gemini Intelligence AI deep integration solution and the all-new Googlebook laptop at its Android event.
  • Why it matters: This marks Google’s full-scale hardware integration of its large model capabilities, aiming to compete directly with Apple Intelligence and Microsoft’s Copilot+ PCs through system-level AI integration.
  • Discussion summary: The discussion is focused on whether the Googlebook has the hardware prowess to challenge the MacBook, and the debate over privacy boundaries versus actual productivity gains resulting from the deep embedding of system-level AI.

Topic 5: Arcadia Mayor Eileen Wang Resigns Over China Agent Charges Link to heading

  • Category: AI · News
  • Summary: Trending for: 1 day ago, Related posts: 308,000
  • What happened: Eileen Wang, the mayor of Arcadia, California, has resigned after being accused of acting as an agent for the Chinese government and has agreed to a plea deal with the Department of Justice.
  • Why it matters: The incident reflects the United States’ strict crackdown on foreign influence. Such geopolitical frictions often spill over into the tech sector, leading to stricter security reviews for the cross-border movement of AI talent and technological cooperation.
  • Discussion summary: Discussions on X are primarily focused on concerns about political infiltration, criticism of the official’s political party, and debates over the penetration of national security risks at the local government level.

Summary of Today’s AI Public Opinion on X Link to heading

The main narrative in AI public opinion is shifting from simple conversational interactions to system-level integration and unlocking ultimate productivity. There is a strong industry consensus that AI will profoundly reshape software development paradigms and hardware forms. While AI-assisted programming and real-time interactive models demonstrate astounding efficiency, significant public disagreements persist regarding the long-term technical debt risks of AI-generated code, the privacy boundaries of system-level AI, and the actual odds of startups successfully challenging tech giants. Furthermore, the risk of political scrutiny stemming from geopolitical friction is gradually spilling over into the tech sector, potentially casting a shadow over the free movement of global AI talent and transnational technological collaboration.

💡 Influencer Insights Link to heading

Hello! I am a senior AI industry analyst. Based on the tweets from several AI leaders (Influencers) on X over the past 24 hours, I have compiled today’s in-depth AI industry observation report for you.

A. Autonomous Agents Enter the “Goal-Oriented” Era Link to heading

The most significant hotspot today is OpenAI Codex’s /goal mode and the chain reaction it has triggered.

  • Codex /goal Mode Goes Viral: Both @zhixianio and @Pluvio9yte have highly praised this feature. A user only needs to input a final goal (e.g., “refactor the module and get the tests to pass”), and the Agent can autonomously complete the closed loop of reading, modifying, testing, and self-checking without human intervention.
  • Convergence of Agent Interaction Interfaces: @dotey observed that top Agent applications like Codex, Claude Code, and Cursor are converging on their interface layouts, all adopting a three-pane layout: “session on the left, conversation in the middle, and workspace on the right.” This is considered the current optimal solution for Agent interaction.
  • Enhanced Browser Control: The Chrome extension launched by Codex provides seamless control over the browser. @Pluvio9yte believes this represents a quantum leap over traditional MCP plugins and products like Manus.

B. Industry Verticalization and Agent Infrastructure (Skills/MCP) Link to heading

  • Deep Customization for the Legal Industry: Anthropic has launched Claude for Legal, which includes 12 plugins for legal roles and over 20 MCP connectors. This signals that large models are beginning to deeply penetrate high-value vertical domains (@dotey).
  • The Rise of the Skill Ecosystem: Developers are shifting from “writing code” to “writing Skills.” @vista8 recommended SkillsVote, a management tool cataloging over 790,000+ Skills, designed to solve the challenges of Skill discovery, adaptation, and iteration. @op7418’s PPT Skill also received a major update, encapsulating design experience into the Agent.

C. Real-time Native Multimodal Interaction Link to heading

  • Thinking Machines Debuts: The company founded by former OpenAI CTO Mira Murati has released an “interaction model” that supports real-time, native interaction across audio, video, and text. @op7418 pointed out that the model uses a dual-layer “frontend interaction + backend inference” architecture to deliver an ultimate user experience where the user can interrupt at any time and the AI responds in real-time.

2. Noteworthy Unique Perspectives and Industry Outlook Link to heading

  • “Testing is the new moat”: @ruanyf argues that in an era where AI can easily replicate large software (like Next.js), code itself is no longer a moat. Only comprehensive test cases are key to preventing replication.
  • HTML will replace Markdown as the preferred output for Agents: @Pluvio9yte shares the view that Markdown limits information density and interactivity. In the future, Agents will increasingly output reports with rich visualizations, charts, and interactive capabilities via HTML.
  • “Tokenmaxxing”—Big Tech’s AI anxiety: @dotey reveals a “token farming” phenomenon within Amazon and Meta. Companies are mandating AI usage, leading employees to use Agents for meaningless tasks to inflate data, reflecting the giants’ extreme anxiety about the ROI of AI.
  • The essence of role-playing is a “granularity axis”: @lijigang, citing a paper, points out that LLM role-playing is not simple template matching but rather the model moving along a “micro to macro” perspective axis. So-called role-playing is essentially adjusting the “scale” at which the world is observed.
  • Prediction of an Agent App Store: @dotey predicts that Codex’s ambition is to become the App Store for Agent platforms. The future opportunity for small and medium-sized teams lies not in building general-purpose Agents, but in developing specialized plugins for major platforms to solve “last-mile” editing problems.

3. Recommended Tools & Resources Link to heading

Development & Agent Tools: Link to heading

  • Codex (OpenAI): Focus on its /goal mode and Chrome plugin (@zhixianio, @Pluvio9yte).
  • Claude Code: Recommended to try its newly launched agent view (manage multiple sessions in the background) and auto mode (@op7418, @zhixianio).
  • UI-TARS (ByteDance): An open-source, locally runnable on-device model that can directly control a computer to perform tasks like booking tickets and configuring software (@Pluvio9yte).
  • CloakBrowser: An open-source browser project that bypasses 57 fingerprinting detections by modifying the Chromium source code, suitable for advanced web scraping needs (@vista8).

Productivity & Learning Resources: Link to heading

  • NotebookLM (Google): Rated as a top-tier knowledge management tool by @AI_Jasonyu, ideal for digesting long documents and generating podcasts.
  • SkillsVote: For searching, installing, and managing AI Agent Skills (@vista8).
  • HeavySkill Paper Analysis: A Skill implemented by @vista8 based on a Meituan paper, supporting multiple sub-agents to think independently and synthesize conclusions.
  • CapWords: A uniquely creative AI foreign language learning tool that enhances the learning experience with real-action animations and sound effects (@nishuang).

Hardware Recommendations: Link to heading

  • Mac Studio / Mac Mini: Both @zhixianio and @op7418 recommend configuring a high-performance Mac as a local Agent server for running on-device models and PA (Personal Assistant) frameworks.

Analyst’s Brief: Today’s developments show the AI industry is shifting comprehensively from the “chatbot” era to the “workflow” era. The rapid iteration of Codex and Claude Code suggests that future core competitiveness will no longer be about model parameters, but about who can more efficiently solve complex real-world tasks through Skill systems and autonomous goal management. Meanwhile, the “data farming” phenomenon within major tech companies also reminds us that the true real-world impact of AI still requires a more objective evaluation system.

📚 Appendix: Today’s Watch List Source Updates Link to heading

Timeframe: Last 3 days; covers 16 sources; 7 updates in total

a16z Podcast (A_full) Link to heading

  • AI Infrastructure, Distribution, and the Next Wave of Software
    • Published: 2026-05-12 18:00 Beijing Time
    • Summary: - Sophie Buonassisi speaks with a16z General Partner Jennifer Li about why infrastructure is becoming one of the most important frontiers in AI.
      • They discuss how the shift to AI-native systems is reshaping everything from storage and compute to developer tools and orchestration.
      • The conversation delves into early insights from companies like ElevenLabs, analyzing why distribution has become a core advantage in AI and how founders can think about product, research, and go-to-market strategies in a rapidly changing landscape.
  • Jennifer also shares her perspective on creative tools, the role of AI in storytelling, and what the next phase of the ecosystem may look like.
  • This episode was originally broadcast on the GTMnow podcast.
    • EN Highlights:
      • Sophie Buonassisi speaks with Jennifer Li, general partner at a16z, about why infrastructure is becoming one of the most important areas in AI
      • They discuss how the shift to AI-native systems is reshaping everything from storage and compute to developer tooling and orchestration
      • The conversation explores early insights from companies like ElevenLabs, why distribution has become the defining advantage in AI, and how founders can think ab…
      • Jennifer also shares her perspective on creative tools, the role of AI in storytelling, and what the next phase of the ecosystem may look like

All-In Podcast (A_full) Link to heading

  • Charles & Chase Koch on How They Quietly Built a $150B Empire
    • Release Time: 2026-05-13 05:03 Beijing Time
    • Abstract: - Axon.ai — AppLovin’s AI advertising platform, reaching over 1 billion daily active mobile game users worldwide.
      • Full-screen video ads with a median watch time of 35 seconds.
      • Advertisers spend hundreds of thousands of dollars daily and achieve profitability; the platform is currently in a closed beta.
      • Charles Koch & Chase Koch: How they quietly built a $150 billion business empire.
    • EN Highlights:
      • (0:00) David Friedberg welcomes Charles & Chase Koch
      • (1:04) Koch Inc
      • Overview: Scale, Business Lines & History
      • (2:21) Building the Business: Early Days & Charles Koch Joins (1961)

Stratechery by Ben Thompson (A_full) Link to heading

  • SpaceX and Anthropic, xAI’s Two Companies, Elon Musk and SpaceXAI’s Future
    • Release Time: 2026-05-12 18:00 Beijing Time
    • Abstract: - The partnership between Anthropic and xAI is shocking, but also expected: Musk should double down on serving other companies.
      • $15 per month or $150 per year.
      • Provides in-depth analysis of the day’s news through three emails or podcasts per week.
      • Stratechery Interviews.
      • Interviews with CEOs of prominent public companies, founders of private enterprises, and in-depth discussions with peer analysts.
    • EN Highlights:
      • The Anthropic xAI deal is shocking but not surprising: Musk should double down on serving other companies.

OpenAI Blog (A_full) Link to heading

  • How finance teams use Codex

    • Release Time: 2026-05-12 23:00 Beijing Time
    • Abstract: - With Codex, finance teams only need to focus on building the business.
      • Start with your existing closing workbooks, revenue and expense dashboards, forecast updates, past monthly business reviews (MBRs), and lead notes.
      • Codex can transform this background information into tangible assets that your team can review, optimize, and share, all without writing any code.

  • Utilize it to reduce the time spent organizing initial drafts, allowing more focus on refining business logic, verifying data, and preparing for subsequent decisions.

  • Top 10 Scenarios for Finance Teams Using Codex. Link to heading

    • EN Key Takeaways:
      • See how finance teams can use Codex to build MBRs, reporting packs, variance bridges, model checks, and planning scenarios from real work inputs.

  • How NVIDIA engineers and researchers build with Codex

    • Release Date: 2026-05-12 08:00 Beijing Time
    • Abstract: - At NVIDIA, engineers are making Codex their tool of choice for tackling complex engineering tasks and running end-to-end machine learning experiments.
      • Codex, built on GPT-5.5 and running in production on NVIDIA GB200 and GB300 infrastructure, is capable of handling longer, more autonomous tasks, not only executing instructions but also proactively identifying issues and ideas not covered in the original prompts.
      • “Codex is our go-to tool for complex engineering tasks. Thanks to GPT-5.5, it helps me find vulnerabilities and defects in programs that other models can’t detect.”
      • ——Dennis Hannusch, Senior Software Engineer.

      • Build and Deliver Production-Grade Systems. Link to heading

    • EN Key Takeaways:
      • Teams use Codex with GPT-5.5 to ship production systems and turn research ideas into runnable experiments.

  • What Parameter Golf taught us about AI-assisted research

    • Release Date: 2026-05-12 08:00 Beijing Time
    • Abstract: - Parameter Golf brought together over 1000 participants and more than 2000 submissions to explore AI-assisted machine learning research, coding agents, quantization, and novel model design under strict constraints.
      • Parameter Golf gathered over 1000 participants and 2000 submissions to explore AI-assisted machine learning research, coding agents, quantization, and…
      • Parameter Golf’s Insights on AI-Assisted Research.
    • EN Key Takeaways:
      • Parameter Golf brought together 1,000+ participants and 2,000+ submissions to explore AI-assisted machine learning research, coding agents, quantization, and no…

  • AutoScout24 scales engineering with AI-powered workflows

    • Release Date: 2026-05-12 08:00 Beijing Time
    • Abstract: - Learn how AutoScout24 Group leverages Codex and ChatGPT to accelerate development cycles, improve code quality, and expand the adoption of AI.
      • This article from the OpenAI blog explains how AutoScout24 is scaling engineering through AI-driven workflows and shaping the broader AI and infrastructure landscape.
      • The article also reveals the practical implications for founders, operators, and investors interested in how AutoScout24 is achieving engineering scale through AI-driven workflows.
    • EN Key Takeaways:
      • Learn how AutoScout24 Group uses Codex and ChatGPT to speed development cycles, improve code quality, and expand AI adoption.