Core Viewpoint - The article discusses the influential contributions of Andrej Karpathy in the AI field, particularly his role in coining significant terms and concepts that have shaped the industry, such as "hallucinations," "Software 2.0," "Software 3.0," "vibe coding," and "bacterial coding" [1][6][9]. Group 1: Naming and Concepts - Karpathy coined the term "hallucinations" to describe the limitations of neural networks, which generate meaningless content when faced with unfamiliar concepts [1][3]. - He is recognized as a master of naming in the AI community, having introduced terms like "Software 2.0" and "Software 3.0," which have gained traction over the years [6][9]. - The act of naming is emphasized as a foundational behavior in knowledge creation, serving as a stable target for global scientific focus [7]. Group 2: Software Evolution - "Software 1.0" refers to traditional programming where explicit instructions are written in languages like Python and C++ [12][14]. - "Software 2.0" represents a shift to neural networks, where developers train models using datasets instead of writing explicit rules [15]. - "Software 3.0" allows users to generate code through simple English prompts, making programming accessible to non-developers [16][17]. Group 3: Innovative Programming Approaches - "Vibe coding" encourages developers to immerse themselves in the development atmosphere, relying on LLMs to generate code based on verbal requests [22][24]. - "Bacterial coding" promotes writing modular, self-contained code that can be easily shared and reused, inspired by the adaptability of bacterial genomes [30][35]. - Karpathy suggests balancing the flexibility of bacterial coding with the structured approach of eukaryotic coding to support complex system development [38]. Group 4: Context Engineering - Context engineering has gained attention as a more comprehensive approach than prompt engineering, focusing on providing structured context for AI applications [43][44]. - The article highlights a shift towards optimizing documentation for AI readability, indicating a trend where 99.9% of content may be processed by AI in the future [45].

Core Viewpoint - The article discusses Andrej Karpathy's new concept of "Bacterial Code," which emphasizes small, modular, self-contained code blocks that are easy to copy and paste, inspired by the evolutionary strategies of bacteria [1][5][6]. Group 1: Concept of Bacterial Code - Bacterial Code has three main characteristics: small code blocks, modularity, and self-containment, allowing for easy replication [1][6][12]. - The idea is that open-source communities can thrive through "horizontal gene transfer," similar to how bacteria share genetic material [2][12]. - Karpathy's insights are derived from the survival strategies of bacteria, which have evolved to colonize diverse environments through efficient genetic coding [7][8]. Group 2: Principles of Bacterial Code - The first principle is "smallness," where each line of code consumes energy, leading to a natural self-optimization mechanism [8][11]. - The second principle is "modularity," where code should be organized into interchangeable modules, akin to bacterial operons, promoting high cohesion and low coupling [11][12]. - The third principle is "self-containment," meaning code snippets should be independent and not reliant on complex configurations or external libraries [13][14]. Group 3: Limitations and Future Directions - While Bacterial Code is effective for rapid prototyping, it is not suitable for building complex systems, which require more intricate structures like eukaryotic genomes [15][16]. - Karpathy suggests a hybrid approach, utilizing the strengths of both bacterial and eukaryotic coding strategies [16]. Group 4: Evolution of Software Development - Karpathy has previously introduced concepts like Software 3.0, which represents a shift towards programming with natural language models [18][25]. - He notes that software has undergone significant transformations in recent years, moving from traditional coding to model training and now to natural language programming [19][23][31]. - The future of software development will involve a collaboration between humans and large models, leading to semi-autonomous applications [28][30]. Group 5: Context Engineering - Context Engineering is highlighted as a crucial skill for effectively utilizing large language models (LLMs), requiring a balance of information to optimize performance [36][39]. - This discipline involves understanding the behavior of LLMs and integrating various elements like task descriptions and multimodal data [40][41].

Core Viewpoint - The future of AI requires a "ten-year patience" and a focus on developing "Iron Man suit" style enhancement tools rather than fully autonomous robots [3][30][34]. Group 1: Software Evolution - The software industry is undergoing a fundamental transformation, moving from Software 1.0 (human-written code) to Software 2.0 (neural networks) and now to Software 3.0 (using natural language as a programming interface) [6][10][11]. - Software 1.0 is characterized by traditional programming, while Software 2.0 relies on neural networks trained on datasets, and Software 3.0 allows interaction through prompts in natural language [8][10][11]. Group 2: LLM as a New Operating System - Large Language Models (LLMs) can be viewed as a new operating system, with LLMs acting as the "CPU" for reasoning and context windows serving as "memory" [12][15]. - The development of LLMs requires significant capital investment, similar to building power plants and grids, and they are expected to provide services through APIs [12][13]. Group 3: LLM's Capabilities and Limitations - LLMs possess encyclopedic knowledge and memory but also exhibit cognitive flaws such as hallucinations, jagged intelligence, anterograde amnesia, and vulnerability to security threats [16][20]. - The dual nature of LLMs necessitates careful design of workflows to leverage their strengths while mitigating their weaknesses [20]. Group 4: Partial Autonomy Applications - The development of partial autonomy applications is a key opportunity, allowing for efficient human-AI collaboration [21][23]. - Successful applications like Cursor and Perplexity demonstrate the importance of context management, multi-model orchestration, and user-friendly interfaces [21][22]. Group 5: Vibe Coding and Deployment Challenges - LLMs democratize programming through natural language, but the real challenge lies in deploying functional applications due to existing infrastructure designed for human interaction [24][25]. - The bottleneck has shifted from coding to deployment, highlighting the need for redesigning digital infrastructure to accommodate AI agents [25][26]. Group 6: Infrastructure for AI Agents - The digital world is currently designed for human users and traditional programs, neglecting the needs of AI agents [27][28]. - Proposed solutions include creating direct communication channels, rewriting documentation for AI compatibility, and developing tools that translate human-centric information for AI consumption [28][29]. Group 7: Realistic Outlook on AI Development - The journey towards AI advancement is a long-term endeavor requiring patience and a focus on enhancing tools rather than rushing towards full autonomy [30][31]. - The analogy of the "Iron Man suit" illustrates the spectrum of autonomy, emphasizing the importance of developing reliable enhancement tools in the current phase [33][34].

Core Insights - The core viewpoint of the article is that software has undergone two fundamental transformations in recent years, leading to the emergence of "Software 3.0," characterized by programming large models using natural language [2][5]. Group 1: Evolution of Software - Software has remained relatively unchanged for the past 70 years, but recent advancements have led to significant changes [2]. - The introduction of large models has transformed neural networks from fixed-function machines to programmable entities, where prompts serve as the programming language [4][5]. - Karpathy predicts that we are at the beginning of Software 3.0, where natural language programming will dominate [5][27]. Group 2: Attributes of Large Models - Large models possess three attributes: tool, factory, and operating system [7]. - As tools, large models require substantial initial capital investment for infrastructure, similar to building an electric grid, and are charged based on usage [8]. - The factory aspect highlights the high capital needed for training large models, akin to semiconductor manufacturing, but software's replicability makes its competitive moat less robust than hardware [9]. - Large models function as complex software ecosystems, with both closed-source giants and open-source communities coexisting [12]. Group 3: Human-like Characteristics and Limitations - Large models exhibit human-like psychological traits due to training on human data, possessing vast memory but also significant cognitive flaws [14][15]. - They can remember extensive information but may produce nonsensical outputs or errors that humans would not make, such as miscalculating simple facts [16]. Group 4: Opportunities in AI Applications - The current major opportunity in AI applications lies in developing semi-autonomous products, allowing human control while leveraging AI capabilities [17][21]. - Examples include AI tools that assist programmers without fully replacing them, maintaining human oversight [21][22]. Group 5: Future of AI and Software Development - The next decade will see a shift towards more autonomous systems, with a gradual increase in AI's role in enterprise workflows, including code, documentation, and data analysis [29]. - Long-term visions include the proliferation of intelligent assistants akin to Jarvis from "Iron Man," where human decision-making remains central [30]. - The industry will require expertise in Software 1.0 (coding), 2.0 (model training), and 3.0 (prompt engineering) [31].

Core Viewpoint - The article discusses the evolution of software in the context of AI, particularly focusing on the transition to "Software 3.0," where natural language becomes the new programming interface, and large language models (LLMs) play a central role in software development [6][8][25]. Group 1: Evolution of Software - Software development is categorized into three phases: Software 1.0 (manual coding), Software 2.0 (neural network weights), and Software 3.0 (LLMs as programming interfaces) [8][25]. - The current shift signifies a transformation where LLMs are viewed as a new type of operating system, centralizing computational power in the cloud and allowing users to interact through natural language [14][48]. Group 2: Characteristics of LLMs - LLMs are described as "defective superheroes," possessing vast knowledge but prone to errors and lacking long-term memory, necessitating careful supervision in their application [14][88]. - The article emphasizes the need for a redesign of digital infrastructure to make it more machine-readable, facilitating the development of advanced AI systems [14][38]. Group 3: Opportunities in AI Applications - The concept of "partial autonomy" in applications is introduced, where tools like Cursor and Perplexity exemplify how LLMs can enhance human capabilities while maintaining user control [101][107]. - The importance of user-friendly graphical interfaces (GUIs) is highlighted, as they improve the efficiency of human oversight in AI-generated outputs [104][117]. Group 4: Future of Programming - The emergence of "vibe coding" is noted, where individuals can create software by describing problems in natural language, thus democratizing programming [138][144]. - The article suggests that the future of software development will involve creating tools that are friendly to LLMs, enabling seamless interaction and enhancing productivity [170][179].

Group 1 - The core idea presented by Andrej Karpathy is that Software 3.0 is revolutionizing traditional programming by introducing a paradigm where "prompts are the program," requiring programmers to adapt or risk obsolescence [2][4] - Karpathy categorizes software evolution into three phases: Software 1.0 (manual coding), Software 2.0 (machine learning), and Software 3.0 (prompt-driven), emphasizing that Software 3.0 is not merely a combination of the previous two but a new entity that significantly disrupts their existence [6][11] - The emergence of large language models (LLMs) is likened to "transformers" in technology, capable of performing multiple roles, thus fundamentally altering the traditional logic of technology commercialization [7][11] Group 2 - Karpathy introduces the concept of "LLM Psychology," highlighting two main challenges: "jagged intelligence," where AI excels in complex tasks but struggles with basic logic, and "anterograde amnesia," where AI lacks memory retention beyond immediate context [10][14] - The analogy of AI as a "forgetful delivery person" illustrates its inability to retain user preferences or past interactions, suggesting the need for a "digital diary" to enhance its learning and memory capabilities [16][14] - Solutions proposed include implementing a "system prompt learning" approach, allowing AI to summarize experiences and improve decision-making over time, akin to writing a work summary after a job [14][16] Group 3 - The concept of "partial autonomy" is introduced, where AI systems are equipped with an "autonomy regulator" to balance decision-making capabilities and human trust, facilitating a more effective human-AI collaboration [18][19] - Karpathy emphasizes the importance of rapid feedback loops in human-AI interactions, suggesting that AI should generate concise proposals for quick human validation, while also setting boundaries to prevent AI from producing non-functional code [21][23] - The transition from demo to product is highlighted as a significant challenge, with the need for developers to find a balance between feature richness and reliability in AI systems [23] Group 4 - The rise of "Vibe Coding" has led to a surge in startups, indicating a transformative moment in software development akin to the early days of Bitcoin [24][27] - The current development tool landscape is described as a mix of old and new, necessitating tools that can bridge the gap and enhance AI's understanding of complex documentation [27][30] - Karpathy calls for a redefinition of user categories in tool development, focusing on human users, API-driven programs, and intelligent agents that can process data and understand human language [30] Group 5 - Karpathy advocates for practical innovation over speculative goals like achieving AGI by 2027, emphasizing the need for semi-autonomous systems that can understand human intent and make decisions [31] - The evolution of software development is framed as a shift from manual coding to a more collaborative process with AI, requiring a complete overhaul of development workflows [31] - The vision for large models is to become foundational infrastructure, similar to utilities, enabling developers to build applications without reinventing the wheel, thus reshaping the entire tech ecosystem [31]

Core Insights - The software industry is undergoing a paradigm shift from traditional coding (Software 1.0) to neural networks (Software 2.0), leading to the emergence of Software 3.0 driven by large language models (LLMs) [1][11][35] Group 1: Software Development Paradigms - Software 1.0 is defined as traditional code written directly by programmers using languages like Python and C++, where each line of code represents specific instructions for the computer [5][6] - Software 2.0 focuses on neural network weights, where programming involves adjusting datasets and running optimizers to create parameters, making it less human-friendly [7][10] - Software 3.0 introduces programming through natural language prompts, allowing users to interact with LLMs without needing specialized coding knowledge [11][12] Group 2: Characteristics and Challenges - Software 1.0 faces challenges such as computational heterogeneity and difficulties in portability and modularity [9][10] - Software 2.0 offers advantages like data-driven development and ease of hardware implementation, but it also has limitations such as non-constant runtime and memory usage [10][11] - Software 3.0, while user-friendly, suffers from issues like poor interpretability, non-intuitive failures, and susceptibility to adversarial attacks [11][12] Group 3: LLMs and Their Implications - LLMs are likened to utilities, requiring significant capital expenditure for training and providing services through APIs, with a focus on low latency and high availability [16] - The training of LLMs is compared to semiconductor fabs, highlighting the need for substantial investment and deep technological expertise [17] - LLMs are becoming complex software ecosystems, akin to operating systems, where applications can run on various LLM backends [18] Group 4: Opportunities and Future Directions - LLMs present opportunities for developing partially autonomous applications that integrate LLM capabilities while allowing user control [25][26] - The concept of "Vibe Coding" emerges, suggesting that LLMs can democratize programming by enabling anyone to code through natural language [30] - The need for human oversight in LLM applications is emphasized, advocating for a rapid generation-validation cycle to mitigate errors [12][27] Group 5: Building for Agents - The focus is on creating infrastructure for "Agents," which are human-like computational entities that interact with software systems [33] - The development of agent-friendly documentation and tools is crucial for enhancing LLMs' understanding and interaction with complex data [34] - The future is seen as a new era of human-machine collaboration, with 2025 marking the beginning of a significant transformation in digital interactions [33][35]

Core Viewpoint - The article discusses a paradigm shift in software development driven by AI, marking the transition to "Software 3.0," where natural language replaces traditional coding as the primary interface for programming [1][2]. Group 1: Evolution of Software - Software is undergoing a profound transformation, with the last 70 years seeing little change until recent years, which have witnessed two major shifts [5]. - The emergence of "Software 2.0" involves using neural network weights instead of traditional code, indicating a new software paradigm [8][16]. - The current "Software 3.0" allows developers to use natural language prompts to interact with large language models (LLMs), simplifying the programming process [17][19]. Group 2: Impact on Developers and Users - The evolution of programming lowers barriers for developers and enhances user interaction, making software more intuitive and collaborative [2][4]. - The relationship between humans and machines is at a historical turning point, with future software acting as intelligent partners rather than mere tools [2][4]. Group 3: Characteristics of LLMs - LLMs are likened to public utilities, requiring significant capital investment for training and offering services through APIs, similar to electricity distribution [29][31]. - LLMs exhibit properties of both a "wafer fab" and an "operating system," indicating their complex nature and the need for substantial infrastructure [38][39]. - The current state of LLMs is compared to the computing landscape of the 1960s, suggesting that they are still in their infancy [51][67]. Group 4: Opportunities and Challenges - LLMs present opportunities for creating partially autonomous applications, allowing for more efficient workflows and collaboration between humans and AI [95][102]. - The need for effective context management and user interfaces is emphasized to enhance the interaction between users and LLMs [97][110]. - The article highlights the importance of refining documentation and tools to make them more accessible for LLMs, which can unlock new applications [152][161]. Group 5: Future Directions - The future of software development will involve a gradual increase in the autonomy of AI systems, with a focus on maintaining human oversight [135][172]. - The concept of "vibe coding" is introduced as a new way for individuals to engage with programming, making it more accessible to a broader audience [140][144]. - The article concludes with a call to action for developers to embrace the new paradigm and build systems that leverage the capabilities of LLMs effectively [170][172].

Group 1 - Google has launched the Gemini 2.5 series, with the Flash-Lite version being the fastest and most cost-effective at $0.1 per million tokens [1] - Gemini 2.5 demonstrates human-like behavior in gaming scenarios, showing panic when health is low, which affects reasoning capabilities [1] - The 2.5 series utilizes a sparse MoE architecture, supporting multimodal inputs and long texts of up to millions of tokens, outperforming previous generations [1] Group 2 - Microsoft introduced three innovative algorithms: rStar-Math, LIPS, and CPL, which enhance large model inference capabilities [2] - rStar-Math improves mathematical reasoning quality through self-evolution and Python code validation, while LIPS optimizes mathematical proof strategies [2] - CPL algorithm significantly boosts cross-task generalization abilities by searching high-level abstract planning spaces [2] Group 3 - MiniMax has released the Hai Luo 02 video generation tool, capable of creating 10-second 1080P videos, ranking second in international video generation projects [3] - Hai Luo 02 achieves realistic physical effects and supports multilingual prompts, generating videos in a single attempt [3] - Four out of the top five video generation companies in the international rankings are Chinese, highlighting China's leading position in this field [3] Group 4 - Meta is collaborating with Italian luxury brand Prada to develop AI smart glasses, expanding partnerships beyond EssilorLuxottica [4] - Meta plans to launch Oakley smart glasses for athletes on June 20, priced around $360, featuring enhanced weather resistance [4] - Since 2023, Meta and Luxottica have sold 2 million pairs of Ray-Ban smart glasses, with plans to increase annual production to 10 million by the end of 2026 [5] Group 5 - Luo Yonghao's digital persona completed its first e-commerce live stream on Baidu, attracting over 13 million viewers and generating a GMV of over 55 million yuan [6] - Baidu's Hui Bo Xing technology enabled a unified five-dimensional presentation during the live stream, with AI accessing its knowledge base 13,000 times [6] - Baidu aims to add 100,000 digital personas and invest 100 million yuan to scale the digital persona live streaming industry [6] Group 6 - The "Six Little Dragons" of large models have faced significant executive turnover, with 22 executives leaving in the past six months [7] - Companies like Zero One and Baichuan Intelligence are shifting strategies, with Zero One abandoning large model training for Alibaba Cloud [7] - Commercialization is critical for survival, and the "Six Little Dragons" must find differentiated applications in the open-source large model era [7] Group 7 - Hong Kong University of Science and Technology has released the first medical world model, MeWM, which simulates tumor evolution and treatment planning [8] - The system achieves a Turing test accuracy of 79% and demonstrates an F1-score of 64.08% in liver cancer TACE treatment, nearing professional doctor levels [8] - MeWM's survival risk prediction C-Index is 0.752, indicating a 13% performance improvement when integrated into physician decision-making [8] Group 8 - Andrej Karpathy introduced the concept of Software 3.0, emphasizing the shift from traditional coding to prompt engineering in AI development [10] - He highlighted the limitations of LLMs, including "jagged intelligence" and "forward amnesia," necessitating new paradigms for storing problem-solving strategies [10] - AI product design should focus on human-agent collaboration, treating agents as new consumers of digital information [10] Group 9 - Sam Altman predicts that AI will achieve autonomous research capabilities within the next 5-10 years, significantly enhancing scientific discovery [11] - OpenAI envisions an "AI companion" that integrates into daily life, understanding user goals and proactively offering assistance [11] - Altman critiques Meta's talent acquisition strategy, suggesting it lacks innovation and that humans will adapt quickly to the superintelligent era [11] Group 10 - Stanford's research indicates a significant mismatch in AI startup investments, with 41% directed towards low-priority areas that do not meet employee needs [12] - A majority of employees prefer a "human-machine equal partnership" model, with only 17.1% in the arts welcoming automation [12] - The value of skills has shifted, with teaching others now ranked second in demand, highlighting the growing importance of interpersonal skills over information processing [12]

Andrej Karpathy 在 YC AI 创业营的演讲火了。 「我们并非处于智能体之年，而是身处智能体的十年时代。」 从软件和大模型发展开始讲起，阐述了今天做软件开发需要具备什么样的技能，以及 LLM 时代，新的交互和人机关系。 以及，如何真正迈向软件 3.0 时代——提示词即应用的时代。 虽然官方还未发布现场视频，但 Latent Space 对推特上相关推文进行了整理，基本上整理出来了 PPT 的完整内容。我们第一时间进行了编译处理。 第一天的 YC AI Startup School 还有其他一些嘉宾的精彩分享，Sam Altman、李飞飞、马斯克和 Aravind Srinivas 等人的观点，我们一并整理了放在文中。 TLDR: 超 6000 人的「AI 产品市集」社群！不错过每一款有价值的 AI 应用。 邀请从业者、开发人员和创业者，飞书扫码加群： 不同于我之前对软件 2.0 图表的修改方式， Andrej 此次推出了全新的示意图，展示了软件 1.0、2.0、3.0 交错共存的状态，并指出「软件 3.0 正在取代 1.0/2.0」，且「大量软件将被重写」。 01 软件 3.0： 提示词即软件 ...