Core Viewpoint - The article discusses the evolution of Li Auto's autonomous driving technology from following Tesla to referencing Tesla, highlighting original innovations made by Li Auto that Tesla has not publicly addressed [2][3]. Group 1: Development Line of Li Auto's Autonomous Driving - Initially, Li Auto's autonomous driving was considered to be following Tesla, but after the introduction of VLM, it transitioned to a reference model, showcasing original innovations not mentioned by Tesla [2]. - The core innovation of Li Auto's VLA is at the DeepSeek MoE level, which is lower than the DeepSeek MLA innovation level [2]. - During the V10-11 period, it was acceptable to say Li Auto was following Tesla, but from V12 onwards, the extent of following has significantly decreased [2]. Group 2: Ashok's Presentation at ICCV 2025 - Ashok Elluswamy discussed Tesla's shift to a single, large end-to-end neural network that directly generates control actions from sensor data, eliminating explicit perception modules [4]. - The reasons for this shift include the difficulty of encoding human values into code, poor interface definitions between traditional perception, prediction, and planning, and the need for scalability to handle real-world complexities [5]. - Key challenges in learning from pixels to control include the curse of dimensionality, interpretability and safety guarantees, and evaluation [6]. Group 3: Solutions to Challenges - To address the curse of dimensionality, Tesla utilizes extensive data from its fleet and employs complex data collection methods to extract valuable corner case data [7]. - For interpretability, end-to-end models can be prompted to predict auxiliary outputs for debugging and safety assurance, with the main focus being on control actions [8]. - The evaluation challenge is addressed through a neural network closed-loop simulator that allows for comprehensive testing and performance assessment [10]. Group 4: Comparison with Li Auto - The article argues that Li Auto's prior announcements on natural language processing and 3D Gaussian representation predate Ashok's presentation, indicating that Li Auto is not merely following Tesla [13]. - The discussion highlights that Ashok's concepts lack groundbreaking ideas, suggesting that Li Auto's innovations are leading rather than following [13]. - The article also notes that Tesla's potential adoption of a VLA-based solution aligns with Li Auto's previously published architecture [16].

Core Viewpoint - Tesla is shifting to a single, large end-to-end neural network that directly generates control actions from pixel and sensor data, eliminating explicit perception modules [1][34]. Group 1: Reasons for Transition to End-to-End Neural Networks - Integrating human values (like driving smoothness and risk assessment) into code is extremely challenging [3]. - Poor interface definitions between traditional perception, prediction, and planning can lead to information loss [4]. - The end-to-end approach is easier to scale for handling long-tail problems in the real world [5]. - It allows for homogeneous computation with deterministic latency, which is crucial for real-time systems [6]. Group 2: Challenges in Learning "Pixel to Control" - The primary challenges include the curse of dimensionality, interpretability and safety guarantees, and evaluation [7][8][9]. - The input context can be extensive, with a 30-second window potentially reaching 2 billion tokens [10][49]. - Tesla leverages its vast fleet data to extract valuable corner case data through complex, trigger-based data collection methods [11][51][56]. Group 3: Solutions to Challenges - For the curse of dimensionality, Tesla refines its extensive driving data to ensure the right correlations are captured [51][56]. - Interpretability is addressed by prompting the end-to-end model to predict various auxiliary outputs for debugging and safety assurance [12][60]. - Evaluation challenges are tackled by creating a neural network-based world simulator that can generate consistent video streams from multiple cameras [19][79]. Group 4: Future Developments - The next step involves the Cyber Cab, a next-generation vehicle designed specifically for robotaxi services, utilizing the same neural network technology [25][83]. - The technology developed for autonomous driving is also being adapted for humanoid robots, such as Optimus [26][86].

Core Insights - The article provides a detailed user experience of a new electric vehicle (EV) over a 40-day period, focusing on various aspects such as driving, interior space, charging efficiency, and smart features [2][3]. Driving Experience - The vehicle's exterior design is described as unique, with a mix of positive and negative aspects regarding its appearance [4]. - The interior is noted for its simplicity and high-quality feel, significantly better than a previous model [6]. Charging Efficiency - The total distance driven is 1470.3 km with a total energy consumption of 234.4 kWh, resulting in an average energy consumption of 15.9 kWh/100 km [7]. - The vehicle's energy consumption aligns with the manufacturer's claims, with a theoretical range of 656 km based on the battery capacity [8]. - Charging times at different charging stations are documented, showing that 5C charging stations are faster than 4C stations by approximately 6 minutes [10][11]. Space and Comfort - The vehicle offers ample space for six passengers, with comfortable seating even in the third row [13]. - The driving experience is enhanced by the vehicle's comfort, allowing for long drives without significant fatigue [13]. Advanced Driver Assistance Systems (ADAS) - The vehicle's ADAS is evaluated as subjective, with both strengths and weaknesses noted in its performance [16]. - The system performs well in recognizing traffic conditions and executing smooth maneuvers, but has limitations in predicting side traffic and making quick decisions [19]. Smart Features - The vehicle's smart cockpit and user interface are praised for their functionality, particularly the customizable desktop feature that simplifies access to various functions [22]. - Voice recognition capabilities are highlighted as strong, accommodating different accents effectively [22]. Overall Assessment - The overall user experience exceeds expectations in terms of passenger space and comfort, with charging efficiency and ADAS being prioritized features [23].

Core Viewpoint - The article discusses the development and strategic importance of the self-developed operating system (OS) by the company, highlighting its advantages over traditional systems like AUTOSAR and the potential for industry-wide collaboration through open-sourcing the OS [1][6][21]. Group 1: Self-Developed Operating System - The self-developed communication middleware connects various distributed systems in the vehicle, enhancing communication and resource coordination [1][13]. - The OS breaks down traditional "black box" barriers from different suppliers, allowing for end-to-end integration and improved real-time performance [1][14]. - The integration of hardware and software is emphasized, similar to Apple's approach, which maximizes system performance [1][8]. Group 2: Technical Advantages - The OS achieves high iteration efficiency through application layer decoupling and various tools, leading to faster development and problem resolution [1][12]. - Compared to AUTOSAR, the OS offers superior cross-domain real-time capabilities and utilizes a distributed communication protocol that enhances QoS, security, and scalability [1][15]. - The system can predict and react to braking or evasive actions 7 meters in advance at 120 km/h, showcasing its advanced real-time capabilities [1][15]. Group 3: Industry Collaboration and Open-Sourcing - The initial motivation for developing the OS was to ensure supply chain security and freedom in chip selection, especially during supply shortages [3][7]. - The company encourages open-sourcing the OS to reduce redundancy in the industry and foster collaboration among various OEMs [6][19]. - The trend towards a unified OS is seen as beneficial for both car manufacturers and chip suppliers, addressing the challenges of system fragmentation [22][23]. Group 4: Challenges in OS Development - Developing a self-developed OS requires a strong foundation in business application software to inform system requirements [4][10]. - Talent acquisition and organizational structure are critical challenges for traditional car manufacturers in developing their own OS [4][11]. - The complexity of operating systems necessitates a focus on real-time performance and safety, making it unsuitable for fragmented development efforts [21].

Core Insights - The article discusses advancements in autonomous driving technology, particularly focusing on Tesla's use of similar techniques as VLA in their V14 model, highlighting the importance of spatial understanding and multitasking capabilities [1][2] - Ashok Elluswamy, Tesla's AI software VP, emphasized the integration of various data sources in Tesla's Full Self-Driving (FSD) system during a workshop at ICCV 2025, indicating a significant upgrade in their autonomous driving capabilities [1][2] Group 1: Tesla's Technological Advancements - Tesla's V14 model utilizes technology akin to VLA, showcasing enhanced spatial comprehension and multitasking abilities, which are critical for long-duration tasks [1] - Elluswamy's presentation at ICCV 2025 highlighted the FSD system's reliance on a comprehensive network that incorporates camera data, LBS positioning, and audio inputs, culminating in action execution [1][2] Group 2: ICCV 2025 Workshop Details - The ICCV 2025 workshop focused on distilling foundation models for autonomous driving, aiming to improve the deployment of large models like vision-language models and generative AI in vehicles [3] - Key topics included foundational models for robotics, knowledge distillation, and multimodal fusion, indicating a broad exploration of AI applications in autonomous driving [6][7]

Core Viewpoint - The company is expanding its operations internationally, with a focus on enhancing its autonomous driving capabilities in overseas markets, particularly in Central Asia and Europe [14][17]. Group 1: International Expansion - The company has opened its first overseas retail center in Tashkent, Uzbekistan, and plans to open two more stores in Kazakhstan in November, collaborating with leading local dealers to sell models L9, L7, and L6 [14]. - 2025 is designated as the company's global expansion year, with the establishment of R&D centers in Germany and the U.S., and plans for new vehicle adaptations for global markets starting in 2026 [14]. Group 2: Autonomous Driving Testing - The company is initiating preliminary tests of its autonomous driving features overseas, as inferred from recent social media posts by its product manager [17]. - The product manager's posts indicate the importance of having assisted driving technology while exploring cities, suggesting a focus on enhancing user experience through advanced driving aids [4][13].

Core Viewpoint - The article discusses the five stages of artificial intelligence (AI) as defined by OpenAI, emphasizing the importance of each stage in the development and application of AI technologies [10][11]. Group 1: Stages of AI - The first stage is Chatbots, which serve as a foundational model that compresses human knowledge, akin to a person completing their education [2][14]. - The second stage is Reasoners, which utilize supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF) to perform continuous reasoning tasks, similar to advanced academic training [3][16]. - The third stage is Agents, where AI begins to perform tasks autonomously, requiring a high level of reliability and professionalism, comparable to a person in a specialized job [4][17]. - The fourth stage is Innovators, focusing on generating and solving problems through reinforcement training, necessitating a world model for effective training [5][19]. - The fifth stage is Organizations, which manage multiple agents and innovations to prevent chaos, similar to corporate management [4][21]. Group 2: Computational Needs - The demand for reasoning computational power is expected to increase by 100 times, while training computational needs may expand by 10 times over the next five years [7][23]. - The article highlights the necessity for both edge and cloud computing to support the various stages of AI development, particularly in the Agent and Innovator phases [6][22]. Group 3: Ideal Self-Developed Technologies - The company is developing its own reasoning models (MindVLA/MindGPT), agents (Driver Agent/Ideal Classmate Agent), and world models to enhance its AI capabilities [8][24]. - By 2026, the company plans to equip its autonomous driving technology with self-developed advanced edge chips for deeper integration with AI [9][26]. Group 4: Training and Skill Development - The article emphasizes the importance of training in three key areas: information processing ability, problem formulation and solving ability, and resource allocation ability [33][36]. - It suggests that effective training requires real-world experience and feedback, akin to the 10,000-hour rule for mastering a profession [29][30].

Core Viewpoints - The definition of pure electric usage scenarios for range-extended vehicles is crucial for long-term market feedback [2] - Analyzing the most suitable product definition details for ideal range-extended users is complex and should not be judged lightly [2] - The core of product definition relies on a comprehensive balance of data and taste, and despite having more data than competitors, the company may not make the most suitable trade-offs [2][3] - The expected level of high-speed charging resources in China over the next 1-4 years is uncertain, although improvements are anticipated [2] - The 2026 generation of ideal range-extended vehicles may have shortcomings in battery product definition, which could be addressed in 2027 [2] Product Definition Insights - Domestic and overseas versions of range-extended vehicles may not be suitable for the same product definition due to differing charging resource expectations [3] - The company maintains a commitment to providing good quality at reasonable prices, with the D series not aimed at high-end positioning [4] - Range-extended vehicles are expected to cater to high-net-worth users who prefer convenience in long-distance travel while primarily using electric modes [4] User Behavior and Data Insights - A significant portion of users prefers to maximize electric usage during high-speed travel, driven by cost savings and smoother driving experiences [5][6] - Data indicates that approximately one-third of users primarily use gasoline at high speeds, while two-thirds prefer to use electric power as much as possible [6][7] - The company recognizes the importance of data in understanding user preferences, although there are challenges in accurately defining and interpreting data metrics [7][9] Complexity of Product Definition - The relationship between data and taste in product definition is complex and should not be oversimplified [8] - The ideal range-extended vehicle's battery definition and expected range are intricate issues that require careful consideration [8][9] - The company faces challenges in applying historical data from pure electric vehicles to the new generation of range-extended vehicles due to differences in design and user experience [9]

Core Insights - The article discusses the differences between VLM (Visual Language Model) and VLA (Visual Language Action) in the context of autonomous driving, particularly focusing on scenarios like blind spot deceleration [1][2]. Group 1: VLM and VLA Differences - VLM operates by perceiving scenarios such as uncontrolled intersections and outputs a deceleration request to the E2E (End-to-End) model, which then reduces speed to 8-12 km/h, creating a sense of disconnection in the response [2]. - VLA, on the other hand, utilizes a self-developed base model to understand the scene directly, allowing for a more nuanced approach to blind spot deceleration, resulting in a smoother and more contextually appropriate response based on various road conditions [2]. Group 2: Action Mechanism - The action generated by VLA is described as a more native deceleration action rather than a dual-system command, indicating a more integrated approach to scene understanding and response [3]. - There are concerns raised in the comments regarding VLM's reliability as an external module, questioning its ability to accurately interpret 3D space and the stability of its triggering mechanisms [3].

Core Viewpoint - The article discusses the challenges and advancements in identifying abnormal noises in vehicles, emphasizing the complexity of vehicle components and the innovative use of AI for diagnostics [2][3]. Group 1: Challenges in Noise Diagnosis - The complexity of components: Over 200 parts in vehicles can be sources of abnormal noises, each producing unique sound characteristics that require precise analysis [3]. - Environmental interference: Normal operational sounds overlap with abnormal noises, making it difficult to isolate specific signals [3]. - Dynamic diagnosis issues: Many abnormal noises are intermittent, complicating the identification process for technicians [3]. Group 2: Technological Solutions - Step 1: Sound digitization: Utilizing Fourier transform and signal processing techniques to convert chaotic sound waves into clear time-frequency graphs, creating unique "waveform fingerprints" for each noise [4]. - Step 2: Massive data training: The development of a self-owned NVH model that incorporates decades of diagnostic experience into an algorithm, allowing real-time analysis and continuous self-optimization [5]. - Step 3: Real-time fault diagnosis: The system operates in real-time on the vehicle, using edge computing to complete diagnostics within one minute and monitor multiple components simultaneously [6]. Group 3: Impact and Benefits - The deployed model helps identify over 30 hidden faults monthly with a diagnostic accuracy of 100%, saving over 3 million yuan in claims costs annually [7]. - The NVH diagnostic model reduces the time cost for after-sales technical support in resolving noise issues by 99%, enhancing customer service experiences [7].