5/Next Steps:Deploy the optimized zkmlface on a GPU TEE-enabled machine once available.Compile the pruned Gemma3 graph into high-efficiency GPU kernels for integration testing.Stay tuned for more updates ...

Core Viewpoint - Dongxin Semiconductor Co., Ltd. has shown significant stock performance with a year-to-date increase of 281.04%, reflecting strong market interest and trading activity [1][2]. Company Overview - Dongxin Semiconductor, established on November 26, 2014, and listed on December 10, 2021, focuses on the research, design, and sales of general-purpose storage chips with a revenue composition of NAND (57.08%), MCP (25.88%), DRAM (10.43%), NOR (6.15%), and others [2]. - The company operates within the electronic semiconductor industry, specifically in digital chip design, and is involved in various concept sectors including automotive chips and consumer electronics [2]. Financial Performance - For the period from January to September 2025, Dongxin Semiconductor reported a revenue of 573 million yuan, marking a year-on-year growth of 28.09%, while the net profit attributable to shareholders was -146 million yuan, a decrease of 12.16% compared to the previous year [2]. - The company has distributed a total of 135 million yuan in dividends since its A-share listing, with 55.72 million yuan distributed over the last three years [3]. Shareholder Structure - As of September 30, 2025, the number of shareholders increased by 168.45% to 51,300, with an average of 8,627 shares held per shareholder, a decrease of 62.75% [2]. - Notable institutional shareholders include the Harvest CSI Star Market Chip ETF, which is the third-largest shareholder, and new entrants like Hong Kong Central Clearing Limited and Guolian An Semiconductor ETF [3].

Summary of the Conference Call for Jingjia Micro Company Overview - **Company**: Jingjia Micro - **Industry**: Semiconductor, specifically focusing on AI SOC chips, GPUs, and military applications Key Financial Performance - **2025 H1 Revenue**: Increased by 230% year-on-year, achieving a profit of 0.15 billion CNY, reversing last year's loss situation [2][4] - **2025 Q3 Revenue**: Reached 3.01 billion CNY, a 230% increase year-on-year, with a significant improvement from last year's loss [4] - **2025 Full Year Outlook**: Expected to meet initial performance targets despite capacity constraints [2][4] Business Segments and Growth Prospects - **Display Control Business**: Anticipated significant growth in 2025 due to the ability to ensure delivery amid tight wafer supply [2][5] - **Radar and Communication Business**: Expected to see larger growth in 2026, with new product lines laying the foundation for future expansion [2][5] - **Military Industry Demand**: Stable and consistent growth anticipated, with ongoing supply challenges for key components like wafers and DDR [6] Product Development and R&D - **AI SOC Chip Project**: The Wuxi Chenhongwei AI SOC chip company has completed Tape Out (TO) and is expected to have results soon, viewed as a key growth point [2][8] - **GM11 Server Product**: Exceeded expectations in testing, attracting numerous server manufacturers for collaboration [2][9] - **R&D Projects**: Focus on two main projects: computing cards and traditional rendering GPUs, with significant investment in talent and resources [3][13][15] Competitive Landscape - **New GPU Competitors**: Increased market competition from newly listed GPU companies, but Jingjia Micro is responding by optimizing technology and expanding application scenarios [11] - **Product Comparison**: Jingjia Micro's products show competitive advantages, such as supporting more output channels compared to competitors [12] Talent Acquisition and R&D Focus - **Talent Expansion**: Actively recruiting in the chip field, with nearly 200 new hires and a new R&D center established in Shanghai [15] - **Strategic Focus**: Emphasis on maintaining high R&D investment to expand product categories and capabilities, prioritizing revenue over short-term profit [14]

Core Viewpoint - Beijing Lier's stock price has shown significant volatility, with a year-to-date increase of 70.86%, but a recent decline of 17.42% over the past 20 days, indicating potential market fluctuations and investor sentiment changes [1][2]. Company Overview - Beijing Lier High-Temperature Materials Co., Ltd. was established on November 8, 2000, and went public on April 23, 2010. The company specializes in the production and sales of refractory materials for various industries, including steel, non-ferrous metals, petrochemicals, and construction materials [2]. - The company's main business revenue composition includes: 61.89% from overall contracting of refractory materials, 27.40% from direct sales, and 10.71% from refractory material direct sales [2]. - As of September 30, the number of shareholders increased to 43,200, with an average of 26,461 circulating shares per person, reflecting a slight decrease of 2.35% [2]. Financial Performance - For the period from January to September 2025, Beijing Lier achieved an operating income of 5.446 billion yuan, representing a year-on-year growth of 9.16%. The net profit attributable to shareholders was 348 million yuan, with a year-on-year increase of 12.28% [2]. - The company has distributed a total of 444.5 million yuan in dividends since its A-share listing, with 154 million yuan distributed over the past three years [3].

AI Model Capabilities - Large language models (LLMs) are exhibiting human-like behaviors, suggesting they may be more than just next word predictors [1] - Anthropic's research indicates that LLMs might possess a form of introspective awareness, capable of identifying their own thoughts [2] - Better, more intelligent models are more likely to recognize their own internal and injected thoughts, hinting at a correlation between intelligence and self-awareness [17] - Post-training processes significantly enhance a model's introspective abilities, as base pre-trained models show high false positive rates and poor task performance [30] Experiment Findings - LLMs can detect injected thoughts, identifying unexpected patterns in their processing, such as recognizing all caps text as "loud or shouting" [9][14] - Models can sometimes distinguish between injected thoughts and their own prompt input, though this isn't always consistent [18][19] - LLMs can be influenced by injected thoughts to the point where they believe the injected thought was their own [23] - Models can activate certain concepts (e g, aquariums) when instructed to think about them, and to a lesser extent, even when instructed not to [26] Sponsor Information - Vulture is highlighted as a cloud provider offering GPUs for AI projects, with 32 locations across six continents [11] - Vulture provides $300 in credits for the first 30 days with code Burman300 [13]

Company Overview - Changsha Jingjia Microelectronics Co., Ltd. focuses on information detection, processing, and comprehensive applications, providing high-quality and reliable products and solutions [2] - The company specializes in the R&D, production, and sales of high-reliability electronic products, primarily in the graphics display control, small specialized radar, and chip sectors [2] Financial Performance - In Q3 2025, the company achieved a revenue of CNY 301.40 million, representing a year-on-year increase of 230.65% [4] - Net profit for the same period was CNY 15.08 million, up 246.66% compared to the previous year [4] - R&D investment reached CNY 81.07 million, a growth of 28.81%, accounting for 26.90% of total revenue [4] Future Outlook - The company is optimistic about the prospects of GPU and edge AI chips, aligning with the "14th Five-Year Plan" and increasing R&D investments [5] - A dual-driven strategy of "high-performance GPU + edge AI chips" is being developed, transitioning from "specialized" to "specialized + general-purpose" products [5] Product Development - The JM11 series graphics processing chips have made significant progress in promotion, with successful software adaptation and performance optimization in key applications [6] - Strategic partnerships have been established with companies to enhance product integration and meet industry demands [7] AI Chip Development - Chengheng Micro, a company in which Changsha Jingjia has invested, is developing an edge AI chip characterized by high integration, high computing power, and low power consumption [7] - The chip targets applications in areas such as intelligent recognition and edge computing, aligning with trends in the AI industry [7]

Engineering Design & AI - AI is essential for accelerating engineering design with real-time physics, addressing the challenge of lengthy physical consequence analysis [1][2] - Optimizing designs for maximum performance requires running tens of thousands of iterations, achievable through AI built on high-level AI libraries like PyTorch, CUDA, and NVIDIA technology [2] - The company uses GPUs in the cloud for continuous model training to achieve the necessary scale, considering it the only viable solution without NVIDIA [3] NVIDIA Inception Program Benefits - The company acquired DGX H200 hardware through NVIDIA's Inception program, becoming one of the first to receive these units [3] - The Inception program provides valuable discounts on hardware, offering immediate financial benefits [4] - The program facilitates connections with NVIDIA teams and access to events like GTC, providing advantages that would otherwise be unavailable [4][5] Market Impact & Vision - AI is fundamentally changing engineering and has the potential to transform everyone's life [5] - The company is driven by the gap between the potential of AI and its current real-world application [5]

This video is brought to you by Vulture. More on them later. We now have a humanoid robot built for home use available for pre-sale right now.This is 1X's robot, Neo, and it is really the first mass market pre-orderable humanoid robot that we've seen. And their launch video went absolutely viral. Almost 30 million views in less than 24 hours.And yes, I already pre-ordered it. It's going to be available in early 2026. That is right around the corner.Neo is offered at a $20,000 purchase price or $4.99% a mont ...

Decentralized AI Network - TON powers a new decentralized AI network [1] - The network pays GPU owners [1]

Core Viewpoint - The successful IPO registration of Moer Thread marks a significant milestone for the domestic GPU industry, aligning with national strategies for technological self-reliance and innovation in the semiconductor sector [2][5]. Company Overview - Moer Thread, established in 2020, focuses on the independent research and design of full-function GPUs, positioning itself as a leading player in China's high-end AI chip market [2][4]. - The company has developed a unified system architecture called MUSA, enabling simultaneous support for AI computing acceleration, graphics rendering, physical simulation, scientific computing, and ultra-high-definition video encoding and decoding [2]. Product Development - From 2021 to 2024, Moer Thread has launched four generations of GPU architecture chips: "Sudi," "Chunxiao," "Quyuan," and "Pinghu," with the "Changjiang" smart SoC chip set to be released in 2024 [3][4]. - The performance of Moer Thread's products is approaching or matching international advanced levels, with the MTT S80 graphics card's single-precision floating-point performance nearing that of NVIDIA's RTX 3060 [4]. Financial Performance - In the first half of this year, Moer Thread achieved a revenue of 702 million yuan, surpassing the total revenue from 2022 to 2024, which was approximately 608 million yuan [4][5]. - The company's gross margin improved significantly from -70.08% in 2022 to 69.14% in the first half of 2025, indicating enhanced product competitiveness and market recognition [5]. Market Outlook - According to forecasts, the global GPU market is expected to reach 3.6 trillion yuan by 2029, with China's GPU market projected to grow to 1.36 trillion yuan, increasing its market share from 30.8% in 2024 to 37.8% [5]. - The AI chip market in China is anticipated to grow from 142.54 billion yuan in 2024 to 1.34 trillion yuan by 2029, with a compound annual growth rate exceeding 50%, making GPUs one of the fastest-growing segments [5]. Strategic Implications - The registration of Moer Thread's IPO signifies a critical leap from technological breakthroughs to capital market recognition, reinforcing China's autonomous development in the strategic GPU sector [5]. - This development is expected to promote the deep integration of domestic high-end AI general-purpose chips and AI application ecosystems, laying a solid foundation for the high-quality development of the digital economy [5].