Core Viewpoint - The article discusses the impact of U.S. export controls on NVIDIA's operations in China, highlighting the significant loss of market share and potential revenue due to legislative changes and restrictions on AI chip sales [1][2]. Group 1: Legislative Impact - The U.S. Congress is pushing for legislation that requires chip manufacturers like AMD and NVIDIA to prioritize domestic supply for AI chip products [1]. - NVIDIA's CEO, Jensen Huang, stated that due to U.S. export controls, the company has completely exited the Chinese market, with its market share dropping from 95% to 0% [1]. Group 2: Financial Consequences - Following the U.S. export controls initiated in October 2022, NVIDIA faced ongoing challenges in selling high-end AI chips in China [1]. - The introduction of compliant AI chips for China, such as A800, H800, and H20, was not sufficient to mitigate losses, as new regulations in April 2025 forced the discontinuation of the H20 chip, resulting in approximately $4.5 billion in inventory losses and $8 billion in potential revenue losses [1]. Group 3: Future Outlook - Despite the challenges, NVIDIA remains hopeful for a policy change and is actively communicating with the U.S. government regarding its operations in China [2]. - Huang emphasized that the U.S. has lost access to one of the largest markets globally, suggesting that policies aimed at China could also harm U.S. interests [2]. - He noted that China possesses about 50% of the world's AI researchers and has a strong focus on AI development, indicating the competitive landscape [2].

Core Viewpoint - The article discusses a significant coding incident in a company, where a physicist's code, despite being assisted by AI, led to a cost of over $600,000 in a single day due to the use of an unfamiliar legacy tech stack, highlighting the risks of non-computer science professionals entering programming without adequate engineering knowledge [5][11]. Group 1 - A coding accident occurred in a company, resulting in a cost of over $600,000 in one day, which is more than the typical funding of many startups [5]. - The physicist involved used an outdated technology stack that the team was not familiar with, leading to a failure in code review and oversight [7]. - Following the incident, the company took immediate action to halt all tasks and negotiate discounts with related companies, while the manager of the physicist's team was dismissed [8]. Group 2 - The incident serves as a cautionary tale about the potential pitfalls of non-computer science individuals entering programming roles, even with AI assistance, due to a lack of fundamental engineering skills [11]. - The company culture encourages learning from failures and embracing risk and innovation, as indicated by the leadership's response to the incident [8].

Core Viewpoint - Microsoft has officially closed the security update channel for Windows 10, impacting millions of users globally, while simultaneously tightening upgrade restrictions for Windows 11, leaving many older device users facing limitations or forced obsolescence [1][5]. Summary by Sections Windows 10 Support Termination - Microsoft has ceased support for Windows 10, meaning no more security updates or bug fixes will be provided, significantly increasing the risk of cyberattacks [5]. - Approximately 42% of Windows devices globally are still running Windows 10, with 36.38% in the U.S. alone [5]. - Nearly 43% of these devices cannot upgrade to Windows 11 due to high hardware requirements, such as a 64-bit processor, 64GB storage, 720p display, and TPM 2.0 support, making many devices from 2017 and earlier unable to upgrade [5]. Windows 11 Upgrade Restrictions - Microsoft has removed the known mechanism for creating local accounts during the Windows 11 installation process, requiring users to be online and log in with a Microsoft account to complete the setup [4]. - This change is seen as a significant shift from previous policies that allowed local account creation, which could mislead users into skipping important setup screens [4]. - Although there are alternative methods for experienced users to perform a local installation, these methods are not accessible to the average user [4]. Economic and Environmental Impact - The termination of Windows 10 support could lead to up to 400 million computers becoming vulnerable to attacks or being discarded as electronic waste, with estimates suggesting over 1 million pounds of e-waste generated in the U.S. alone [5][6]. - The cost of upgrading to Windows 11 is a burden for many consumers and small businesses, with entry-level devices starting around $300 and high-performance models exceeding $3,000 [6]. - The current economic climate, with the U.S. unemployment rate at a near four-year high of 4.3%, exacerbates the financial strain on low-income families and unemployed individuals [6]. - Critics argue that Microsoft's actions contradict its previous commitments to achieve carbon-negative operations and zero waste by 2030, as the discontinuation of Windows 10 support will lead to significant electronic waste [6].

Core Viewpoint - Tata Consultancy Services (TCS) has executed one of the largest layoffs in the Indian tech industry, reducing nearly 20,000 jobs in just three months, marking the largest decline in employee numbers in the company's history [1][2]. Group 1: Layoff Details - The total number of employees at TCS fell from 613,069 at the end of June to 593,314 by September 30, 2025, resulting in a net reduction of 19,755 employees [1]. - This is the first time since 2022 that TCS's employee count has dropped below 600,000 [2]. - The layoffs primarily affected mid-level and senior employees as part of a restructuring strategy aimed at aligning the company with future technological developments [3]. Group 2: Reasons Behind Layoffs - The layoffs are attributed to a "skills mismatch," with TCS's new HR director stating that the layoffs resulted from both voluntary and involuntary departures, with approximately 6,000 being involuntary [3]. - TCS has set a broader plan to reduce its global workforce by 2% by March 2026, with half of this plan already completed [3]. Group 3: Financial Implications - TCS has allocated 11.35 billion rupees for severance payments and other restructuring-related costs, highlighting the scale and seriousness of the company's transformation from traditional outsourcing to high-value digital services and AI-driven offerings [4]. - The actual number of layoffs exceeded initial expectations, which estimated around 12,000 job cuts, indicating a 66% increase over the forecasted figure [4]. Group 4: Industry Context - The layoffs occurred amid challenging global conditions, including a slowdown in traditional IT service demand, tightening tech budgets, and increased automation, prompting IT giants to reassess long-standing operational models [5]. - The uncertainty in global relations, particularly between the U.S. and India, has further exacerbated industry concerns, with proposed policy changes potentially impacting TCS's reliance on U.S. clients [5]. Group 5: Transparency and Employee Impact - The National Association of IT Employees (NITES) has accused TCS of lacking transparency regarding the actual number of layoffs, claiming that nearly 8,000 employees have disappeared from the roster beyond what TCS has acknowledged [6]. - Despite a slight decrease in voluntary resignation rates, the significant layoffs overshadow this positive trend, raising concerns about the overall employment landscape in the industry [7]. Group 6: Broader Industry Trends - The phenomenon of "invisible layoffs" is spreading across the $280 billion Indian IT industry, where companies are not formally announcing layoffs but are instead encouraging employees to leave through various means [8]. - Labor rights advocates are calling for increased employee protections and transparency in layoffs, as the current restructuring methods leave employees with limited avenues for support [8].

Core Insights - OpenAI's CEO Sam Altman announced on October 15, 2025, that ChatGPT will ease restrictions previously imposed to prevent mental health issues, which had negatively impacted user experience [3] - A new version of ChatGPT is set to launch in a few weeks, featuring a more human-like personality and the ability to interact with users in a friendlier manner, including the use of emojis [3] - OpenAI plans to implement age restrictions in December, allowing for more content, including adult content for verified adults [3] User Reactions - Some users expressed concerns about children accessing adult content through shared accounts, highlighting the need for parental supervision [5] - Others countered that children can easily access adult websites independently, emphasizing that parental responsibility is crucial [5] - Suggestions were made for creating separate child accounts to allow parents to maintain control over content access [5]

调侃留言 往期趣图 （点击下方图片可跳转阅读） 走心 留言 ...

大家好，我是小程程 程序员忙起来一整天都不回消息？ 最近，一位四川小姐姐的求助在网上引发了不少讨论。 她正在和大厂程序员接触，可让她困惑的是，这位程序员整个白天都不主动发信息，只有到了晚上下班之后才 会联系她，她不确定这种情况是否正常。 这个问题一抛出，网友们纷纷留言，各抒己见，看法差异还挺大。 在众多讨论中，有一条留言获得了 900 多个赞，还有不少网友跟帖回应，这条留言是：" 可能是怕了回了就一 来一回聊起来了 "。跟帖的网友们纷纷表示："笑死这个很真实"，"完全是我了.... 我可能白天看到了，但是晚上 才会集中回复"，"哈哈。我工作忙的时候也是这么想的"。 但随着讨论深入，她也补充了更多藏在"求助"背后的真实处境： 她并非闲着没事等消息，自己也有工作要忙， 甚至因为怕影响对方，白天都不太主动发信息。对方下班后果真会秒回，还会一直陪到她睡觉，见面时也耐心 体贴。可即便如此，她还是绕不开心里的疙瘩 —— 自己情感需求比较高，忙里偷闲时会和别人聊聊，所以总 觉得"一整个白天都没空" 有点离谱，"正常人不都有摸鱼、吃饭、上厕所的时间吗？" 他两没确定关系，只是在接触阶段，她不想过度猜忌，却又忍不住困惑：这种" ...

Core Insights - The article discusses the launch of "nanochat," a simplified version of ChatGPT created by Andrej Karpathy, which can be built with minimal resources and code [1][2][4]. - The project aims to provide an accessible framework for training language models, emphasizing ease of use and modification [11][13]. Project Overview - "Nanochat" is a full-stack training and inference pipeline that allows users to create a basic ChatGPT-like model with approximately 8000 lines of code [2][4]. - The total cost to train this model is around $100, using a cloud GPU server for about 4 hours [4][16]. - The model is built using Rust and includes a custom tokenizer, with training conducted on the FineWeb dataset [5][19]. Performance Metrics - After approximately 12 hours of training, the model's performance on the CORE metric surpasses that of GPT-2 [8]. - Specific performance metrics include: - CORE: 0.2219 - ARC-Easy: 0.3876 - GSM8K: 0.0758 - HumanEval: 0.0854 - MMLU: 0.3151 [7][56]. Training Process - The training process involves several stages: pre-training, mid-training, supervised fine-tuning (SFT), and reinforcement learning (RL) [45][50]. - The pre-training phase utilizes a large dataset to teach the model about the world, while mid-training focuses on adapting the model for conversational tasks [28][45]. - The SFT phase further refines the model using high-quality dialogue data [48]. Community Engagement - The project has gained significant attention, with over 4.8k stars on GitHub shortly after its release, indicating strong community interest [14]. - The framework is designed to be easily modifiable, allowing users to experiment with different parameters and configurations [59]. Future Potential - Karpathy envisions "nanochat" evolving into a research tool or benchmark framework, similar to previous projects like nanoGPT [13]. - The project is still in its early stages, with potential for further optimization and enhancement [13][50].

Group 1 - The article discusses a peculiar incident where a candidate received a job interview email from Ant Group at 1 AM, raising concerns about the timing and potential errors in the system [5][6]. - The timing of the email, which is 8 hours before a typical 9 AM work start, suggests a possible bug in the scheduling system [5]. - The situation sparked humorous comments from netizens, indicating that the company may be looking for candidates who can work late hours, reflecting a culture of long working hours at Alibaba [6]. Group 2 - The article highlights the reactions from other users who also received similar emails, emphasizing the unusual nature of the timing [5]. - There is a mention of the broader implications for work culture in tech companies, particularly regarding expectations for employees to be available outside of regular hours [6].

转自：CSDN（ID：CSDNnews） 最近 Python 软件基金会（PSF）正式发布了 Python 3.14.0 。这是一次里程碑式的更新，带来了多个开发者期待已久的新特性，包括 自由线程（支 持 、 子解释器（Subinterpreters）引入 、 结构化模式匹配增强 ，以及对 异步编程机制 的深度优化等。 从功能上看，3.14 的变化可谓诚意十足。但许多开发者最关心的，依然是那句老问题——"它到底有多快？" 为此，软件工程师 Miguel 对新版 Python 进行了一轮非官方性能测试，希望看看这次升级在速度上究竟能带来多大提升。 不过，他也在测试开头就打了"预防针"： "在我展示结果之前，必须提醒大家，这类基准测试（benchmark）其实很容易产生误导。运行这些测试当然很有趣，但要想仅靠几个简单脚本就准确 描绘出像 Python 解释器这样复杂系统的性能全貌，几乎是不可能的。" 接下来，我们将看看完 整的测试。 测试矩阵 这次的基准测试涵盖了一个五维度的测试矩阵，具体如下： Miguel 表示，他设计的测试 全部运行纯 Python 代码 ，刻意避免任何第三方依赖，尤其排除了所有用 C 编写 ...