Articles

The article reveals the severe executive turbulence Apple has recently faced. Johny Srouji, Senior Vice President of Hardware Technologies and a key figure in Apple Silicon, has informed Cook of his intention to resign, becoming the fourth executive whose future is uncertain this month. Previously, Apple's Head of AI John Giannandrea and Chief UI Designer Alan Dye had already departed, with the latter moving to Meta. The article details Srouji's core contributions to the Apple Silicon project and notes that Cook is actively trying to retain him, even considering appointing him as CTO. Furthermore, the article mentions that power within Apple is concentrating among four key executives, and that the 65-year-old Cook may have health concerns, making the question 'Who will be Apple's next CEO?' a hot topic. Overall, Apple is experiencing its largest executive reshuffle in decades, posing challenges to the company's future development.

This issue of JavaScript China Weekly celebrates the 30th anniversary of JavaScript's birth, reviewing the language's widespread application in desktop, mobile, embedded fields, and on the Web. The weekly highlights the development progress of TypeScript 7, with the new version being rewritten in Go to achieve a 10x speed increase. It also reports on Anthropic's acquisition of the Bun JavaScript runtime and its commitment to continue open-source development. In addition, Vite 8 Beta was released. The 'News' section covers security vulnerabilities in React Server Components, full support for WebGPU in major browsers, among other things; the 'Articles and Videos' section discusses technical articles on npm security practices, JSDoc type system applications, browser Base64 data processing performance, Web Audio API applications, etc.; the 'Code and Tools' section introduces new tools such as TanStack AI, Remend, Tinybench; the 'Other Interesting Content in the Ecosystem' and 'Releases' sections list multiple framework and library version updates such as GitHub Spec Kit, Web performance review, and ESLint v10, Prisma v7.1, offering developers comprehensive industry trends and technical insights.

This article presents CKDA, a new Visible-Infrared Lifelong Person Re-identification method from Peking University's Peng Yuxin team at AAAI 2026. CKDA continuously learns new person discriminative information while maintaining recognition of known data, offering significant value in applications like public security surveillance. Existing methods often hinder the retention of common cross-modal old knowledge when learning new modality-specific knowledge, leading to knowledge conflicts. CKDA explicitly decouples and purifies common and modality-specific discriminative information using a cross-modal general prompt module and a single-modal dedicated prompt module to avoid interference. Subsequently, the decoupled old and new knowledge is aligned in independent intra-modality and inter-modality feature spaces, efficiently balancing cross-modal knowledge. Experiments demonstrate that CKDA achieves state-of-the-art performance on multiple benchmark datasets.

This article summarizes Ren Zhengfei's discussion with the ICPC chairman, coaches, and contestants on AI and education. Ren Zhengfei highlighted the profound changes AI will bring, predicting significant efficiency improvements in industries like manufacturing and healthcare within 3-5 years. He stressed that AI's true value lies in industrial applications, not just conceptual inventions, and anticipated that computing power will become abundant, with the key being the transformation of Large Language Models (LLMs) into tools for various sectors. Regarding education, he believes AI will promote equitable distribution of educational resources, enabling access to high-quality education even in remote areas. He encouraged young people to aim high, embrace experimentation, and not fear failure, emphasizing that the purpose of education is learning, and the purpose of enterprises is business. Finally, he urged China to continue opening up, absorbing global knowledge and talent, and promoting high-quality development through Innovative Productive Forces.

The article analyzes in detail the entire process of the fresh yogurt industry from explosive growth to rapid decline around 2023. In the early stages, brands such as Blueglass and Moyou Yogurt attracted a large number of middle-class consumers and gained capital favor by shaping a "healthy" positioning and mid-to-high-end price range, leading to a surge in the number of stores and even the myth of "million-yuan monthly income" for franchisees. However, the period of prosperity was short-lived. As the "healthy" persona was exposed for using non-dairy creamer, high sugar, and high calories, consumer trust quickly collapsed. At the same time, the discrepancy between high price and actual value, severe product homogenization, inherent limitations of the category (such as being affected by seasons, limited consumption scenarios), and a fragile supply chain system, all contributed to the rapid cooling of the industry. The article points out that leading brands such as Moyou Yogurt, Blueglass, and Yizhi Yogurt Cow are all facing the dilemma of a large number of store closures, with franchisees bearing the brunt. Finally, the article emphasizes that the industry is undergoing a value regression, and companies need to balance between "delicious" and "cost-effectiveness" and build strong supply chain barriers to survive the reshuffle.

The article uses "10 snippets" to outline the significant achievements of DeepMind, an Artificial Intelligence company. It covers key milestones from its founding vision to AlphaGo's victory over the Go World Champion and AlphaFold's solution to the protein folding problem. The article emphasizes DeepMind's pursuit of Artificial General Intelligence (AGI) and how its groundbreaking research has transformed science and human society. Through these snippets, readers can understand DeepMind's reflections and practices in AI ethics, scientific exploration, and technological application, inspiring imagination about the future development of AI and fostering appreciation for their rigorous approach.

This article, presented as 'Today's Open Source,' introduces six cutting-edge AI projects, categorizing them into foundation models and framework tools. Foundation models include: LLaVA-UHD v3, co-developed by Tsinghua University and the Chinese Academy of Sciences (CAS), which leverages Progressive Visual Compression (PVC) for efficient native resolution encoding, performing comparably to mainstream models while significantly reducing Time to First Token (TTFT); NetraEmbed, a multilingual multimodal embedding model focused on visual document retrieval, supporting efficient similarity search through Matryoshka representation learning; and GLM-4.6V, a model that achieves state-of-the-art performance in visual comprehension for its parameter scale and is the first to integrate native function calling capability, providing a unified technical foundation for multimodal agents, supporting high-quality mixed-media creation and multi-document input. Framework tools featured are the Light-X video generation framework, the PATN privacy protection framework, and the VieNeu-TTS Vietnamese Text-to-Speech model. While comprehensive in scope, the article provides relatively brief introductions to each project.

The article deeply analyzes the profound influence of Silicon Valley tech elites, represented by the "PayPal Mafia," on American politics and national strategy. It sets the stage with a fictional "Genesis" AI project, indicating that the policy focus in the Trump 2.0 era has shifted towards the assertions of tech capital and elites. Specifically, core tech figures like Peter Thiel, David Sacks, and Alexander Karp, through funding political agents, overseeing policymaking, and promoting collaboration between businesses and defense agencies, elevate the "technological accelerationism" ideology to the national strategic level. They advocate for the government to adopt an "engineering mindset," viewing AI and other deep technologies as the cornerstone of national strength, unburdened by the constraints of "diversity and inclusion" and "political correctness." The article also references the book "The Technology Republic," emphasizing that the U.S. should shape the international power balance through the development of AI weapon systems, and warns that tech elites need to return to the national mission. Ultimately, the article points out that this interest-based alliance is attempting to transform the United States into a "Tech Republic" sustained by deep technology rather than financial hegemony.

This article introduces `hl`, an open-source log viewer built in Rust, specifically designed for efficient processing of structured logs in JSON or logfmt format. It highlights `hl`'s exceptional performance, boasting parsing speeds of up to ~2 GiB/s with automatic indexing and ~10 GiB/s for reindexing growing files, significantly outperforming existing alternatives. The tool is presented as highly valuable for DevOps engineers managing very large log files from the command line, capable of handling hundreds of local files totaling hundreds of gigabytes. Key features include an integrated pager, comprehensive filtering by log level, field, and time ranges, a `tail -f`-like follow mode, customizable output, automatic timezone switching, and support for UI themes. A standout capability is its ability to construct complex queries using logical and comparison operators, including explicit handling of missing fields. The article also references a Hacker News discussion where `hl` was suggested as a simpler alternative to more complex tools like Grafana Loki for single-instance applications. `hl` is available under the MIT license for macOS, Linux, and Windows.

Vitest, the Vite-native testing framework, has released version 4.0, bringing significant advancements in testing capabilities. Key features include the stabilization of Browser Mode, allowing tests to run in real browsers for increased accuracy, and integrated visual regression testing via `toMatchScreenshot` for detecting UI changes. The update also adds Playwright Traces integration, improved debugging with a VSCode extension, and refined reporter behaviors. While praised for its developer experience and performance, a minor TypeScript compatibility issue with older Vite versions has been noted. The release includes breaking changes, with a detailed migration guide available.