Hacker News Top 10
- English Edition

Published on December 28, 2025 at 18:01 CET (UTC+1)

1. Growing up in “404 Not Found”: China's nuclear city in the Gobi Desert (367 points by Vincent_Yan404)

   This article explores life in a secret Chinese nuclear city, built in the Gobi Desert during the Cold War and intentionally omitted from maps. It discusses the experiences of people who grew up in this isolated, closed community, highlighting the contrast between its secretive military purpose and the ordinary lives of its residents. The piece uses the "404 Not Found" internet error as a metaphor for the city's historical invisibility.

2. Calendar (702 points by twapi)

   This is a simple, elegant web tool that generates a minimalist, printable calendar for any given year (shown for 2026). It displays the entire year on a single page, designed to be folded and carried. The creator emphasizes practicality and a philosophy of kindness, offering a physical, at-a-glance alternative to digital calendar management.

3. Replacing JavaScript with Just HTML (541 points by soheilpro)

   The article argues for replacing common JavaScript-driven web components (like accordions, modals, and filters) with native HTML elements and CSS where possible. It provides specific code examples, such as using <details> and <summary> for accordions, to reduce JavaScript dependency. This approach aims to improve performance, reduce download size, and free up JavaScript for more complex tasks only it can handle.

4. Building a macOS app to know when my Mac is thermal throttling (66 points by angristan)

   The article details the creation of a custom macOS application designed to monitor and alert the user when the Mac's CPU is undergoing thermal throttling. It explains the motivation (performance concerns) and likely delves into the technical process of accessing system-level thermal data and building a native macOS menu bar or notification utility.

5. Hungry Fat Cells Could Someday Starve Cancer to Death (68 points by mrtnmrtn)

   Researchers at UCSF have developed a novel cancer treatment approach using CRISPR to engineer ordinary white fat cells into "beige" fat cells, which consume large amounts of nutrients. When implanted near tumors in mice, these hungry cells out-compete the cancer for resources, effectively starving the tumors. This bio-inspired method leverages existing plastic surgery techniques and presents a potential new form of cellular therapy.

6. One year of keeping a tada list (104 points by egonschiele)

   The author reflects on one year of maintaining a "tada list" or "to-done list," a journal where they record daily accomplishments instead of future tasks. They share the format, which includes monthly summaries with drawings, and discuss the pros (better recognition of progress, combating anxiety) and cons (potential for guilt, feeling like a chore) of the practice for personal productivity and mental framing.

7. Designing Predictable LLM-Verifier Systems for Formal Method Guarantee (6 points by PaulHoule)

   This academic paper presents a formal framework with provable guarantees for systems that combine Large Language Models (LLMs) with automated software verification tools. It introduces a convergence theorem, modeling the process as a Markov chain, to ensure these LLM-Verifier pipelines will eventually terminate successfully, addressing current issues of unreliability and lack of theoretical foundation in AI-assisted formal methods.

8. Floor796 (869 points by krtkush)

   "Floor796" is an interactive, animated homage to internet and pop culture, presented as a seemingly endless pixel art animation of a large office floor. It contains hundreds of references to movies, games, memes, and tech history, with characters and scenes interacting in a continuous, scrolling view. It is a creative and dense visual project celebrating digital nostalgia.

9. Rex is a safe kernel extension framework that allows Rust in the place of eBPF (101 points by zdw)

   Rex is an open-source framework that allows developers to write safe kernel extensions in Rust as an alternative to using eBPF (Extended Berkeley Packet Filter). It aims to provide the safety and expressiveness of Rust for kernel-level programming, enabling more complex and performant in-kernel functionality than eBPF typically allows, while maintaining security guarantees against crashes and memory errors.

10. How we lost communication to entertainment (562 points by 8organicbits)

    The article argues that modern digital platforms (like social media) have fundamentally shifted from communication tools to entertainment distribution networks. This paradigm clash explains conflicts within decentralized networks like the Fediverse, where older users prioritize reliable message delivery (communication) while newer platforms prioritize content consumption and algorithmic curation (entertainment), often at the cost of genuine human connection.

AI/ML Insights & Trends

1. Trend: Formal Guarantees for AI-Assisted Systems
2. Why it matters: As LLMs are integrated into critical technical workflows (like code verification, theorem proving, or system design), their inherent unpredictability becomes a major risk. The research in Article 7 moves beyond empirical testing to provide mathematical proofs of system behavior (e.g., termination guarantees).

3. Implication: This signifies a maturation phase where AI/ML engineering must adopt formal methods from traditional computer science. It will be crucial for deploying AI in safety-critical domains like aerospace, finance, and infrastructure code generation.

4. Trend: AI Driving Hardware-Aware Software Optimization

5. Why it matters: The pursuit of efficiency, highlighted by the JavaScript-replacement (Article 3) and thermal throttling monitor (Article 4), is directly relevant to AI. Efficient client-side code reduces load on devices running local AI models, while understanding thermal constraints is vital for on-device ML inference performance and hardware longevity.

6. Implication: ML developers will need deeper system-level knowledge. The trend points toward co-design of lightweight AI models and the extremely efficient software stacks they run on, especially for edge and mobile AI.

7. Trend: AI as a Catalyst in Synthetic Biology & Cellular Engineering

8. Why it matters: Article 5's cancer treatment, powered by CRISPR, exemplifies a field where AI/ML is already indispensable. AI models are used to predict protein structures, design genetic sequences, and simulate biological pathways. This bio-engineering approach to therapy relies heavily on data-driven design and optimization.

9. Implication: AI's role in life sciences is shifting from analysis to direct design and fabrication of biological systems. This accelerates the pipeline from concept to cellular therapy, with ML models acting as the primary design tool for novel living treatments.

10. Trend: The Infrastructure Shift for AI Workloads (Rust & Kernel-Level Code)

11. Why it matters: Article 9's Rust-based kernel framework reflects a broader infrastructure trend toward performance, safety, and control. High-performance AI data pipelines, custom hardware orchestration, and network routing for distributed training all benefit from safe, low-level systems programming, moving beyond scripting languages and even C/C++.

12. Implication: The AI infrastructure stack is deepening. Knowledge of systems programming languages like Rust will become increasingly valuable for building the next generation of high-efficiency, secure ML deployment platforms and data processing engines.

13. Trend: The Data Quality Crisis Fueled by "Entertainment-First" Platforms

14. Why it matters: Article 10's critique highlights that much of the public data generated online is optimized for engagement, not truthful communication or reliable information. This creates a significant problem for training and fine-tuning LLMs, as they ingest data skewed by entertainment algorithms, potentially amplifying bias, misinformation, and sensationalism.

15. Implication: Reliance on open-web data for AI training is becoming riskier. There will be a growing premium on curated, high-quality, and verifiable datasets, as well as AI alignment techniques that can discern communicative intent from entertainment-driven content. This also pushes value toward decentralized or private communication data that may be more genuine.

16. Trend: Human-AI Interaction and Cognitive Ergonomics

17. Why it matters: Articles 6 (tada list) and 2 (simple calendar) reflect a human need for simplicity, clarity, and a sense of accomplishment. As AI tools become more complex and autonomous, managing user cognitive load, providing transparent progress tracking, and designing for human psychological needs (like celebrating "wins") become critical for adoption.

18. Implication: AI product design must incorporate principles from behavioral science and cognitive ergonomics. The focus will expand from raw model capability to the entire user experience loop, ensuring AI assistants and tools reduce anxiety and enhance human agency, not just raw productivity.

19. Trend: Generative AI for Creativity and Cultural Archiving

20. Why it matters: Article 8 (Floor796) is a massive, human-made tapestry of digital culture. This reflects a space where generative AI is both a tool for creating such content and a subject of its references. It shows the deep intertwining of AI with contemporary creativity and the preservation/remixing of cultural memory.
21. Implication: AI is becoming a central medium for artistic expression and historical documentation. This leads to new research areas in AI for style transfer, cultural pattern recognition, and the creation of interactive, generative worlds. It also raises important questions about copyright, authorship, and the algorithmic curation of our collective culture.

---

Analysis generated by deepseek-reasoner

Deutsch