1. I still like Sublime Text
Total comment counts : 138
Summary
The article discusses why the author continues to use Sublime Text as their preferred text editor in 2025, despite the availability of many modern alternatives. Here are the key points:
Historical Use: The author has been using Sublime Text since around 2010-2011, initially due to its speed, cross-platform compatibility, and portability. Over the years, it has become ingrained in their workflow.
Performance: Sublime Text is praised for its speed, low resource usage, ability to handle large files, and stability.
Modern Features:
- LSP (Language Server Protocol): Sublime Text has adopted LSP, enhancing its capabilities with features like code-aware completion and hover information, similar to what’s available in more contemporary editors like VS Code.
- Snippets, Project Workspaces, Build Systems, Multiple Cursors, and Key/Mouse Bindings: These features make development and coding tasks more efficient.
Updates and Community Support: Despite its age, Sublime Text is still actively updated, with the latest build mentioned being from January 2025. The community contributes through plugins like Package Control, which is crucial for extending the editor’s capabilities.
Customization and Extensibility: The editor allows for deep customization, especially with LSP settings, enabling users to configure servers on a per-project or per-file basis, enhancing its utility for different programming needs.
Legacy and Influence: Sublime Text has influenced many modern editors, suggesting that its foundational concepts are still relevant and appreciated.
In summary, the author remains a loyal user due to Sublime Text’s enduring performance, comprehensive feature set, and the extensive customization options that keep it relevant in the evolving landscape of text editors.
Top 1 Comment Summary
The article expresses the author’s fondness for Sublime Text, a versatile text editor they use continuously across all their devices for personal notes, blog posts, and writing books. However, for professional programming, the author prefers using specific IDEs tailored to the language they are working with, such as Visual Studio or IntelliJ. They appreciate Sublime Text for being lightweight yet feature-rich, ideal for managing documents and small coding tasks without the bulkiness of full-fledged IDEs.
Top 2 Comment Summary
The developer of Sublime Text expresses gratitude for the appreciation received and looks forward to future achievements in the current year, offering to answer any questions from the community.
2. OpenAI Furious DeepSeek Might Have Stolen All the Data OpenAI Stole from Us
Total comment counts : 4
Summary
The article discusses accusations by OpenAI, Microsoft, and AI czar David Sacks against the Chinese AI startup DeepSeek. They claim that DeepSeek developed its advanced large language model, R1, by unfairly using data from OpenAI’s models, possibly through a technique called distillation where one AI model learns from another by asking it numerous questions. This practice, according to Sacks, effectively allows the student model to mimic and extract knowledge from the parent model, which he alleges DeepSeek did with OpenAI’s technology. The irony highlighted in the article is that OpenAI itself has been criticized for similar practices, obtaining vast amounts of data from the internet often without explicit permission or in violation of terms of service, yet now complains when it perceives the same methods being used against them. This situation underscores a double standard in the AI industry regarding data usage ethics.
Top 1 Comment Summary
The article provides guidance on submitting content to Hacker News, emphasizing the importance of sharing original sources rather than secondary reports or sensationalized versions of stories. It advises users to avoid submitting “knock-off” articles that increase linkbait and indignation, as Hacker News aims to be a platform for more substantive and less sensationalized content. The comments related to the article have been moved to a specific thread on Hacker News for further discussion.
Top 2 Comment Summary
The article criticizes clickbait journalism for sensationalizing corporate disputes. It mentions that Microsoft and OpenAI are investigating whether DeepSeek, another AI company, improperly used their technology to train its R1 model, but clarifies that OpenAI’s reaction is not one of anger but a calm verification process regarding potential agreement violations.
3. Composable SQL
Total comment counts : 35
Summary
The article discusses the limitations of SQL in handling complex queries and business logic, particularly focusing on two major issues:
Testing SQL Queries: Testing SQL is inherently difficult because even simple queries require a vast amount of test data. Each query depends on multiple columns and rows, necessitating the population of an entire object graph, which becomes exponentially complex with joins and deep queries. This leads to:
- Uneven test coverage where simpler queries are tested but complex ones are not.
- High maintenance cost due to the need for extensive test fixtures and the difficulty in maintaining and understanding these tests.
Reusing Business Logic: SQL’s lack of abstraction means business logic must often be replicated across multiple queries or denormalized, leading to:
- Duplication of Logic: Each query might need to include logic for computed properties, increasing the chance of inconsistencies as the logic evolves.
- Denormalization: To manage this, one might denormalize data by pre-computing values, but this introduces new challenges like ensuring data consistency across the system.
The author proposes a solution involving composable query fragments with statically-typed interfaces to address these issues:
- This would allow for the definition of logic once, making it easier to test and reuse across different queries.
- It aims to reduce the need for duplicating logic and simplify testing by focusing on smaller, reusable components rather than entire object graphs.
Overall, the article critiques SQL’s current capabilities in handling modern data needs and suggests an enhancement to make SQL more modular and testable, thereby improving both the development and maintenance of database-driven applications.
Top 1 Comment Summary
The article discusses the limitations and challenges of using views in PostgreSQL for query optimization. The author explains that ideally, PostgreSQL would inline views into a single, highly optimized query, but this often does not occur effectively. Here are the key points:
PostgreSQL’s View Handling: The database does not always manage to inline and optimize views into a “gigaquery” as hoped, leading to suboptimal query performance.
Comparison with Other Databases: Unlike PostgreSQL, databases like Oracle and SQL Server are better at inlining views and optimizing the resulting query.
Issue with Views: Views are inherently designed for general use, which means they cannot be parameterized, leading to inefficiencies when specific filtering is required post-view creation.
SQL Server’s Solution: SQL Server offers inline table-valued functions which act like parameterizable views, although they have their own limitations in terms of composability. For instance, direct composition of functions that return and accept tables isn’t straightforward, often requiring intermediate storage.
Utility Despite Limitations: Despite these issues, functions in SQL Server are highlighted as very useful for building complex queries that reuse logic.
Overall, the article points out a significant area where PostgreSQL could improve in handling views and suggests that while SQL Server provides a better alternative, it still has its own set of challenges in function composition.
Top 2 Comment Summary
The article discusses the inefficiencies associated with Common Table Expressions (CTEs) in PostgreSQL, particularly focusing on the lack of predicate pushdown which affects performance. Here are the key points:
PostgreSQL’s CTE Limitations: The article critiques PostgreSQL for not effectively pushing down predicates in CTEs, leading to performance issues.
Comparison with BigQuery: The author contrasts PostgreSQL with Google BigQuery, where CTEs benefit from predicate pushdown, leading to predictable performance improvements. BigQuery also supports user-defined table functions, though these are not commonly used.
Desired Features: The author expresses a desire for better scoping capabilities in SQL, akin to how CTEs are structured, but also allowing for inline function definitions within SQL statements. This would enhance readability and manageability, especially in environments where separate function definitions are cumbersome.
BigQuery’s Additional Features: BigQuery supports user-defined aggregate functions, which, while not often necessary, are highly beneficial when needed.
Call for Feedback: The author invites feedback from BigQuery developers, indicating a willingness to share insights on user experiences and potential improvements.
Overall, the article highlights a need for enhancements in how databases handle CTEs and function definitions to improve efficiency and user experience.
4. Science YouTuber physicsgirl (Dianna Cowern) stands for the first time in 2 yrs
Total comment counts : 38
Summary
error
Top 1 Comment Summary
The article is a personal account by a 22-year-old who has been dealing with Long COVID for three years. The individual shares their ongoing struggle with the condition, which has significantly altered their life, though they are not bedridden. They’ve documented their experiences on a blog and offer words of encouragement to others suffering from Long COVID and ME/CFS, urging them not to lose hope.
Top 2 Comment Summary
The article discusses the personal experience of the author’s sister who developed Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) following a burnout after the birth of her second child. The sister’s condition fluctuated, requiring periods of rest and recovery, although she was never permanently bedridden. At the time of her illness, ME/CFS was not widely recognized or accepted as a diagnosis, leading to challenges in getting her condition acknowledged. The author expresses hope for Dianna’s recovery, indicating a connection or comparison between Dianna’s situation and their sister’s.
5. What’s OAuth2, anyway?
Total comment counts : 28
Summary
Summary of the Article:
The article discusses OAuth2, a widely-used authorization framework that enables users to securely share access to their data across different services without exposing their credentials. Here are the key points:
OAuth2 Usage: Many users unknowingly utilize OAuth2 when they log into websites using their Google, Facebook, or GitHub accounts.
Simplicity for Users: OAuth2 is designed to be user-friendly, making integration between applications straightforward without requiring users to understand the underlying protocol.
Problem Solving: The article explores the scenario of integrating a user’s Git repositories into a deployment platform like Fly.io or Vercel. It discusses the initial, problematic solution of sharing user credentials and introduces alternatives like Personal Access Tokens (PATs).
Personal Access Tokens (PATs): PATs are an improvement over sharing main credentials as they can be limited in scope and lifespan, reducing security risks.
OAuth2 vs. PATs: While PATs simplify integration for users, OAuth2 offers a more automated and secure method for applications to request access to user data. OAuth2 focuses on delegation of access permissions, allowing users to control which applications can access their data and to what extent.
Design Philosophy: OAuth2 is built around the principle of empowering users to manage permissions for third-party applications, enhancing security and privacy by design.
The article concludes by highlighting the importance of OAuth2 in modern web services, where it streamlines the process of granting and managing access permissions, making it an integral part of the software ecosystem.
Top 1 Comment Summary
The article discusses the author’s interest in OAuth2 and OIDC (OpenID Connect) protocols, particularly enjoying an article titled “OAuth from First Principles” which explains the protocols in detail. The author appreciates how different explanations of these protocols can enhance understanding. However, in a critique section, the author points out a misuse of the term “Trust-on-first-use” (TOFU). They clarify that TOFU refers to a specific security scenario where initial trust is established without prior shared secrets, which can be vulnerable to man-in-the-middle (MitM) attacks, unlike the simple caching of authorization scopes as suggested in the article. The most common example of TOFU is described with SSH connections where keys are not initially verified.
Top 2 Comment Summary
The article expresses frustration with the complexities and perceived unnecessary limitations of authentication systems, particularly in web development. The author criticizes:
Complexity and Misimplementation: The author finds authentication overly complicated and often incorrectly implemented, citing examples like not checking JWT (JSON Web Token) audiences.
Lack of Understanding: Developers frequently use libraries without understanding the underlying security mechanisms, which leads to poor security practices.
User Experience (UX): From an application developer’s perspective, the author desires a simple process where a user provides a token from a trusted source to prove identity and permissions. The method of token acquisition (e.g., via redirect or user action) should not require special labels or separate authorization flows, as this adds unnecessary complexity.
The author essentially calls for a more straightforward, less jargon-filled approach to authentication that focuses on security and user ease rather than on convoluted procedures and terminology.
6. Bacteria (and their metabolites) and depression
Total comment counts : 17
Summary
error
Top 1 Comment Summary
The article from The Hill, titled “White House insists no quid pro quo in Trump call with Ukraine”, discusses the controversy surrounding a phone call between President Trump and Ukrainian President Volodymyr Zelensky. Here’s a summary:
Context: The call has come under scrutiny due to a whistleblower complaint suggesting Trump might have used the call to solicit foreign interference in the 2020 election by pressuring Ukraine to investigate Joe Biden, a potential election rival, and his son Hunter’s activities in Ukraine.
White House Stance: White House Press Secretary Stephanie Grisham stated there was “no quid pro quo” in the call. She emphasized that Trump was merely interested in rooting out corruption, a stance he has consistently maintained, and that discussing corruption in Ukraine was part of this broader initiative.
Transparency: Trump has declassified and released the transcript of the call, which shows him asking Zelensky to look into the Bidens. However, the White House insists this was not a condition for U.S. aid to Ukraine, despite reports suggesting aid might have been withheld around the same time.
Political Reaction: Democrats, including House Speaker Nancy Pelosi, have criticized the call, suggesting it showed Trump abusing his power for personal political gain. This has led to calls for investigations and even impeachment inquiries.
Public and Congressional Response: The release of the transcript has led to a mixed reaction. Some see it as evidence of wrongdoing, while others, including Trump’s allies, argue that the conversation was appropriate given the context of anti-corruption efforts.
This summary captures the essence of the article’s content, focusing on the main points of contention, the White House’s defense, and the political implications of the call.
Top 2 Comment Summary
The article discusses how the author managed to alleviate a severe depression by fasting for two weeks. While the exact reason for its effectiveness isn’t clear, there’s speculation that changes in the microbiome due to the drastic dietary change might be involved. This experience provides the author with a potential avenue for further investigation into managing depression.
7. Discovery Coding
Total comment counts : 44
Summary
The article introduces the concept of “discovery coding” in programming, paralleling it with the distinction between outline writers and discovery writers in literature. Here’s a summary:
Discovery Coding vs. Outlining: Discovery coders start writing code without a predefined plan, exploring the problem through coding itself. This contrasts with outline programmers who plan their approach before coding.
Cultural Perception: There’s a cultural bias towards structured methodologies in programming, similar to preferring outline writing in literature. Discovery coding is often seen as messy or undisciplined.
Process vs. End-Artifact: The article argues that the method of creation (discovery coding) does not necessarily reflect the quality or structure of the final product. A discovery coder can produce well-structured code.
Design Approaches: Discovery coding is not synonymous with bottom-up design, although there might be an overlap in preference. Top-down design focuses on the problem approach rather than the problem understanding process.
Benefits of Discovery Coding: This approach helps avoid jumping to familiar solutions prematurely, instead allowing the coder to understand the system’s constraints through interaction. This method suits those who find pre-written plans unhelpful or restrictive.
Tooling: Current programming tools are not typically designed with discovery coding in mind. Features like live programming and system visualization could enhance this approach.
Community Acceptance: The author calls for the programming community to recognize and accept discovery coding as a legitimate method, just as the writing community has accepted discovery writing.
The article emphasizes the importance of recognizing diverse coding styles and suggests that embracing these differences can lead to better problem-solving and creativity in software development.
Top 1 Comment Summary
The article discusses the inefficiencies of extensive planning and documentation in software development:
Design Docs as Abstractions: The author argues that design documents are inherently incomplete until the actual coding begins, where the real complexities of the problem are encountered.
Exploration Over Documentation: Instead of spending time on perfecting a document, the author suggests that “hacking around” to explore the problem space is more effective. This hands-on approach helps in understanding the full scope of the problem and estimating the time needed to implement solutions.
Practicality at Scale: While acknowledging that large teams might require detailed documentation for coordination, the author implies this can be cumbersome and less efficient for actual coding tasks.
Critique of Rigid Methodologies: There’s a critique of methodologies like Test-Driven Development (TDD) for being overly meticulous, suggesting they might appeal more to those who prefer not to engage directly with coding or have a need for high levels of organization.
Flexibility in Process: The author encourages a more flexible approach to coding, where writing extra code or experimenting with solutions is seen as less wasteful than over-documenting.
Overall, the article advocates for a pragmatic approach to programming where initial exploration and iterative improvement are prioritized over exhaustive preliminary planning.
Top 2 Comment Summary
The article discusses the appropriateness of different programming styles during various phases of software development. Specifically, it highlights that a particular style of programming is very effective during the technical design phase, while emphasizing the importance of having a plan during the actual development phase.
8. Why DeepSeek had to be open source
Total comment counts : 48
Summary
Summary of the Article:
The article discusses DeepSeek, a Chinese AI lab, which developed R1, an open-source reasoning model that rivals OpenAI’s proprietary model but was trained with significantly less resources. Here are the key points:
Why Open Source?: DeepSeek chose to open-source its model despite having a potentially market-leading product, which might seem counterintuitive from a traditional business perspective. However, this decision was driven by:
- Trust and Market Access: Being a Chinese company, there are concerns about data privacy and compliance in Western markets. Open-sourcing helps build trust by allowing users full control over the model.
- Resource Constraints: Due to export controls on advanced hardware, DeepSeek had to innovate more efficient training methods, which are shared through open-source.
Market Dynamics:
- Open-source models are gaining traction as they offer similar performance at a lower cost compared to proprietary models like those from OpenAI.
- The commoditization of LLMs means that the performance differences between models are lessening, making cost a significant factor.
Infrastructure vs. Consumer Software:
- In infrastructure, open-source is often preferred because it requires customization and maintenance regardless of whether it’s proprietary or not. Engineers benefit from being able to audit and modify the codebase.
- However, for consumer products, the maintenance burden of open-source can be a deterrent.
Future of AI:
- Despite the rise of open-source models, proprietary AI like OpenAI’s still holds advantages due to first-mover status and the foundational work that has enabled newer models through techniques like LLM distillation.
Conclusion: While open-source models like DeepSeek’s R1 challenge the dominance of proprietary models, the latter still have a role due to their pioneering work and market positioning. However, the trend towards open-source in AI infrastructure is clear, driven by cost, customization, and the inherent nature of technical products needing user control and transparency.
Top 1 Comment Summary
The article discusses DeepSeek, a Chinese startup focusing on large language models (LLMs), which has chosen to remain open source. Unlike other major Chinese tech companies like ByteDance, Tencent, Baidu, and Alibaba, which also work on LLMs but keep their models proprietary, DeepSeek has opted not to commercialize its technology or raise capital, instead focusing on community engagement and organic growth through open-source contributions. This approach, according to the CEO, helps in attracting top talent and emphasizes the value of the team, similar to the early days of OpenAI before it shifted towards a more closed model under Microsoft’s influence. The article highlights that while this choice initially kept DeepSeek in obscurity, its commitment to open source has begun to gain recognition and traction within the community by November 2024.
Top 2 Comment Summary
The article appears to be a brief query or statement questioning whether a certain entity only made available the weights of a model, similar to what others have done, rather than fully open-sourcing the entire model or codebase.
9. Seagate: ’new’ hard drives used for tens of thousands of hours
Total comment counts : 23
Summary
The German publication Heise.de has reported potential fraud involving Seagate hard drives sold as new but found to be heavily used. Readers who purchased Seagate Exos HDDs from various retailers, including official Seagate retailers, discovered through FARM tests that these drives had been used for thousands of hours, contrary to what was advertised. These drives, identified as OEM models rather than retail, did not come with the expected five-year warranty. Seagate is investigating the issue, which seems to involve multiple retailers and a significant number of drives, with usage ranging from 15,000 to 50,000 hours. The situation raises questions about whether there was intentional misrepresentation or a significant error in inventory management. Seagate has not commented on the issue yet, but they do engage in selling refurbished drives through an official eBay store in the US, which does not seem connected to this particular problem.
Top 1 Comment Summary
The article discusses how to access FARM logs on Seagate hard drives using the command smartctl -l farm /dev/sd<n>. These logs are considered more reliable than the standard SMART statistics. The author mentions that their two relatively new 16TB Seagate Exos drives, with dates of manufacture in June 2021, showed a discrepancy in usage hours: zero hours according to regular stats but approximately 18,000 hours according to FARM logs. Additionally, when trying to format these drives with ZFS, it failed because the drives were already formatted as part of a DDF (Disk Data Format) RAID setup.
Top 2 Comment Summary
The article discusses the author’s preference for buying used Western Digital Ultrastar drives from Amazon for their NAS (Network Attached Storage) system over new Seagate Exos drives due to several reasons:
Cost and Reliability: Used Ultrastar drives are cheaper and have proven reliable, especially with RAID-6 configuration which allows for two drive failures.
Quick Replacement: It’s faster to replace a failed drive by purchasing from Amazon than waiting for an OEM warranty replacement.
Backblaze Data: The author references Backblaze’s data, which shows that Ultrastar drives have a good reliability record.
Reseller Reliability: The reseller on Amazon seems trustworthy.
Remanufactured Drives: There’s evidence suggesting that many “remanufactured” drives might actually be OEM products with high no-trouble-found rates when returned, indicating they might work perfectly well.
The author sees an additional benefit in knowing they are buying used drives, which aligns with their understanding of the typical quality of returned hardware.
10. Apple and SpaceX link up to support Starlink satellite network on iPhones
Total comment counts : 20
Summary
The article is an instruction for users to verify they are not robots by clicking a specific box, ensuring their browser allows JavaScript and cookies, and provides guidance on where to find more information or how to get support if needed. It includes a reference ID for further inquiries.
Top 1 Comment Summary
Here is a summary of the article linked:
Title: How to Use ChatGPT to Write Code
The article details how ChatGPT, a language model by OpenAI, can be utilized to assist in writing code. Here are the key points:
Code Writing: ChatGPT can generate code for various programming languages, from simple scripts to complex applications. It can write functions, provide full code snippets, or even complete small programs. However, the code might not always be perfect or optimized, requiring human review for correctness and efficiency.
Debugging: Users can feed error messages or problematic code into ChatGPT to get debugging help. It can suggest fixes or explain potential issues, although its understanding might be limited to the context provided.
Explaining Code: If you provide code, ChatGPT can explain its functionality, which is particularly useful for learning or understanding inherited codebases.
Learning: For educational purposes, ChatGPT can be a tool for beginners to learn programming by asking for examples, explanations, or step-by-step guides on how to code.
Limitations:
- Outdated Knowledge: ChatGPT’s knowledge is up to date only until 2021, meaning recent libraries, frameworks, or updates might not be known to it.
- Contextual Understanding: It might not fully grasp the broader context or project requirements which could lead to less relevant or usable code.
- Security Concerns: There’s a caution against using AI-generated code in production without thorough review due to potential security vulnerabilities.
Best Practices:
- Always review and test the code provided by ChatGPT.
- Use it as a learning or starting tool rather than a complete solution provider.
- Keep in mind its limitations regarding real-time information and complex problem-solving.
The article emphasizes that while ChatGPT is a powerful tool for coding assistance, it should be used in conjunction with human judgment, particularly for critical applications.
Top 2 Comment Summary
The article expresses dissatisfaction from a Verizon/Android user regarding Verizon’s choice to support AST SpaceMobile’s (ASTS) technology over the established Starlink satellite internet service. The user appreciates that integrating support into the operating system helps prevent customer dissatisfaction and avoids having to mark features with disclaimers or limitations.