Rock Solid: Turn Cursor Into a Rock-solid Software Engineering Companion

,

“Rock solid” aims to enhance Cursor's role as a software engineering companion, defining comprehensive policies for AI coding agents and users. It emphasizes task-driven development, strict adherence to product requirements, controlled file creation, and clear roles. Key principles include user authority, task granularity, and documentation standards. The policy details backlog management, task workflows, testing strategies, and change management to maintain project integrity, accountability, and automation, ensuring quality and compliance throughout the development process.

https://gist.github.com/boxabirds/4a8a8a16b1f8431fd64a790209452380

A Break From Programming Languages

User reflects on a decade in programming languages, noting a decline in motivation partly due to health issues and the community's dynamics. They've had a long-standing passion for programming, from early experiments to professional work on languages like Haskell and Racket. Despite enduring satisfaction from solving software problems, increasing challenges in language design and community conservatism have led to disillusionment. The author recognizes accomplishments but expresses a desire to move on from programming languages, citing a disconnect with the Haskell community, particularly its focus on fintech and crypto. Ultimately, they conclude it's time to step back and reassess their interests and contributions, acknowledging the bittersweet nature of leaving this chapter behind.

https://lexi-lambda.github.io/blog/2025/05/29/a-break-from-programming-languages/

We Tested 7 Languages Under Extreme Load and Only One Didn’t Crash (It Wasn’t What We Expected)

TLDR: Testing seven programming languages under extreme load revealed that Erlang alone didn't crash. The experiment highlighted that architectural resilience matters more than performance, with memory management being a critical bottleneck across languages. Key insights include designing systems for failure, the importance of isolation, effective recovery capabilities, and resource management in maintaining reliability.

https://freedium.cfd/https://medium.com/@codeperfect/we-tested-7-languages-under-extreme-load-and-only-one-didnt-crash-it-wasn-t-what-we-expected-67f84c79dc34

A11y 101: How to Test Manually

Manual accessibility testing requires multiple tools: a computing device, screen readers (like NVDA), Chrome browser, spreadsheets for logging issues, and contrast checkers. Start with a site scope and objective, then test with mouse and keyboard navigation, followed by screen reader checks and color contrast analysis. Finally, assess writing clarity based on target audiences. Document findings and prioritize key issues for client follow-up, emphasizing continued support and training for ongoing improvement.

https://tarnoff.info/2025/05/15/a11y-101-how-to-test-manually/

Why Are 2025/05/28 and 2025-05-28 Different Days in JavaScript?

,

In JavaScript, 2025/05/28 and 2025-05-28 yield different results due to parsing behavior. The former is interpreted as local time, while the latter is treated as UTC, causing discrepancies in the displayed date. This inconsistency arises from varying interpretations of date formats across browsers, influenced by standards updates, leading to conflicting behaviors over time. Upcoming JavaScript Temporal APIs aim to resolve such ambiguity by allowing explicit handling of date-only strings without timezone confusion.

https://brandondong.github.io/blog/javascript_dates/

Vibe Coding for Teams, Thoughts to Date

LLMs are game-changers in coding, but haven't altered the fact that understanding code is harder than writing it. Each line of code represents technical debt due to limited understanding. Changing engineers' mindsets is tough; LLMs face similar challenges in grasping priorities from existing codebases. Current LLMs create additional, unnecessary functions, complicating codebases. The explosion of bespoke code can lead to inefficiencies and risks. While LLMs lower coding costs, they might enable better tools for code comprehension and organization. How are you adapting leadership styles with LLMs?

https://laughingmeme.org//2025/05/25/vibe-coding-for-teams.html

DuckLake Is an Integrated Data Lake and Catalog Format.

,

DuckLake: An integrated data lake and catalog format from DuckDB, using Parquet and SQL databases for advanced features without complexity. Supports PostgreSQL, SQLite, MySQL, and DuckDB for metadata management. Key features include snapshots, time travel queries, ACID transactions, and performance optimization. DuckLake offers a lightweight solution for collaborative data access and is licensed under MIT. Documentation available for setup and usage.

https://ducklake.select/

Making Video Games in 2025 (without an Engine)

,

Noel Berry discusses making video games in 2025 without using commercial engines like Unity or Unreal. He finds this approach more enjoyable and efficient, as it allows for customized tools and better control over development. Berry uses modern C# for programming, SDL for cross-platform support, and other libraries for assets and UI. He values open-source solutions and prefers lightweight setups, having transitioned to Linux for development. While he recognizes advantages in larger engines like Godot, he opts for simpler, tailored frameworks to match his indie game development style.

https://noelberry.ca/posts/making_games_in_2025/

Detecting Malicious Unicode

Curl contributor highlighted a security flaw where an ASCII character in a URL was replaced with a visually similar Unicode character, going unnoticed by reviewers or CI jobs. This could lead to severe impacts. The curl project has since implemented a CI job to spot malicious Unicode by flagging non-ASCII UTF-8 sequences to prevent potential exploits. Discussions emphasized the importance of recognizing such vulnerabilities, and GitHub is reportedly addressing the issue.

https://daniel.haxx.se/blog/2025/05/16/detecting-malicious-unicode/

After Months of Coding With LLMs, I’m Going Back to Using My Brain • Albertofortin.com

,

TLDR: After extensive use of LLMs for coding, I've reverted to manual coding due to chaotic code quality and frustration. I realized I'd been relying too much on AI instead of applying my own software engineering skills. Now, I limit AI's role to minor tasks while focusing on understanding and organizing code myself. I'm concerned about the detrimental effects of AI reliance on programming skills and worry that non-coders face greater challenges with AI tools.

https://albertofortin.com/writing/coding-with-ai

WebAssembly

WebAssembly (Wasm) is a binary instruction format for a stack-based virtual machine, designed for fast, portable web applications. It enables programming language compilation for client and server, focusing on efficiency, safety in a sandboxed environment, and open debuggability. Wasm supports seamless integration with JavaScript and adheres to web standards, allowing access to browser functionalities.

https://webassembly.org/

WCAG in Plain English

AAArdvark provides an accessible guide to WCAG (Web Content Accessibility Guidelines), summarizing key principles and criteria without replacing official documents. It covers essential accessibility features, auditing, visual tools, and supports various user needs across multiple levels (A, AA, AAA). For more details, visit their resources and sections focusing on organizations, agencies, and freelancers.

https://aaardvarkaccessibility.com/wcag-plain-english/

If AI Is so Good at Coding … Where Are the Open Source Contributions?

, ,

AI coding claims hype; lack of open-source contributions raises skepticism. CEOs cite minor AI contributions to code, but no substantial public evidence. High-profile skeptics urge sharing valid AI-produced pull requests. AI struggles with complex coding; challenges arise from inexperienced users submitting subpar work. Open-source projects resist AI-generated code aimed at maintaining quality.

https://pivot-to-ai.com/2025/05/13/if-ai-is-so-good-at-coding-where-are-the-open-source-contributions/

Building Software on Top of Large Language Models

,

Workshop on building software with Large Language Models (LLMs) at PyCon US by Simon Willison on May 15, 2025. Covering setup, prompting, a text-to-SQL tool, structured data extraction, semantic search, and tool usage techniques. Emphasis on practical exercises, material accessibility, and the diverse LLM landscape, including OpenAI, Gemini, and Anthropic models. Discussion on LLMs' cost-effectiveness, prompt injection security challenges, and evaluation methods for LLM applications. Concluding demos of local models with an invitation for private workshops.

https://simonwillison.net/2025/May/15/building-on-llms/#atom-everything

Scroll to Top