Loading...
Loading...
AI-assisted development is shifting from autocomplete to agentic workflows that can plan, execute, and iterate across repositories—often in the cloud and running in the background. New tooling reflects that transition: agent orchestration apps, emerging “agentic engineering” patterns, and integrations like an MCP interface for Chrome DevTools that let agents debug and inspect browser behavior directly. At the same time, agents are expanding beyond app code into performance work, with reports of meaningful success diagnosing GPU bottlenecks and new projects promoting GPU learning with agents alongside GPU-native languages like OctoFlow. The trend is reshaping expectations for developers and hiring, while raising concerns about skill atrophy and understanding.
How to use storytelling to fit inline assembly into Rust | Hacker News Hacker News new | past | comments | ask | show | jobs | submit login How to use storytelling to fit inline assembly into Rust ( ralfj.de ) 6 points by vinhnx 1 hour ago | hide | past | favorite | discuss help Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact Search:
The diminished art of coding
Writing my own text editor, and daily-driving it | Hacker News Hacker News new | past | comments | ask | show | jobs | submit login Writing my own text editor, and daily-driving it ( jsbarretto.com ) 13 points by todsacerdoti 1 hour ago | hide | past | favorite | discuss help Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact Search:
Paul Graham famously framed how to build software in the simplest possible terms: make something people want. Soon, it will be make something agents want. [翻译] Paul Graham 曾用最简单的话概括如何构建软件:做人们想要的东西。 很快,这将变成:做 AI 代理想要的东西。
Eyot is a new programming language prototype that treats the GPU as just another threadable worker, letting developers spawn GPU jobs with the same syntax and runtime convenience as CPU threads. Authored by Duncan Steele, Eyot transparently compiles the same source for CPU and GPU and handles memory, compilation, scheduling, and communication in the runtime, aiming to simplify GPU/offload programming for games, numerical analysis, and AI. The blog showcases a simple example where a function can be invoked directly, as a CPU worker, or as a GPU worker via keywords like cpu and gpu, with send/receive-style message passing. Early-stage and experimental, Eyot is not yet production-ready but invites feedback and experimentation via a playground.
Eyot is an experimental programming language that treats the GPU like a first-class thread: functions can be invoked normally, spawned as CPU background workers, or compiled and scheduled as GPU kernels with the runtime handling memory, compilation, and communication. Creator Duncan Steele positions Eyot for domains that heavily use accelerators—games, numerical analysis, AI—aiming to remove the friction that currently deters developers from offloading work to GPUs. The language already demonstrates GPU-side printing and a simple worker model in its playground, but remains early-stage and not production-ready. The author plans rendering and Vulkan support and hopes to replace OpenCL; development is hobbyist-paced and seeking feedback and sponsors.
Show HN: 生存指数——人工智能代理会选择哪些开发工具?
To put this in other words: Writing code by hand was walking Agentic IDEs were getting in a car Agentic CLIs in an existing repo are an airplane Cloud agents in a new repo are getting in a rocket And yes, I agree. https://t.co/Lzl353Y4ch [翻译] 换句话说:手写代码是步行,代理式 IDE 是开车,在现有仓库中使用代理式 CLI 是坐飞机,在新仓库中使用云代理是坐火箭。是的,我同意。
Every function is getting rewritten around one filter: can this person work with AI systems? Indeed tracked a 130% surge in AI-mentioning job postings since pre-pandemic. Total postings? Flat. Only 6% above 2020 levels. Companies aren't hiring more people. They're replacing the [翻译] 每个职位都在围绕一个筛选标准重写:这个人能否与 AI 系统协作?Indeed 追踪到自疫情前提及 AI 的职位发布激增130%。总职位数?持平。仅比2020年水平高6%。公司没有招更多人,而是在替换
The article discusses the categorization of software errors into two types: expected and unexpected. Expected errors, such as user input mistakes or network failures, are part of normal operation and should be handled gracefully by developers. In contrast, unexpected errors, like null pointer exceptions, indicate bugs and should lead to program crashes. The author emphasizes the importance of effective error handling in software development, arguing that distinguishing between these error types can enhance user experience and software reliability. This perspective is informed by the author's extensive experience in web and application development, making it relevant for developers and tech professionals.
Lysium is an app I've been building specifically for background agent orchestration. Features: → Mobile-first + cross-platform → Run multiple agents in parallel across repositories → Launch agent-based requests from issues and PRs → Swipe actions for close, merge, create PR, https://t.co/LWUyrVdONW [翻译] Lysium是我一直在开发的专门用于后台智能体编排的应用。功能:→ 移动优先+跨平台 → 跨仓库并行运行多个智能体 → 从issues和PR启动基于智能体的请求 → 滑动操作关闭、合并、创建PR
The third era of AI software development
The third era of AI software development
The article discusses the challenges faced by Python type checkers when inferring types for empty containers, such as lists and dictionaries. It highlights two primary strategies employed by different type checkers: inferring 'Any' type for container elements, which simplifies implementation but sacrifices type safety, and inferring the container type based on its usage in the code, which enhances type safety but can be more complex. The piece evaluates the pros and cons of each approach, providing insights for developers on selecting an appropriate type checker. This analysis is particularly relevant for Python developers focused on improving code quality and reducing runtime errors.
Cursor 工程师很好的文章,AI 编程的三个阶段: 1 以代码自动补全(Tab 早期Copilot) 2 同步 Agent,A在编辑器中实时处理更复杂的任务,但仍局限于本地环境、单次交互、即时响应模式。 3 云端长期运行 Agent(cloud / background / long-running [翻译] Cursor 工程师很好的文章,AI 编程的三个阶段:1. 代码自动补全(Tab 早期 Copilot)2. 同步 Agent,在编辑器中实时处理更复杂的任务,但仍局限于本地环境、单次交互、即时响应模式。3. 云端长期运行 Agent
@simonw: New chapter of my Agentic Engineering Patterns guide. This one is about having coding agents build c
In a reflective blog post, a programmer discusses the impact of AI coding agents on his work, comparing his experience to that of a 747 pilot who feels stagnant in his career. The author notes that while AI tools, particularly large language models (LLMs), have significantly improved efficiency by automating coding tasks, they also hinder skill development and deep understanding of programming. He expresses concern that reliance on these agents may lead to a lack of growth in coding proficiency, as tasks can be completed without the programmer fully engaging with the underlying code. This commentary highlights the dual-edged nature of AI in software development.
Andrej Karpathy / @karpathy: AI coding agents made a huge leap forward since December, completing complex projects with minimal oversight, meaning “programming is becoming unrecognizable” — It is hard to communicate how much programming has changed due to AI in the last 2 months: not gradually and over time in the …
An article titled "If I hear \"design pattern\" one more time, I'll go mad" appears to address frustration or fatigue with the frequent use of the term "design pattern" in software engineering discussions. With no body text available, details such as the author’s specific arguments, examples, or any referenced frameworks, languages, or companies cannot be confirmed. Based on the title alone, the piece likely comments on how design patterns are discussed, taught, or applied in programming, and may question overuse of jargon or rigid adherence to pattern-based thinking. The topic matters because design patterns influence software architecture decisions, developer communication, and engineering culture, and debates about their usefulness versus overapplication are common in the industry.
A new programming language called OctoFlow has been introduced, designed specifically for GPU-native execution. Unlike traditional languages that treat GPU processing as an auxiliary function, OctoFlow allows data to reside on the GPU by default, optimizing performance for tasks such as statistical analysis and machine learning. It supports operations on arrays with over 10 million elements and includes features like an interactive REPL and HTTP client capabilities. The language is still in its early stages, and the creator is seeking feedback from users to improve its functionality. This development could significantly impact fields requiring high-performance computing, particularly in AI and data processing.
A blog author recounts learning programming while incarcerated, framing it as part of a personal “redemption arc.” The writer says they are drafting the post on a ThinkPad using the NVIM editor, after returning to prison following years of substance use and repeated arrests starting at age 14. During COVID-era intake, they spent 14 days in quarantine on “23 and 1” lockdown in a county of about 67,000 people, which they describe as a turning point. Over the next 13 months, they studied widely—economics, engineering, technology, and finance—taking extensive notes. In October 2022, a judge sentenced them to 18 years with 7 to serve and 11 suspended; they were transferred to an intake facility (MCC) and then Maine State Prison. The post offers limited technical details beyond tools used.
Chrome DevTools for coding agents Language: TypeScript Stars: 34 Forks: 0 Contributors: OrKoN,Lightning00Blade
Show HN: Learn GPU programming with coding agents