Tag
Gergely Orosz 的 Pragmatic Engineer 通讯文章回顾 2025 年软件工程与 LLM 的结合现状，涵盖 Anthropic、Cursor、Google、Amazon 等公司如何使用 Claude Code、Cursor、Windsurf 等 AI 开发工具，以及资深工程师们的观察。
一位运营5家一人公司的创业者分享他基于 Cursor 的 AI 工作流系统：用 AGENTS.md、每条业务线独立文件、CEO agent 排程、日志与审查机制，把 AI 产出控制在可审查范围内以稳步走向 $50K MRR。
Matt Pocock lists common mistakes developers make when working with AI coding agents, including misaligned effort, overly long prompts, weak automated checks, poor codebase navigation, and insufficient agent autonomy.
Bradley Gibbs 的文章提出"循环工程"方法，介绍如何通过优化终端评估、内层循环与学习循环来降低 AI 辅助代码生成带来的验证瓶颈，保留人类的时间、注意力与判断力。
Simon Willison tests Mistral Large 4 alongside frontier models (Claude Opus 5.5, GPT-6.1, Gemini 3.8 Flash) by generating an SVG of an armadillo in fishnet tights jaywalking on Mars, highlighting how saturated benchmarks push toward absurd creative prompts.
OpenHuman is an open-source AI agent harness with a Rust core that has become the #1 trending tool on GitHub, claiming better efficiency, lower RAM/CPU usage, and faster results than OpenClaw and Hermes.
A group member released Paper-Xray, a paper-breakdown skill with 15 XML sections that generates detailed technical reports on papers (and other content). Tested against models like dsv4.1f, space bunny, and gpt6.1sol, it reportedly produces more comprehensive breakdowns than public blog posts, though slightly less human-sounding.
GNOME developer Michael Catanzaro argues that AI vulnerability scanning is now essential for maintaining secure, high-quality open-source software, noting that AI-generated vulnerability reports have improved dramatically in 2026 despite still causing issues like verbose or occasionally fabricated reports.
A user shares their setup for free, simple web search extraction in Open WebUI using OpenSERP, including two PRs: one making OpenSERP compatible with OWUI's extractor and another adding OpenSERP as a search provider in Open WebUI.
OpenAI developers' thread-reader roundup highlighting developer updates around the Decisions API, alongside other announcements like GPT-Live-1, Codex skills, and the open-source Open Responses spec.
The article explores why TypeScript has become the dominant language for building AI agents and agent frameworks, questioning why alternatives like Rust or C++ aren't more prevalent.
This reference implementation demonstrates how to run an LLM agent securely within a local sandbox to process and analyze various document types using Rust, LiteParse, and microsandbox. The open-source CLI leverages OpenAI's GPT models and native bash commands to perform file retrieval and analysis in an isolated environment.
Hermes Agent v0.13.0 ('The Tenacity Release') ships with durable Kanban, persistent goals, Checkpoints v2 with rollback, and 8 P0 security fixes, positioning itself as a runtime persistence layer alongside coding agents like Claude Code and Codex. The release coincides with cheap 1M-context models like DeepSeek V4-Pro and MiMo-V2.5-Pro, making long-running agentic software work more viable.
Moonshot AI has open-sourced the Kimi Vendor Verifier (KVV), a tool designed to help users verify the accuracy and correctness of inference provider implementations for open-source models like Kimi K2. It uses six critical benchmarks to detect infrastructure-level issues such as KV cache bugs, quantization degradation, and parameter misuse.
A developer argues that modern AI agent terminology (harnesses, orchestrators, memory layers, etc.) is largely rebranding of familiar software engineering patterns, coining the phrase 'Vocabulary Tax' for the intimidation new jargon creates. The post aims to demystify agentic AI concepts for experienced developers.
In this Anthropic guest post, Prof. Matthew Schwartz describes BootLoops, an open-source toolkit of scientific software and protocols built with Claude to do exact calculations in quantitative science. By deliberately targeting "Claude-shaped" problems, the tools surfaced cross-domain connections to ecology, population genetics and other fields, which domain experts then helped steer toward scientifically meaningful questions.