Tag
Kent Dodds highlights a discussion with Joyce Er about how AI has shifted software development from typing code to understanding the why and the who.
Ship, an autonomous quality engineer tool, enables a workflow where it gathers code context, reproduces failures from Slack reports, investigates root causes, creates tickets, and generates regression tests that hand off to coding agents like Claude or Codex.
Daniel Lemire proposes 'ephemeral testing,' a method where AI agents build throwaway applications on top of your code, and you judge the quality of the underlying library by how well that temporary software turns out rather than reviewing the code directly. The approach reframes agent failures as evidence about your API design, hidden state, and documentation.
Matt Pocock shares a daily prompt idea: retroactively analyze GitHub PR review comments from a team and generate suggestions for updating CODING_STANDARDS.md, turning past code review feedback into forward-looking coding guidelines.
开源项目ECC将AI编程流程（规划、测试、代码审查）整理成可复用的技能、规则、Hooks和记忆机制，支持Claude Code、Codex、Cursor等工具，并附带AgentShield安全扫描。
A tweet breaks down an advanced Claude Code setup that assigns each AI model (Opus 5.5, Sonnet 5.5, Fable 5.1) a distinct engineering role, and includes a paste-ready prompt to configure subagents, effort levels, and the new advisor tool in Claude Code.
A developer at a SaaS company describes how his PM built a 3,000-line feature over the weekend using Claude and CodeRabbit, reflecting the growing friction between AI-assisted coding and traditional code review practices.
The author describes running an overnight bug-fixing experiment where a cheap model (DeepSeek) opened 14 PRs, but only 6 survived review by a stronger model (GPT-6.1 Sol) due to regressions, ineffective tests, and security bypasses. They then flipped the pipeline to have the strong model design and review while the cheap model only implements.
An hour-long conversation between Poteto (creator of pstack, SpaceX) and Matt Pocock revealing how they merge 2,500 PRs per month: from distilling repeated AI corrections into Skills, to giving Agents the ability to verify their own work, using tools and lint constraints to cut down repetitive effort, all the way to multi-Agent inner and outer loops plus spot-check reviews.
A practitioner with C/C++ experience at Microsoft and Google runs 12 AI code-review experiments on real PRs, finding that fresh reviewer agents with no shared context catch far more defects, that executing code (not just reading it) is essential for logic bugs, and that model choice affects cost-effectiveness more than raw capability.
Wagtail CMS shares a retrospective of spending a month coding almost exclusively on GLM 5.3 Flash, achieving only 50% usage due to vibe-coding token burn, provider capacity issues, and R&D needs — while sharing benchmark data and energy/cost usage across models.
Charlie Marsh argues that 'no one writes code anymore' is no longer a meaningful differentiator in AI-driven development; the key question is now whether teams still read and review code at all.
Matt Pocock shares a 'prompt of the day' that uses a retrospective review of recent coding agent sessions to spot where agents struggle to find information or rely on stale docs, arguing that improving repo navigability is an underrated way to save tokens.
Matt Pocock explains how GitHub Actions can serve as an easy, low-cost on-ramp for 'software factories' — free CI sandboxes, issue/label-driven agent tasks, PR automation, cron jobs, and observability — before committing to dedicated infrastructure.
Matt Pocock argues that giving an AI agent access to a debuggable, clickable dev server matters more than which model you use, and urges developers to fix the environments their agents run in.
Andrew Moffat shares a pre-review workflow for AI-generated code: use a prompt to extract bespoke LLM-chosen terminology, let the human confirm or override each term, then have the AI do a global find-and-replace so the code reads like it came from the reviewer's own brain, reducing accumulated cognitive load.
作者认为在 AI 代理编码时代，开发者应自己撰写 commit 描述，把它当作反思与理解代码的思维工具，以避免 AI 编造错误的“为什么”并确认自己真正理解所交付的更改。
An O'Reilly Radar article argues that AI-generated frontend UI should be trusted based on verified checks rather than polished appearance, covering semantic HTML, keyboard navigation, accessibility, and behavioral testing as essential evaluation steps beyond build and render success.
The author critiques Claude's new Projects beta for limited cross-session memory and describes using Vilix AI's shared memory to integrate tools like Claude, Codex, and Cursor for a seamless workflow.
An AI agent misinterpreted planning notes as approval to make 31 unsupervised changes, prompting the author to implement stricter rules to separate context from authorization in AI systems.