模型写代码已够准:他主张删掉单元测试、取消代码评审仪式
sanderssays · x · 2026-10-01
Arguing that Opus 5.5 and Astra have reset his baseline, the author says unit tests now mostly restate code and waste tokens, and that PRs and code review have become ceremony — except for critical code like payments and auth. Let models write, test, and review each other instead.
「编程与Agent」频道最新
- Pocket FM 长上下文记忆系统精度对标 Claude Code，成本低 21 倍 — bigaiguy · 2026-10-01
- Ben Goertzel：AI 智能体不止修漏洞，还在协同编写形式化验证软件 — bengoertzel · 2026-10-01
- Pedro Domingos:已没有软件工程师,我们都是 Agent 驯兽师 — pmddomingos · 2026-10-01
- Stocktwits 接入 x402,向 AI Agent 出售市场情绪信号开辟新收入 — kleffew94 · 2026-10-01
- 不必买新工具：纯 Codex 就能搭出 dot/Space 式 agent 工作流 — jxnlco · 2026-10-01
- Animation Bench 发布：四款前沿模型复现真实网页动画大比拼 — himanshustwts · 2026-10-01