最新资讯
dev.to 银行 77 意图分类最便宜 LLM:GPT-6.1 Terra 每次成功任务 $0.000727 theinformation.com Chesky的副业:Airbnb向CEO新设AI实验室出售IP theverge.com OpenAI 宣布数据中心项目不再与公职人员签署保密协议 dev.to NVIDIA 发布四个开放 Cosmos 3 仿真模型 dev.to 真实 MCP 客户端击穿我们服务器的五种方式 theverge.com Airbnb 正在开发 AI 旅行 agent,计划 2027 年上半年推出 wechat 这可能是美国AI最魔幻的一天。 nytimes.com 美众议院共和党领袖要求 Nvidia 黄仁勋闭门简报 pub.towardsai.net 忘掉我:为记忆而设计的系统中的删除难题 marktechpost.com Google 发布 Gemini Omni 1.2 Flash:15 秒视频,每秒 0.07 美元 unite.ai Airbnb 计划推出覆盖规划、预订与客服的 AI 旅行 agent techcrunch.com Airbnb 称将于 2027 年上半年推出 AI 旅行代理
多元动态
登录 注册
大模型 www.aibase.com ·

Claude 编写 Anthropic 八成代码,CI 任务半年激增 25 倍

原文 我的收藏

据 AIbase 转述 Anthropic 内部数据,Claude 已编写该公司约 80% 的代码,工程师季度代码交付量达到 2021 至 2025 年平均水平的 8 倍,代码审查和合并审批也大量由 Claude 处理。由于 AI 倾向提交极细粒度 PR 并全天持续提交,测试用例数量激增十倍,CI 运行任务半年内增长 25 倍,导致 CI 系统过载。Anthropic 先后尝试增加核心、分片和每日强制重启三种补丁均未奏效,最终采纳 Claude 建议,以分布式无状态架构重建服务才解决问题。

AI推荐理由 内容给出Anthropic内部披露的具体数字:Claude编写约80%代码、工程师季度交付量为2021-2025年平均的8倍、CI任务半年增长25倍,并描述三次补丁失败后改用分布式无状态架构解决过载。这些细节对理解AI编程对CI/CD等工程基础设施的冲击有实际参考价值。但来源为AIbase转述,未提供原始披露链接或可验证证据,数字无法独立核实,且带有'差点被砸瘫'等标题化表述,故质量分受限。

72

正文

Claude wrote 80% of Anthropic's code, CI tasks surged 25 times in half a year, nearly causing system breakdown AIbase基地 Published inAI News · 3 min read · Sep 18, 2026 7 Internal data disclosed by Anthropic paints an exaggerated picture: 80% of the company's code is written by Claude, and engineers' quarterly code delivery volume reaches eight times the average level between 2021 and 2025. Even tasks such as code reviews and merge approvals are largely handled by Claude. However, the "behavior pattern" of AI writing code is completely different from that of humans. It prefers to submit PRs with extremely fine granularity, constantly submitting throughout the day, and causing the number of test cases to surge tenfold - this approach directly crashed the CI system: the number of CI run tasks increased 25 times within six months. Anthropic made three patches - adding core, sharding, and daily forced restarts - but all failed. Finally, they listened to Claude's suggestion and completely rebuilt the service, adopting a new distributed stateless architecture, which truly solved the CI overload issue. The signal of this incident is clear: the impact of AI programming has spread from the question of whether "programmers can finish writing code" to the capacity of the entire software engineering system. When writing code is no longer constrained by human rhythms, the first things to fail are often the pipelines designed according to human habits. AI jargonClaudeCI systemDistributed architecture This article is from AIbase Daily Scan to view Welcome to the [AI Daily] column! This is your daily guide to exploring the world of artificial intelligence. Every day, we present you with hot topics in the AI field, focusing on developers, helping you understand technical trends, and learning about innovative AI product applications. —— Created by the AIbase Daily Team © Copyright AIbase Base 2024, Click to View Source -

来源报道 · 1 篇 · 以下链接均前往原始媒体

← 返回精选动态 原文

更多内容