←新闻详情

马东锡 NLP更新于 10月21日 22:46

Anthropic 每日合并代码变更从 2,500 增至超 30,000,作者解读面向 code agent 的 infra 形态

从五月到九月, Anthropic 每天合并代码变更从 2,500 到超过 30,000 个。为人类规模构建的 infra 成了瓶颈。

AI 导读

Anthropic 工程博客披露,其每日合并代码变更从五月的约 2,500 个增至九月的超过 30,000 个,为人类规模构建的开发者基础设施因此成为瓶颈。作者据此归纳,面向 code agent 的 infra 不再假设客户端是人,也不再优化笔记本上的离线体验和人类逐条审 PR,而是更服务端中心、API 驱动、稀疏访问、高吞吐、强治理,目标是支撑成千上万个始终在线的 agent 同时读写、合并、测试和落地代码。

推荐理由

AI 评分 59

Anthropic 合并量半年增长十余倍,作者据此归纳出面向 code agent 的基础设施该具备哪些特征。

主题

标签

引用

Anthropic

@AnthropicAI

New on our Engineering blog: Creating in-house source control. In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale. Here, we share how we built a new version control system—and fully migrated our…

展开完整原文

New on our Engineering blog: Creating in-house source control. In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale. Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: https://t.co/wzNkwHs7IN

查看原帖

正文

从五月到九月, Anthropic 每天合并代码变更从 2,500 到超过 30,000 个。为人类规模构建的 infra 成了瓶颈。

面向 code agent 的 infra长什么样?

Anthropic 在这篇博客里其实已经给出了比较具体的答案:

不再假设客户端是人,它不再优化笔记本上的离线体验和人类逐条审 PR。

面向 agent 的 infra 会更服务端中心、API 驱动、稀疏访问、高吞吐、强治理。优化目标是“成千上万个始终在线的 agent 同时读写、合并、测试、落地代码”的场景。

来源

马东锡 NLP