Skip to content
View original post on X: Dongxi NLPX· Same story68/100AI score68/100

Anthropic says daily code merges rose from 2,500 to over 30,000 and outgrew its tooling

AISummary

Dongxi NLP relays an Anthropic engineering post saying daily merged code changes rose from about 2,500 in May to more than 30,000 in September.

Anthropic says its developer infrastructure, built for human rather than agent scale, became the bottleneck, and it built a new version control system and migrated its monorepo off git in five months.

The author argues that infrastructure for code agents should be server-centric, API-driven, sparse-access, high-throughput and strongly governed.

Post on XView on X
Dongxi NLPVerified on X
@dongxi_nlp

From May to September, Anthropic's daily code merges grew from 2,500 to over 30,000. The infrastructure built for human-scale work became the bottleneck.

What does infrastructure for code agents look like?

Anthropic's blog post actually gives a fairly concrete answer:

It no longer assumes the client is a human. It no longer optimizes for offline experience on a laptop or for humans reviewing PRs one by one.

Agent-oriented infrastructure will be more server-centric, API-driven, sparsely accessed, high-throughput, and strongly governed. The optimization target is scenarios where thousands of always-on agents simultaneously read, write, merge, test, and ship code.

AITranslated from Chinese
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 monorepo from git—in five months: https://www.anthropic.com/engineering/redesigning-source-control
View quoted post on X

This story is featured in top picks: “Anthropic builds in-house source control after merge volume outgrows git”

Source: Dongxi NLP · x.comPublished · added here