--- title: Redesigning Source Control at Anthropic | daily.dev description: Anthropic describes building Scion, an in-house version control system that replaced git across its monorepo in five months, driven by a surge from 2,500 to... canonical: https://daily.dev/posts/redesigning-source-control-at-anthropic-knpezbnsb twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Redesigning Source Control at Anthropic | daily.dev og:description: Anthropic describes building Scion, an in-house version control system that replaced git across its monorepo in five months, driven by a surge from 2,500 to... og:url: https://daily.dev/posts/redesigning-source-control-at-anthropic-knpezbnsb og:image: https://api.daily.dev/og/posts/knPezbnSb.png og:image:alt: Redesigning Source Control at Anthropic og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Redesigning Source Control at Anthropic **[Hacker News](https://daily.dev/sources/hn)** · 19 min read · 1 upvotes · 0 comments ## Summary Anthropic describes building Scion, an in-house version control system that replaced git across its monorepo in five months, driven by a surge from 2,500 to over 30,000 code merges per day as Claude-written code scaled dramatically. Unlike git's decentralized, offline-first design, Scion assumes clients are always online (mostly agents and CI jobs), uses an always-reachable server for history, blame, merges, and commits, supports sparse and virtualized checkouts, and exposes source control operations via API so jobs rarely need a local repository. The system reduced infrastructure footprint 200x versus prior git mirroring and sped up checkouts by two orders of magnitude. A seven-person team built Scion plus a code review tool (Antique) largely with heavy use of Claude, which also found optimizations and rewrote components with major performance gains. Anthropic frames this as a preview of challenges other organizations using AI coding agents at scale may eventually face, while noting Scion itself isn't meant as a general git replacement, and suggests git platforms could adopt server-side RPC and sparse-checkout ideas. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why did Anthropic replace git with its own version control system called Scion? Anthropic's merge rate grew from about 2,500 code changes per day in May to over 30,000 per day by September, driven largely by Claude writing code autonomously, overwhelming git's offline-first, decentralized design. Git's main branch history reached almost 100GB with over 25 million CI checkouts daily, which would have meant over 600PB of git traffic per day if fetched naively, so a seven-person team built Scion, an always-online, server-centric system, in five months. _Teams scaling CI and agent workloads past git's limits can track architecture lessons like this one on daily.dev._ ### How does Scion's client-server architecture differ from git's model? Scion assumes clients are always online and shifts most computation to the server, unlike git where every clone is a full self-contained repository. The Scion server handles file history, blame, commits, merges, and rebases, while the client maintains only a bounded local object cache and supports sparse or virtualized checkouts that materialize files on demand, drastically reducing local data requirements. _Developers evaluating server-centric versus distributed source control designs can follow these tradeoffs on daily.dev._ ### What performance improvements did Anthropic see after migrating from git to Scion? Anthropic's Scion infrastructure footprint became 200 times smaller than its previous git mirroring setup in CPU and memory usage, each caching tier maintains over a 99% cache hit rate, and CI and research jobs check out code two orders of magnitude faster, dropping from minutes to seconds, while CI alone gained a 25% CPU capacity improvement. _Engineers weighing infrastructure migrations at scale can track real-world benchmarks like this on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 187 comments across hackernews (as of 2026-10-14). **TL;DR:** Reactions mix genuine technical interest in Scion's architecture with heavy skepticism about whether 30,000 daily merges represents real productivity or just AI-driven churn, and whether any of this generalizes beyond a company with Anthropic's unique resources and incentives. **Sentiment:** 20% positive · 35% mixed · 45% skeptical **The case for** - Some find the engineering genuinely interesting and see value in rethinking source control for agent-heavy workflows beyond just reflexively using git. - One commenter notes the described rewrite of the ingestion pipeline by an internal model was a real, impactful performance win for their migration. **The pushback** - Many question whether high commit/merge volume reflects actual productivity or is just brute-forced busywork with little real-world benefit. - Several doubt this approach generalizes to typical companies, seeing it as only feasible given Anthropic's unusual scale and lack of normal profit constraints. - Some are skeptical about code quality and review rigor given the volume of AI-generated merges, questioning whether humans are meaningfully reviewing changes. - A few dismiss the whole framing as hype/marketing timed around an IPO rather than a genuine engineering necessity. - Commenters compare it unfavorably to older centralized systems like SVN or Perforce, implying it's not as novel as presented. **By community** - hackernews (heated): Discussion splits sharply between those intrigued by the technical approach and those skeptical it's meaningful progress, hype, or even good for society, with pointed debate about AI-driven code churn and whether commit counts matter. **Hottest debate:** Whether massive automated merge volume represents real engineering value or just frenetic, low-substance churn enabled by having money and inferior accountability to burn. **Open questions** - What is actually being built with all these merges, and does the output justify the churn? - How much human review actually happens on AI-generated code before it ships? - Does this approach generalize to companies without Anthropic's resources and incentives? **Highlights** > I genuinely don't understand the core value they're pitching here. The tech is interesting, and I'm not trying to under cut that, but I question the idea that this is directionally valuable advice to nearly any company. This amount of churn feels like a thing only possible with no real profit motives and a lot of money to waste. Lots of activity is not necessarily synonymous with productivity, and it carries real cost. So I think the design here is interesting, and the concepts seem applicable to where some agentic development may end up, but I question the underlying principles that drive this being the right solution space at all. At least if the goal is a profitable company. > — [horsawlarway on hackernews · 1 comments](https://news.ycombinator.com/item?id=50076054) > I find the accelerationist mindset utterly ridiculous. All that busywork and for what exactly? What effect has all this overproduction of LOC had on the physical world and on our existence (beyond the externality of erecting huge data centers in every corner of the globe and the consequent ecological impact)? Is software getting better (in terms of quality, security, efficiency)? Are we getting better government services? Are we getting better food? A better environment (silly question)? Are we working less, as tech titans keep promising us? Are we getting anything better? Not to mention all the negative aspects of this technology and the financial apparatus driving it. > — [ciconia on hackernews · 2 comments](https://news.ycombinator.com/item?id=50077218) > They invented a premise that fit their current business model, and then reinvented server-based version control to “solve” it. This is how idiotic these people are at Anthropic/OpenAI/etc. I can’t wait until the bubble bursts and all these idiots are jailed. And we can move on as a society where AI is a tool, not a personality trait. > — [hejdenxi on hackernews · 1 comments](https://news.ycombinator.com/item?id=50076159) > > Earlier in the year, we had largely handwritten the first version of the git-to-Scion ingestion pipeline as a prototype. We thought it was good—until we got our hands on an early internal snapshot of Claude Mythos Preview. We asked it to rewrite the whole thing, and with no further instructions, it produced a version that consumed 20x less memory and was 4x faster, while preserving all of the implicit invariants around ingestion ordering and data structures. I feel like the advantage of having these models in house can't be understated. Companies with access to inferior models, trying to compete, will be left in the dust when it comes to general development. > — [trnvll on hackernews · 2 comments](https://news.ycombinator.com/item?id=50075355) > Is this really something to be proud of? I make heavy use of AI, but for anything that is eventually going to production there's still a human reviewer. Based on this I assume Anthropic is shipping code with no human review? Is the future of our industry going to be oneshotting prompts with out manual code review and then having AI fix the bugs found in production buy customers? I'm not trying to be a Curmudgeon, but am I missing something? > — [jareds on hackernews · 2 comments](https://news.ycombinator.com/item?id=50075604) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=50073633) · 203 points · 187 comments ## Similar posts on daily.dev - [Google’s Scion Gives Developers a Smarter Way to Run AI Agents in Parallel](https://daily.dev/posts/google-s-scion-gives-developers-a-smarter-way-to-run-ai-agents-in-parallel-os7hdmwvq) · DevOps.com · 0 upvotes · 0 comments - [GitLab and Anthropic: Governed AI for enterprise development](https://daily.dev/posts/gitlab-and-anthropic-governed-ai-for-enterprise-development-jtseezzdo) · GitLab · 1 upvotes · 0 comments - [Google Open Sources Experimental Multi-Agent Orchestration Testbed Scion](https://daily.dev/posts/google-open-sources-experimental-multi-agent-orchestration-testbed-scion-lnrn47ksh) · InfoQ · 0 upvotes · 0 comments --- Tags: [#general-programming](https://daily.dev/tags/general-programming), [#git](https://daily.dev/tags/git), [#anthropic](https://daily.dev/tags/anthropic), [#version-control](https://daily.dev/tags/version-control) [View this post on daily.dev](https://daily.dev/posts/redesigning-source-control-at-anthropic-knpezbnsb) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Redesigning Source Control at Anthropic","url":"https://daily.dev/posts/redesigning-source-control-at-anthropic-knpezbnsb","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/redesigning-source-control-at-anthropic-knpezbnsb"},"datePublished":"2026-10-13T17:53:50.731Z","dateModified":"2026-10-14T04:20:36.375Z","description":"Anthropic describes building Scion, an in-house version control system that replaced git across its monorepo in five months, driven by a surge from 2,500 to...","image":"https://media.daily.dev/image/upload/s--HRgLpUt6--/f_auto/v1722860399/public/Placeholder%2003","thumbnailUrl":"https://media.daily.dev/image/upload/s--HRgLpUt6--/f_auto/v1722860399/public/Placeholder%2003","isAccessibleForFree":true,"articleSection":"Hacker News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Hacker News","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/hn","url":"https://daily.dev/sources/hn"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/redesigning-source-control-at-anthropic-knpezbnsb","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"general-programming,git,anthropic,version-control","timeRequired":"PT19M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Hacker News","item":"https://daily.dev/sources/hn"},{"@type":"ListItem","position":3,"name":"Redesigning Source Control at Anthropic"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/redesigning-source-control-at-anthropic-knpezbnsb#faq","mainEntity":[{"@type":"Question","name":"Why did Anthropic replace git with its own version control system called Scion?","acceptedAnswer":{"@type":"Answer","text":"Anthropic's merge rate grew from about 2,500 code changes per day in May to over 30,000 per day by September, driven largely by Claude writing code autonomously, overwhelming git's offline-first, decentralized design. Git's main branch history reached almost 100GB with over 25 million CI checkouts daily, which would have meant over 600PB of git traffic per day if fetched naively, so a seven-person team built Scion, an always-online, server-centric system, in five months. Teams scaling CI and agent workloads past git's limits can track architecture lessons like this one on daily.dev."}},{"@type":"Question","name":"How does Scion's client-server architecture differ from git's model?","acceptedAnswer":{"@type":"Answer","text":"Scion assumes clients are always online and shifts most computation to the server, unlike git where every clone is a full self-contained repository. The Scion server handles file history, blame, commits, merges, and rebases, while the client maintains only a bounded local object cache and supports sparse or virtualized checkouts that materialize files on demand, drastically reducing local data requirements. Developers evaluating server-centric versus distributed source control designs can follow these tradeoffs on daily.dev."}},{"@type":"Question","name":"What performance improvements did Anthropic see after migrating from git to Scion?","acceptedAnswer":{"@type":"Answer","text":"Anthropic's Scion infrastructure footprint became 200 times smaller than its previous git mirroring setup in CPU and memory usage, each caching tier maintains over a 99% cache hit rate, and CI and research jobs check out code two orders of magnitude faster, dropping from minutes to seconds, while CI alone gained a 25% CPU capacity improvement. Engineers weighing infrastructure migrations at scale can track real-world benchmarks like this on daily.dev."}}]} ```