Cursor's deep dive: why hosting Git repositories at scale is a nightmare
zainhas · x · 2026-08-21
Cursor published a 27-minute research post "Git at any scale" by engineer Vicent Martí.
The piece traces Git's origins: Linus Torvalds designed it to replace BitKeeper for Linux kernel development, and its distributed architecture fit the kernel's decentralized maintainer workflow. Twenty years later, Git is an industry standard, but its distributed nature is more a hindrance than an advantage — the average open-source project and virtually every company relies on a centralized host.
Hosting Git at scale turns out to be extremely hard: in a distributed design, every repository copy is identical, and the server-side copy is no different from one on a laptop. While putting an HTTP daemon in front of an on-disk repo looks like a working Git server, serious scalability and reliability challenges lurk — the article dives into these design-level problems and Cursor's engineering practice.
Related event: Cursor Details Its S3-Centric Architecture for Hosting Git at Scale(3 posts)→
More from Infra
- Nvidia shorted over cuBLAS not using asymptotically optimal algorithms — basedjensen · 2026-08-21
- Fastest Qwen 3.8 27B version for AMD GPU? — Gloomy_Letterhead395 · 2026-08-21
- Unsloth V3 Qwen Model Broken on Dual AMD GPUs, V2 Fix Available — Equivalent-Ear-8016 · 2026-08-21
- CUHK Team Open Sources Libra: 3x Throughput for Agentic Training — jiqizhixin · 2026-08-21
- Open-source x402-cleanweb-agent saves 80% tokens by cleaning web content — EstablishmentTough18 · 2026-08-21
- Pretraining Potential: Coding Agents and the Compute Bottleneck — zeeshanp_ · 2026-08-21