OpenAI’s Pando cuts Git push latency by about 80% for its monorepo
OpenAI has developed its own Git backend called Pando, which reduces code push latency to the company's monorepo by approximately 80%, or nearly fivefold.
The custom backend, announced on November 2, 2026, went into production in April after about 10 weeks of development. Led by one engineer with substantial help from teams of Codex agents, Pando addresses two major scaling challenges: simultaneous writes to different branches and heavy read traffic.
Technical Innovation
Pando enables parallel updates to unrelated branches while maintaining causal consistency within each branch, ensuring edits land in sequence. Crucially, it preserves full compatibility with the Git protocol and existing clients, with OpenAI contributing work back to the Git project itself.
Driving Factors
The pressure came from Codex, with AI-driven commits surging 1,378% from Q1 2024 to Q3 2026. Forecasts predict 2,400% growth by end of 2026, causing traditional Git servers to struggle with reliability often dipping below 90%.
Industry Impact
At peak performance, Pando pushed over 57 TB of Git pack data in one hour, with over 99% from local disk cache. Clone and fetch operations can utilize up to 10 Gbps bandwidth on standard cloud VMs. The project demonstrates how AI assistants can accelerate infrastructure development.
Related
