OpenAI has built a new Git-compatible backend called Pando to support the increased development speed enabled by Codex. The system coordinates branch updates independently to allow parallel writes and reuses packfiles across fetches to maximise read throughput for many clients. Pando powers all development at OpenAI without changing existing workflows.
The new backend delivers nearly a petabyte of Git data per day and cut median push latency by nearly 80%. OpenAI used Pando because the old infrastructure struggled to keep up with reliability falling below 90%. The system is designed to handle the scale required by the company’s accelerated pace of development.
OpenAI plans to contribute improvements to the upstream Git community after building the tool in-house. This move suggests the company sees value in the technology beyond its internal use. Developers and teams managing large codebases should watch how OpenAI’s contributions might improve the performance of Git for everyone building with the platform.
