OpenAI Unveils Pando to Speed Monorepo Pushes
OpenAI announced Pando, a custom Git backend built to handle surging commit volume from its AI coding agent Codex.
Updated on Nov. 3, 2026 in Artificial Intelligence
OpenAI announced Pando on November 2, 2026, a custom Git backend for its monorepo that reduced median push latency by about 80%. The system was built in about 10 weeks and went into production in April 2026.
Why it matters
OpenAI's repository came under strain from a surge of commits generated by Codex, its AI coding agent, which Pando was designed to absorb. The project shows how AI-driven development is pushing the limits of conventional version-control infrastructure.
The figures come from OpenAI's own internal measurements of Pando, which serves nearly one petabyte of Git data per day on a handful of CPU cores. Codex-driven commits rose 1,378% between Q1 2024 and Q3 2026, and the company forecasts growth of 2,400% by the end of 2026, while independent validation of these numbers has not been reported.
The players
OpenAI
OpenAI is the artificial intelligence company that built and operates Pando and develops the Codex coding agent.
The details
Pando allows parallel updates to unrelated branches while keeping changes ordered within each branch, providing causal consistency per branch. It maintains full compatibility with the Git protocol and existing clients, and at its recorded peak pushed more than 57 TB of Git pack data in one hour, with over 99% served from local disk cache. OpenAI has also contributed work back to the Git project.
Timeline
• April 2026: Pando went into production after about 10 weeks of development.
• November 2, 2026: OpenAI announced Pando.
The Tech Race
Pando keeps full compatibility with the Git protocol, so existing clients continue to work while the backend underneath is rebuilt. The project sits within a wider race among AI companies to adapt developer infrastructure to machine-generated code volume.
Developers using Git with large shared repositories may see faster push and fetch times if comparable backends are adopted by their teams. The protocol compatibility means existing clients and workflows are not expected to require changes.
The takeaway
Teams scaling AI-assisted development should expect commit volume to grow sharply and plan version-control infrastructure accordingly. Caching and ordered parallel updates are practical levers for keeping repositories responsive.
Further reading
For more on the AI tools driving this demand, explore our Artificial Intelligence coverage.
