OpenAI has swiftly introduced Pando, a new Git backend, through Codex agents within just 10 weeks, replacing the system behind its primary shared code repository This article explores test git pando. . The company claims that agents run around the clock to test both Git and Pando, ensuring the software supporting its expanding AI-driven development workload.
In its November 2022 technical report, OpenAI detailed how one engineer designed Pando’s core, teams implemented it, and a smaller group deployed it.
Production traffic began flowing through the system in April 2026, now supporting all development in OpenAI’s monorepo, where teams keep code together. A single engineer could concurrently handle multiple changes, with each push triggering multiple automated jobs that fetched identical code. The previous setup was inefficient, with slow transfers, failed merges, and reliability often dropping below 90%.
It stores incoming packfiles, Git's compressed bundles of objects, in blob storage and records branch versions and dependencies in a separate metadata store.
The team developed tests that compare its behavior with standard Git, alongside repeatable simulations that verify system rules and expose race conditions. This approach aligns with Codex Security's vulnerability validation workflow and OpenAI's broader Defense Factory security model, both dedicated to investigating flaws and preparing fixes. This scenario highlights how faster AI-assisted development can be combined with continuous testing, maintaining human-centric design and deployment roles.
