OpenAI Says Its In-House Git Backend Pando Has Taken Over the Main Code Repository

Article

OpenAI has also taken action to transform Git: Codex participated in rebuilding the backend, launched in 10 weeks, with push speeds nearly 5 times faster

Dongcha Beating AI Flash News: As AI writes code faster and faster, the Git backend is starting to buckle under the pressure. Following Cursor and Anthropic, OpenAI has also announced its self-developed Git backend, Pando, which has now taken over the company's entire main code repository.

Over the past 1 years, Codex has enabled engineers to deploy large numbers of Agents to write code simultaneously, causing code branches, pushes, merges, and pull requests to skyrocket. OpenAI's original system experienced frequent lag and merge failures, with reliability even dropping below 90%. As a result, one engineer was responsible for the core architectural design, while multiple Codex Agent teams participated in the implementation, taking only 10 weeks to bring Pando into production.

Pando allows code pushes from different branches to be processed in parallel, eliminating the need to queue up for each other. When pulling code, it no longer generates data packets individually for each client; instead, it reuses existing Git data packets and distributes them directly through caching. After the transformation, the median latency for code pushes has been reduced by nearly 80%, distributing nearly 1 PB of data daily, with a cache hit rate exceeding 99%. However, different branches are no longer guaranteed to present the state at the same moment.

The three companies are solving similar scaling challenges, but their choices are vastly different: Cursor's Origin is still based on Git and primarily aims to replace GitHub's code hosting services; Anthropic's Scion completely abandons Git, moving even commits, merges, and history queries to the server side. OpenAI chose to retain the Git client and protocol, rebuilding only the backend, so the original development workflows for engineers and Agents remain largely unchanged.

Original link https://m.theblockbeats.info/flash/375180

Key points

Dongcha Beating AI reports that OpenAI's in-house Git backend Pando has now taken over the company's entire main code repository.
Reports indicate that over the past 1 year, Codex has enabled engineers to deploy a large number of Agents simultaneously to write code, resulting in an increase in code branches, pushes, merges, and pull requests.
The report states that OpenAI's existing system frequently experienced lag and merge failures, with reliability dropping below 90%.
One engineer took charge of the core architecture design, multiple Codex Agent teams participated in implementation, and Pando was put into production within 10 weeks.
Pando allows code pushes from different branches to be processed in parallel, and reuses existing Git data packages when pulling code, distributing them through caching.
After the overhaul, median push latency was reduced by nearly 80%, nearly 1 PB of data is distributed daily, and the cache hit rate exceeds 99%.
Different branches are no longer guaranteed to reflect the state at the same point in time.
The article states that Cursor's Origin is still based on Git, Anthropic's Scion abandons Git, while OpenAI retains the Git client and protocol and rebuilds the backend.

AI-assisted interpretation

The following is analysis, separate from reported facts. Verify important claims independently.

Pando essentially replaces the backend processing part of OpenAI's internal Git system while retaining the Git client and protocol that engineers and Agents are familiar with. It reduces wait times and expands processing capacity through parallel push processing, data package reuse, and cache-based distribution.

Why it matters to readers

If the data in the report holds up, Pando demonstrates that large numbers of coding Agents can significantly increase code repository load, and that rebuilding only the backend can improve processing speed without substantially changing existing development workflows.

Beginners can understand it this way: the tools for writing code haven't changed, but the backend system responsible for receiving, storing, and distributing code has been redesigned to handle more automatically generated code operations.

Risks and unknowns

The material does not include OpenAI's original announcement or technical documentation; Pando's deployment scope and performance data remain to be verified by primary sources.
It remains unclear what impacts the lack of synchronization across different branch states will have in actual development.
The material does not provide information on Pando's long-term reliability, maintenance costs, or whether it will be made available externally.

OpenAI Says Its In-House Git Backend Pando Has Taken Over the Main Code Repository

Article

AI-assisted interpretation

Why it matters to readers

Risks and unknowns

Related Developments

Related concepts

Keep following

Security Center