On November 4, 2026, OpenAI announced that it will acquire Pierre Computer Company, a developer-tools maker. OpenAI plans to integrate the features of Pierre's code storage service, Code Storage, into its AI coding tool, Codex. The acquisition is subject to the usual closing conditions, including regulatory approval, and the two companies will continue to operate independently until the deal closes.
Pierre's technology is already used in parts of Codex. Trees, which lets users browse files, and Diffs, which compares changes in code, are both in use. With this acquisition, the scope of integration expands beyond those viewing and review features to the infrastructure that stores code and change history. OpenAI also runs its own Git backend for internal development, called Pando, so how the two relate will be worth watching.
From Code Viewing to Storage Infrastructure: A Wider Scope of Integration
According to OpenAI's acquisition announcement, Trees and Diffs, both developed by Pierre, are already built into the Codex app. Trees is a feature for browsing files within a project, and Diffs compares and displays code changes. In other words, this acquisition is not the first time Pierre's technology has been adopted in Codex.
Pierre also develops other tools that support code editing and display. Diffs' Edit mode lets users fix problematic lines directly while reviewing changes. "Highlights" emphasizes syntax according to code a developer is typing or code an AI agent is generating. However, OpenAI has not said that these features are currently used in Codex.
What is newly announced is the plan to bring Code Storage's capabilities into Codex. Code Storage is a service that uses the distributed version control system Git to let applications store and manage code and change history. While Trees and Diffs mainly handle viewing and reviewing code, Code Storage handles storing the code itself and managing its history.
Based on OpenAI's acquisition announcement, its technical write-up on Pando, and the Code Storage product site, the roles and current status of each are summarized below.
| Technology / Service | Main role | Current status |
|---|---|---|
| Trees / Diffs | Browsing files; comparing and displaying code changes | Already adopted in Codex |
| Code Storage | Storing and managing code and change history via Git | Offered as an external service; integration into Codex planned |
| Pando | Git backend supporting OpenAI's large monorepo | Running inside OpenAI |
These are technologies with different roles, and it is important to distinguish between what is available now and what is planned for integration. In particular, Pando is an in-house foundation OpenAI developed itself, not a technology it is newly acquiring through Pierre.
OpenAI says existing Code Storage customers will be able to keep using the service after the acquisition closes. It also plans to keep Pierre's open-source software publicly available and to contribute to its development. However, it has not disclosed an expected closing date, the timing of integration into Codex, or pricing and packaging after integration. According to Axios, financial terms such as the purchase price have not been made public either.
Parallel AI Agents Put New Strain on Git Infrastructure
Key to understanding this acquisition is how the spread of AI agents is changing software development. Even after an AI generates code, the work isn't done. Changes must be saved, merged with other work, and deployed to test or runtime environments. In its technical explanation of Pando, OpenAI says that when many AI agents work in parallel, read and write requests to Git surge.
In Git, developers and agents can advance changes on independent lines of work called "branches." Sending changes to a remote repository is called a "push," but if many agents update different branches at the same time, write operations can conflict. If every update funnels through a single coordination process, wait times grow and the efficiency of parallel work drops.
The load of distributing the same code and change history to many agents and execution environments is also significant. Pando, developed by OpenAI, coordinates updates per branch and processes writes to different branches in parallel. It also reuses already-generated Git pack data to distribute data to many clients more efficiently, while maintaining compatibility with the existing Git protocol and clients.
Pando is the foundation that supports a "monorepo," in which multiple development projects are managed in a single repository. According to OpenAI, migrating to Pando cut the median time to push to its internal monorepo by about 80%. However, this reflects Git write performance in a specific internal environment; it does not mean that AI models generate code 80% faster or that the overall development process takes 80% less time.
Pando and Code Storage both handle Git, but for now they are separate technologies and services. Pando's track record shows that more AI agents demand higher throughput from code storage and distribution infrastructure, but it does not directly demonstrate Code Storage's performance or the effects after integration into Codex.
How Might Competition With GitHub Change?
Axios characterizes the acquisition as a move by OpenAI to take on Microsoft-owned GitHub. If Codex can handle not only generating and reviewing code but also storing it and managing its history, the range of tools developers use could broaden. However, this announcement alone does not support the conclusion that OpenAI intends to push developers to move off GitHub entirely.
Code Storage is currently designed to work alongside existing Git services. According to its product site, it supports syncing with GitHub, GitLab, and Bitbucket, and it also supports standard Git operations such as clone, push, and fetch. It offers SDKs for TypeScript, Python, and Go so that applications can work with code and change history. Whether this compatibility with existing services will be maintained after integration into Codex will be an important point for developers.
On the technical side, the difference in consistency design between Pando and Code Storage is also notable. Pando prioritizes independent updates per branch, so the result of reading multiple branches does not necessarily represent the state of the whole repository at a single point in time. Code Storage, by contrast, says it ensures durability and consistency of reads and writes through a three-phase commit process and consensus among multiple servers. However, the technical relationship between the two and how they might be combined have not been disclosed.
The focus going forward is where code generated and reviewed in Codex will be stored and how it will connect with existing Git repositories. It also remains unclear whether Code Storage's current sync features with external services will be maintained after integration, and how pricing and packaging might change. These terms will help determine how seamlessly developers can handle everything from code generation to review, storage, and history management within Codex.