OpenAI has agreed to buy Pierre Computer Company, the software startup behind Code Storage, a Git-based code storage service. The deal is meant to strengthen Codex, its programming agent, as AI agents generate ever more repositories and parallel commits. Financial terms have not been disclosed, and the transaction has not yet closed.
🔑 Key takeaways
- The deal was reported on November 4, 2026. It remains subject to approval, with no closing date or price disclosed.
- Code Storage supports repositories of up to 32 TB and at least 30 simultaneous writes.
- Pierre’s Trees, Diffs and Highlights libraries are already integrated into Codex.
- OpenAI’s Pando engine cuts median code transfer times by nearly 80%, according to DevOps.com.
- OpenAI pledges to keep Pierre’s open source projects open and to keep contributing to them.
A deal announced, a transaction still pending
On November 4, 2026 at 17:56 UTC, The Information reported in a brief by Jemima McEvoy that OpenAI had agreed to buy Pierre Computer Company, Inc., a young software company meant to reinforce Codex. Coverage varies in its wording: The Verge says OpenAI “is acquiring” Pierre, The Decoder speaks of a buyout, and DevOps.com describes a planned acquisition. These terms describe different stages. The agreement has been announced, but the transaction is not complete.
According to The Decoder, the deal remains subject to approval from the relevant authorities. No closing date has been set, and the financial conditions have not been published. Because both OpenAI and Pierre are privately held, the sources consulted report no stock market reaction.
Code Storage: Git infrastructure built for agents
Pierre Computer Company develops Code Storage, a Git-based code storage service, along with several open developer tools. The service stores code and tracks its history through Git. According to DevOps.com, which cites the company, the platform supports repositories of up to 32 TB and at least 30 simultaneous writes. It is designed for workloads generated by AI agents, which can operate repositories directly without developers performing every operation by hand.
| Feature | Reported figure | Source |
|---|---|---|
| Maximum repository size | 32 TB | DevOps.com |
| Simultaneous writes | At least 30 | DevOps.com |
| SDKs | Python, Go, TypeScript | Sources consulted |
| Integrations | GitHub, GitLab, Bitbucket | Sources consulted |
| Named customers | Lovable, Poke, Anything, Bolt, Amp, Parahelp | Code.Storage website |
Customers already rely on the service. The company’s website states that Code.Storage “powers the largest agentic workloads in the industry” for storing code and other machine-generated artifacts at scale using Git. Lovable is presented as a use case requiring many repositories for the projects it generates. According to a description relayed on Hacker News, customers keep generated code, memory files, design artifacts, legal documents and product assets that evolve over time.
Part of Pierre’s technology already runs inside Codex
Part of Pierre’s technology is already live in Codex. The Trees library lets users navigate project files quickly. Diffs displays code changes and lets developers correct errors directly in the diff view. Highlights provides real-time syntax highlighting, both while a developer types and while an agent generates code. The Code Storage infrastructure itself, however, still has to be added to Codex after the deal closes.
“Connecting the code a company ships to the decisions its members make.”
OpenAI, official announcement
Pando, OpenAI’s answer to the limits of Git
On November 2, 2026, OpenAI announced Pando, its own version control engine designed for developers and agents alike. Pando handles changes on separate branches independently, which reduces delays caused by concurrent operations. OpenAI’s previous development infrastructure suffered slowdowns and failures when the growing number of tasks run by Codex sharply increased repository activity. Its reliability sometimes fell below 90%.
According to DevOps.com, Pando cuts median code transfer times by nearly 80% and now processes almost one petabyte of Git data per day. Built in ten weeks with the help of Codex, it has been carrying production workloads since April 2026. OpenAI is evaluating whether to offer the technology to its customers.
A more direct challenge to GitHub
OpenAI presents the deal as a response to how software is now built. Developers and AI agents write code side by side, and agents handle more tasks in parallel, which requires Git tooling capable of supporting far higher concurrent activity. The move positions OpenAI as a more direct competitor to GitHub, which is owned by Microsoft.
- Cursor introduced Origin in August 2026, presented as a GitHub competitor.
- Former GitHub CEO Thomas Dohmke announced a competing offering in October 2026.
- Pierre brings storage capacity sized for agent-generated volumes, while Pando targets the limits of existing Git tooling.
The competition extends beyond GitHub, and it unfolds alongside Microsoft’s role as an OpenAI shareholder. Microsoft has invested more than $13 billion and held 27% of OpenAI Group PBC after the restructuring adopted in 2025. DevOps.com summarizes the stake as “almost a quarter of the capital,” but the 27% figure is more precise. The same source cites 4,500 employees in 2026, $13.1 billion in revenue for 2025 and a $852 billion post-money valuation in March 2026. These figures concern OpenAI and are not the price of the Pierre acquisition, which has not been disclosed.
“Because code by itself has no value. Coding is the easy part.”
mitchellh, Hacker News commenter
Reactions on Hacker News are mixed. Several commenters view the deal as an acquisition of talent, while acknowledging a very capable team. Others point out that Code.Storage is more than an open library: it reportedly has a business model, customers and revenue, even though no figures have been released. Some question the paradox of companies that boost code output with AI while cutting jobs or letting certain skills erode. Others suspect that large AI labs could reuse these teams to document their own production methods.
Code.Storage’s marketing has also drawn criticism. Some commenters find the website unclear, note that it once displayed an error accompanied by a tombstone image, and object to a passage comparing a spill to toxic waste. The discussion also covers the slowness of the diff interface and the decision to hide the largest changes by default. Skeptics see the strategy as a sign that OpenAI lacks internal processes, or as diversification into a developer tools market that is already saturated and unprofitable. Others read the deal as a signal that AI is not yet able to replace these skills.
OpenAI has committed to keep Pierre’s open projects open source and to keep contributing to them. Existing Code Storage customers will keep access to the service after closing. This commitment partly answers concerns raised on Hacker News about the future of the libraries and their maintenance once they are part of an OpenAI offering, with some commenters fearing a slower release pace over time.
Outlook and scenarios
Three scenarios stand out. In the first, the deal closes without obstacles and Code Storage becomes a native building block of Codex, strengthening OpenAI’s position against GitHub. In the second, regulators or integration requirements delay the transaction, leaving more room for Cursor and other players. In the third, the open source commitment translates into steady contributions, making Pierre a credibility lever with developers. Without figures on Pierre’s revenue or headcount, the financial impact of the deal remains impossible to measure at this stage.
Sources
Cet article est publié à titre informatif et éducatif. Il ne constitue en aucun cas un conseil en investissement. Faites vos propres recherches (DYOR) avant toute décision.
