OpenAI announced that it will acquire Pierre Computer Company, a developer tooling startup, to expand what Codex can do for building, understanding and reviewing code. Pierre sells Code Storage, a Git-based storage service that other coding-agent products use, which makes the deal a question about suppliers as well as about Codex.
What happened
OpenAI’s announcement describes the purchase as expanding Codex’s tools for building, understanding and reviewing code. Pierre describes itself on its corporate site as a maker of low-level, composable product infrastructure for machine work. Its product is Code Storage, and its open-source projects include Diffs, Trees and DiffsHub.
Why it matters
Code Storage’s own site says it stores generated code, agent memory files, design artifacts and legal documents using Git, and names Lovable, Poke, Anything, Bolt, Amp and Parahelp among the products that rely on it. It lists total storage as unlimited, a 32 TB repository size limit and more than 30 concurrent writes. The pages we read do not say how those arrangements change once OpenAI owns the company.
For a buyer, that creates a dependency to map. A company that runs coding agents from one of the named vendors may now have OpenAI’s acquired infrastructure under its generated code, whether or not it uses Codex. Our report on OpenAI’s pooled ChatGPT and Codex allowance covers the commercial side of the same product.
One original insight
Agent platforms keep their output in a repository, and the repository host is becoming a procurement line of its own. Security and legal reviews of an agent vendor usually ask where prompts and model traffic go. They should also ask who hosts the code the agent writes, because that is where the company’s new intellectual property accumulates. Ask each coding-agent vendor in use for its storage subprocessor, and ask whether its terms change if that subprocessor is owned by a competing model provider.
Source: OpenAI News
