OpenAI is attempting to seize the power to define software development in the AI era. On November 4 local time, OpenAI announced it would acquire developer tools startup Pierre Computer Company.

On the same day, Cat Wu (Catherine Wu), the product lead for Anthropic's coding agent Claude Code and its desktop general-purpose work agent Cowork, announced she was leaving Anthropic to join OpenAI, where she will be responsible for product management on the core product team. The capital acquisition and key personnel change reveal that OpenAI is trying to take another step forward in the AI programming field.

OpenAI said that Pierre's developer tools, including Code Storage as well as Trees and Diffs, have already been integrated into its coding agent Codex, while the openness of Pierre's open-source projects is being maintained. The deal amount was not disclosed.

Pierre was founded in 2023 and has raised about $24 million cumulatively. Its main commercial product, Code Storage, is positioned as Git hosting infrastructure for large-scale AI agent workloads.

What is truly valued in Pierre is not a single code tool, but the value of Git as the underlying protocol for code collaboration in the AI agent era. Git is the infrastructure developers use to manage code versions, branches, merges and reviews, and platforms such as Microsoft (ASX: MSFT)-owned GitHub provide hosting, collaboration and permission management on top of Git.

In the past, this process was designed around human developers: a person commits code a few times a day, opens a few branches, and submits a small number of merge requests. But AI agents are changing the pace of code production. A single developer can call multiple agents at once to create branches and modify code in parallel. Rate limits, permissions and review processes on traditional Git hosting platforms, designed around low-frequency human operations, are starting to become bottlenecks.

Also this week, OpenAI unveiled its self-developed Git backend Pando, which supports OpenAI's entire codebase and reduces median code push latency in production by nearly 80%. The Pierre acquisition is meant to fill in another piece of the puzzle. Its Code Storage provides Git-based code storage and history tracking, while Trees and Diffs are used for file browsing and code change review, respectively. OpenAI ultimately hopes to expand Codex from a code generation tool into a complete development platform.

For this reason, the industry interprets the deal as OpenAI entering the heartland of Microsoft (ASX: MSFT)'s GitHub. As AI takes on more coding work, GitHub remains an important code repository and collaboration hub for software development, but AI-native development is creating new infrastructure needs. Besides OpenAI, Cursor and Entire, a new company founded by former GitHub CEO Thomas Dohmke, are also competing for this market.

If the Pierre acquisition addresses infrastructure, Cat Wu's arrival addresses product. Wu joined Anthropic in 2024 and was one of the product leads for Claude Code, working with the product's founder Boris Cherny to help Claude Code grow from an early internal tool into one of Anthropic's most influential developer products. Public information shows that Claude Code expanded rapidly after launching a research preview in 2025, becoming an important product in Anthropic's push for enterprise developers and the AI programming market.

For OpenAI, today's related moves both add Pierre's technical assets and absorb a frontline lead who directly participated in the product growth of competitor Claude Code. Previously, OpenAI's relationship with GitHub had always been unusual. Microsoft (ASX: MSFT) is both GitHub's parent company and an important investor in OpenAI, yet the two sides also compete in areas such as AI programming tools. Now, as OpenAI moves further down into code storage, version control and collaboration infrastructure, it means competition may extend from the application layer to the underlying layer of developer workflows.

However, OpenAI did not announce that it intends to compete directly with GitHub. Its official statement says it hopes to build on developers' existing Git workflows and adapt them to new ways of collaboration between teams and coding agents. As AI agents become increasingly important code producers, what is truly scarce is no longer just a model that can write code, but a full set of development infrastructure capable of carrying massive parallel production, review, storage and delivery of code by agents. From Pando supplementing write throughput, to Pierre supplementing storage and review, to Codex taking on the workspace where developers and agents collaborate, OpenAI's layout is trying to define in advance the interface through which developers review, collaborate on and deliver code, and to consolidate the related processes into a single work entry point, seizing the power to define software development in the AI era.