Anthropic and OpenAI Are Buying Their Way Into the Developer Workflow

More than 80% of Anthropic's third-quarter revenue came from models released in the preceding six months. That single figure, cited by analyst Benedict Evans, explains almost everything Anthropic and OpenAI have done since — including OpenAI's November 4 agreement to acquire Pierre, the company behind Code Storage and the open source libraries Codex already uses to browse files and review changes.

The math is unforgiving. A lab can make good money serving its best model while spending heavily on the next one. It could make amazing money, Evans argues, if it could go 18 months between major training runs instead of six. Think of a gold mine: running it is profitable; finding and digging the next one is what drains the treasury.

Owning the developer's workflow extends the life of the mine. OpenAI wants Codex to be where developers work, from the first branch an agent creates through deployment of the application. The Pierre deal is a concrete step toward that.

Anthropic is moving the same direction. Its redesigned Projects feature, introduced in beta in September, shows what full workflow visibility looks like: a coordinator parcels out work to parallel Claude Code sessions that share a memory. The system can remember when a release date moved, why a feature got cut, or whom to ask before touching a service. Anyone who has tried to reconstruct why a seemingly obvious fix was rejected six months ago can see the attraction.

What does this change for engineering teams?

The immediate risk is lock-in dressed up as convenience. Developers should be wary of turning a preference for this quarter's best model into a much longer commitment to the company supplying it.

Evans points out that Google and Meta can watch what people already do in their apps to figure out what an agent might do for them. Frontier labs lack that vantage point — so their coding agents gain far more when they can start with the work a team already has underway. Project history becomes the asset.

That knowledge comes from people. A maintainer remembers which customer still needs an old API; a security engineer explains why a change needs a second reviewer. When a team settles a question in an agent session, the answer belongs in the project's record — so the next person, and the next agent, can find it.

Is neutrality still on the table?

The labs have all benefited from neutral spaces, even as they build around them. OpenAI developed Pando, a Git back end for its own monorepo, to handle more engineers and agents reading and writing code concurrently. Crucially, its engineers kept using Git: OpenAI retained compatibility with the protocol and client and contributed improvements upstream. That compatibility let it rebuild the back end around the way its engineers already worked.

Standardization helps everyone, not just one lab. In a neutral space, an outside contributor using Codex can send a patch to maintainers who use Claude, discuss it with the project's tools, and run it through the same checks as any other patch. Inside a company, teams that prefer different agents still maintain one system together. If only one agent can read the project's most useful history, its supplier gets a head start from decisions its own customer made.

GitHub is testing whether neutrality can be a competitive edge. Its preview of third-party coding agents puts Claude and Codex alongside Copilot in the issue and pull request workflow. It requires a paid Copilot plan, and GitHub decides which agents it supports. Microsoft has plenty of ways to favor its own products — but it also has a reason to make rivals useful on GitHub. If Claude does good work there, the team has one more reason to keep its projects there.

Where will developers actually prefer to work?

Apple is the obvious objection to all of this. Developers put up with its rules because an integrated environment makes it easier to ship software for Apple's customers. Expect the labs to find things they can do better by building model and tools together; some features may work with exactly one model.

Yet even Apple opens its tools to others. On November 4, GitHub announced a preview of Copilot inside Xcode's built-in coding assistant. Apple still gets the apps.

OpenAI or Anthropic could build the place developers prefer to work. The real test comes when a team likes that place but wants a rival's agent for a particular job — say a startup tool that can track down the intermittent test failure that has stumped the team for a week. Give it access to the code and test history the team has authorized, and it gets a chance to prove itself. Ask the team to move platforms first, and it probably never gets to try.

Git gave OpenAI room to build a better back end while its engineers kept working. The next company with a better way to review or test code ought to be able to sell that idea to OpenAI's customers without moving anything. Neutrality is good for the industry — and can be good for particular businesses, too. We don't need the frontier labs to consume development.