Running a gold mine can be profitable. Finding and digging the next one is expensive. As analyst Benedict Evans argues, Anthropic’s business has the same challenge, with more than 80% of its third-quarter revenue coming from models released in the preceding six months. A lab can make good money serving its best model while spending heavily on the one it will need next, but it could make amazing money if it could go 18 months between major training runs instead of six.
Reality, however, bites.
This pressure to maximize profitability helps explain why frontier labs like Anthropic and OpenAI are aggressively adding functionality to own more of the developer’s workflow. On November 4, OpenAI announced an agreement to buy Pierre, the company behind Code Storage and the open source libraries Codex already uses to browse files and review changes. OpenAI wants Codex to be where developers work, from the first branch an agent creates through deployment of the application. There’s money in extending the life of that gold mine.
The frontier labs are betting that better models will win a developer’s business and that they can keep that business by taking over the complete development workflow. But developers should be wary of turning a preference for this quarter’s best model into a much longer commitment to the company supplying it. As I’ve noted, we need places like GitHub where competing agents can work together, and a practical way to change our minds about which ones we use.
The decisions belong to the project
As powerful as the frontier models are, they would be even more useful to developers if they had greater visibility into developer workflows. Evans points out that Google and Meta can draw on what people already do in their apps to figure out what an agent might do for them. The frontier labs can do some of that, but their coding agents gain more advantage when they can start with the work a team already has underway.
Anthropic’s redesigned Projects feature, introduced in beta in September, shows what that can look 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. If you’ve ever tried to figure out why a seemingly obvious fix was rejected six months ago, you can see the attraction.
That knowledge comes from the people doing the work. A maintainer remembers which customer still needs an old API, just as 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 can find it. Even more importantly for the frontier labs, the next agent should be able to find it, too.
Neutrality is a way to compete
This is clearly where the frontier labs are moving, even though they’ve all benefited from an industry that provided neutral spaces to build. For example, OpenAI developed Pando, a Git back end for its own monorepo, to handle more engineers and agents reading and writing code concurrently. 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.
That industry standardization helps more than any particular lab. A neutral place to build lets an outside contributor using Codex send a patch to maintainers who use Claude. They can discuss it using the project’s tools and run it through the same checks as any other patch. Inside a company, teams that prefer different agents still have to maintain one system together. If only one agent can use the project’s most useful history, its supplier gets a head start from decisions its own customer made.
A neutral space can also foster competitive differentiation. For example, GitHub’s preview of third-party coding agents puts Claude and Codex alongside Copilot in its issue and pull request workflow. It takes 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.
Neutrality, in short, is good for the industry, and can be good for particular businesses. We don’t need the frontier labs to consume development.
Where developers prefer to work
Apple is an obvious objection. Developers put up with plenty of its rules because an integrated environment makes it easier to build and ship software for Apple’s customers. I’d expect the labs to find things they can do better by building the model and the tools together. Some features may work with exactly one model. Still, even Apple has made its own tools more useful by bringing in those of others. GitHub has announced a preview of
[…]
Content was trimmed to protect the source. Please visit the original article for the full text.
Read the original article:
Anthropic and OpenAI are coming for the development workflow