Library Installs Do Not Equal Customers

The key point is that Pierre has built a wide top of funnel, not yet a proven revenue base. Trees and Diffs can spread quickly because they are free UI components that any developer can drop into a coding product, while Code Storage is a separate metered infrastructure purchase that asks a company to trust Pierre with repositories, traffic, and long term operating costs. Codex use shows product quality and distribution, not paid storage conversion by itself.1 2 3

• The split is structural. Pierre gives away the parts a developer sees on screen, file trees, diffs, comments, and edit controls, but charges for the hidden backend that stores repos, serves clones and fetches, replicates writes, and archives idle data after seven days. A team can adopt the first layer without ever buying the second.1
• That pattern is common in agent infrastructure. E2B publishes an open source runtime but makes money when customers pay for managed sandboxes, and it explicitly notes that large signup counts do not prove paid enterprise use. Daytona follows a similar model, using SDK adoption to seed demand for a billed execution service.4 5
• The strongest signal would be workflow lock in, not npm or pip installs. A paying Code Storage customer routes repository creation, branch writes, sync jobs, and network traffic through Pierre’s API inside its own product, similar to how GitHub and GitLab monetize hosted Git after teams entrust them with the system of record for code.1 2 6 7

The next step is whether Pierre can turn front end developer love into backend ownership of agent generated repositories. If Codex and other AI coding products standardize on Git native storage behind the scenes, the winners will be the vendors that become the default system of record, not the ones whose libraries are merely installed in the interface.1 2 3 6 7