Per-developer environments were the goal. Agents moved the goalposts.
Multi-tenancy has moved in one direction for 60 years: the tenant keeps getting smaller. Mainframe time-sharing carved a single machine The post Per-developer environments were the goal. Agents moved the goalposts. appeared first on The New Stack .
The goal of platform engineering was to provide a per-developer environment for each coder. Initially, mainframe time-sharing divided a single machine among an organization's departments, making the tenant the organization. Virtualization and Kubernetes namespaces further divided, making the tenant the team or individual developer.
However, coding agents have changed the dynamics. A developer running multiple agent sessions simultaneously needs its own working version of the system, leading to the realization that the tenant is no longer the developer but the change itself. Tenancy demand now scales with the number of changes in flight, not with headcount.
Under this new assumption, a 50-developer organization can have hundreds of changes in various stages of validation, each requiring its own resources. This shift reveals that treating the change as the tenant allows for a more efficient platform design. Change-level tenancy focuses on creating tenants near-instantly, isolating only what changed, and binding the tenant's lifecycle to the change itself.
This approach mirrors how multi-tenant production services operate, with tenants sharing the substrate, privately owning distinct parts, and being provisioned self-service. Implementing change-level tenancy means changes become tenants of the development platform, owning only what changed and sharing everything else. This enables scaling to handle the demands of a large engineering organization without the overhead of per-developer namespaces or seat-based capacity plans.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
