Thinking about this after a lot of time in scaling docs. In most stacks the move from “works locally” to “runs for many users on many machines” means touching a lot of things that have nothing to do with what the app does: auth middleware, a tenancy filter on every query, serialization schemas, deployment manifests.

We work on Jac, and its approach is to treat that as a runtime concern, the way garbage collection is. The promise is that the same program runs as a local script, as an API server, and on Kubernetes, with walkers (its functions that move through a graph of data) exposed as endpoints and each user’s work scoped to their own data.

It doesn’t pretend the hard parts away.…