For a long time, a development environment was a laptop with a carefully maintained collection of packages. The machine was personal, stateful, and increasingly precious. We learned how to reproduce the application, but rarely the computer around it.
Coding agents are changing that assumption. An agent does not need a desk, a window manager, or even a long-lived machine. It needs a repository, enough compute, a clear set of tools, and permission to act. The development machine starts to look less like a possession and more like a resource.
A workspace, on demand
The interesting part is not merely moving a Docker container into the cloud. A useful remote workspace needs identity, private networking, persistent storage, snapshots, observability, and a trustworthy lifecycle. It should appear quickly, disappear cleanly, and remain understandable when something breaks.
My ideal interface is deliberately boring: create a Linux workspace, connect over a standard protocol, and run the same commands I would run locally. Everything else should support that loop rather than replace it.
This is placeholder writing for the first version of the site. I’ll revisit the idea with real implementation notes as the experiments develop.