Most repositories are built for people who already know them.

That works for a while. The maintainer remembers which setup step matters. The team knows which script to run first. Someone understands why the README says one thing but CI does another. The repo is "runnable," but only because enough context lives outside the repo itself.

That model is starting to break.

More software work now happens through systems that do not have that context: CI jobs, ephemeral dev environments, remote runners, automation scripts, and coding agents. They do not know the history of the repo. They cannot ask the maintainer what changed last week.…