The spec replaced the repo

For twenty years the repository was how software travelled. You shared a URL, someone cloned it, and the code was both the artifact and the proof.

That has come apart. The spec is now what travels — a few files describing intent, constraints and build order — because anyone who receives it can generate the implementation. What you hand someone is no longer the thing; it is the instructions for the thing.

But the proof moved in the opposite direction. When generating a plausible codebase is cheap, a repository stops being evidence that anything works. The only thing that still carries weight is the artifact actually running.

So the two halves separated. Distribution got lighter and evidence got heavier, and a lot of the habits built around the old arrangement — code review as the quality gate, the README as the pitch, stars as the signal — are measuring the wrong half.