You're iterating quickly on experimental AI prototypes and want a repository structure that lets you move fast without creating chaos. The challenge is balancing short-lived experimentation with enough organization, reproducibility, and shared context that other engineers can understand, reuse, or productionize what works.
How do you structure your code repositories when iterating rapidly on experimental AI prototypes?
