Version control for everything
Summary
Author argues that non-code use cases of AI-assisted development lack version control, creating stress when tracing changes or reverting state. The piece proposes two approaches: a proxy layer without changing services, or integrating everything into Git to unify code, issues, and docs, highlighting the challenges of atomicity, reverts, and merge conflicts. It also envisions a broader future where VCS-like tooling improves productivity for AI-assisted and non-programming workflows.