Some Data Should Be Code
Summary
This article argues that data should be treated as code: Makefiles work for small tasks but fail to scale, so infrastructure and CI definitions should be generated by a real programming language. It uses CDK for CloudFormation and GitHub Actions as examples of moving from hand-authored YAML to code-generated workflows and pipelines. The author advocates writing code to dynamically create build graphs and infrastructure for better safety, maintainability, and parallel processing.