Humans Don't Have an API
AI coding tools generate code quickly but require precise human specifications. Since humans lack a formal "API" for communicating intent, the bottleneck shifts from writing code to defining requirements clearly.
Background
- The essay argues that relying on human communication (meetings, docs, Slack) as the primary interface between engineering teams is fragile and doesn't scale — because humans lack a stable, well-documented "API" for interaction.
- An API (Application Programming Interface) is a clean, structured way for software systems to talk to each other, with clear contracts and versioning; humans have none of that.
- The author advocates treating the codebase and its automated tooling (tests, CI/CD pipelines, linters) as the single source of truth, rather than expecting humans to manually coordinate and remember conventions.
- This is part of a broader engineering philosophy sometimes called "Docs as Code" or "shift left" — moving quality enforcement earlier in the process and making it machine-checkable.
- CodeRabbit is an AI-powered code review tool, so the piece also subtly makes the case for automated review systems that enforce standards without relying on human memory or goodwill.