AI Coding Took Off Because the Tooling Was Already There

Coding has taken off as a use case for AI because of the prior existence of mature tooling. The software industry had tools for creating (files, CLI, scripting), collaborating (git, pull requests), verifying (linters, test suites), and measuring success (compilers, logging, analytics) — all before AI wrote a single line of code.

When AI started generating code, it landed in an environment already designed to absorb, validate, and correct machine output. The feedback loop was there.

Most other industries don’t have any of this. There’s no git diff for a building design (I think?). No pull request for a film edit. No CI pipeline that validates a legal argument.

The fields that get comparable tooling for creating, verifying, and measuring will be where AI appears to suddenly “get good.” But it won’t be the AI that changed. It’ll be the infrastructure around it.