Skip to content

How I audit the code a model writes

Image pending

A model writes in two seconds what then has to be read, understood, and tested. That part doesn’t get delegated.

Generated code usually compiles and usually looks tidy. The problems live in what you don’t see at first glance: the input nobody validates, the error that gets swallowed, the new dependency nobody asked for.

This is the review I do before approving a change that came from a model.