I don't review AI code. My build does.
Originally published on indiecore.net . There's a ratio I spent a while pretending not to know about. Claude produces a plausible two-hundred-line diff in roughly a minute. Reading two hundred lines of plausible code properly — not skimming for shape, actually checking it — takes me fifteen or twenty. That gap is the whole problem with "never merge AI-generated code without reading it," which is…
Claude generates a plausible two-hundred-line diff in about a minute, but reading such code properly takes fifteen to twenty minutes. The problem with not merging AI-generated code without reading it is that bugs slip through. The author noticed some bugs that careful reading would have caught, such as a dead internal link, an image with no alt text, and a canonical URL pointing to the wrong host.
These issues were caught by a 169-line verifier script written in Node.js. This script checks the built dist/ directory and exits non-zero on anything that must never ship. The verifier runs on every pull request and the deploy job depends on it, ensuring that a branch that fails it doesn't reach the internet. The author stopped trying to read faster because the verifier script could find dull, mechanical bugs in milliseconds.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
