htmx author's AI debugging postmortem: fast at finding causes, weak at proposing fixes

Working With AI: A concrete example

Carson Gross walks through a real hyperscript bug to show where AI coding helps and where it misleads. Claude found the root cause in minutes: a refactored fetch command accidentally let the expression parser consume the `as` keyword, treating `as JSON` as a type conversion instead of a fetch modifier. But when asked for fixes, AI proposed two subpar solutions—one too hacky and specific to string literals, the other adding an unnecessary context flag when the parser's existing `follows` mechanism already handled it cleanly. Gross nearly fell into the Sorcerer's Apprentice trap of accepting AI-generated complexity without understanding the system.

Why it matters: First-person debug log from Carson Gross showing the classic AI coding boundary: fast at diagnosis, unreliable at fixes. Has concrete code and decision-making, not hand-waving. Score capped at 72 because the topic is frontend tooling, which has narrower reach for an AI-industr...