> The author does not understand what LLMs and coding tools are capable of today.

Not really, I would say they used it well and understood the limitations of LLM exactly. No matter how much polished the output or how good it is, LLMs can't build mental models of a codebase like a human does because they are just statistical machines.