You're forgetting that code quality also requires time. Developers make tradeoffs all the time on how much time to invest into improving the quality of what they write, for both new and existing code. When someone claims that LLMs can produce higher-quality code it can include quality levels that may be unjustifiably slow to hand-craft depending on constraints and needs.

Related - agentic LLMs may be slow to produce output but they are parallelizable by an individual unlike hand-written work.