Martin Fowler explores the phenomenon of "LLM Voice"—a detectable, uncanny quality in AI-generated prose that triggers reader revulsion before any detection tool could confirm its origin. He argues that LLM Voice persists even after human editing, making it nearly impossible to expunge once an LLM drafts the initial text; one exception required a professional editor at a major publication to fully remove it. Fowler advises authors to write their own prose, using LLMs only for outlining or verified search, and mandates disclosure of AI assistance on his site. He emphasizes writing as a cognitive tool: the act of wrestling with words forces coherent thought, and delegating writing to LLMs risks delegating thinking itself. Exceptions exist: Jim Highsmith relies on LLMs to overcome Parkinson's disease, and non-native English speakers use them for translation, though Fowler prefers authentic, clunky human prose over smooth but soulless output. He concedes LLMs may suit bureaucratic communication (form letters, unread reports) where the recipient is a process, not a person, citing Michael Nygard's observation that such writing addresses a "cybernetic organism." Ultimately, Fowler insists that human communication demands humanity to "shine (or at least ooze)" through prose, warning that surrendering to LLMs undermines the practice essential to developing writing skill.