tl;dr: AI-generated writing (design docs, PR summaries, blog posts, personal messages) is unreadable because readers lack the context the prompter had, and it strips out the voice, vulnerability, and narrative that make writing worth engaging with. The author found AI genuinely useful as a writing *assistant*—for fact-checking against source code, fixing grammar, generating citations, and drafting abstracts—but never for producing prose from scratch. The takeaway: use AI to sharpen your own writing, not replace it, because readers increasingly disengage from (and distrust) authors who ship LLM output as their own thinking.