LLM-polished messages erase authentic human voice

This essay cuts at something most AI discourse ignores: the receiver's experience of AI-smoothed text, not just the sender's convenience.

• –Aigner's core insight is that personal writing is a signal layer — accumulated over years, it lets recipients read emotional nuance from word choice and cadence that AI normalization erases
• –The HN thread surfaced a key distinction: AI as a writing scaffold (brainstorm, then rewrite in your own voice) vs. AI as a voice replacement — the community broadly accepts the former, rejects the latter
• –The irony of scaling text communication efficiency only to lose communicative bandwidth is real: over-polished messages waste reader time and strip the metadata that makes them worth reading
• –This tension between productivity gains and authenticity costs is becoming a genuine design question for AI writing tools — prompting the question of whether tools should warn users when output sounds too generic