New Berkeley paper measured what happens to voice when AI revises prose. Even the "preserve voice" prompt drifted in the same direction.

via

ClaudeAI

New arxiv paper just landed that's worth reading if you're interested in stylometry, AI revision, or the prose-writing strand of the 4.7 discussion. Berkeley researcher Tom van Nuenen ran 300 personal narratives through three frontier models (Claude-class, ChatGPT-class, Gemini-class) under three prompt conditions: generic "improve this," generic "rewrite this," and explicitly "revise this while preserving the original voice." He measured 13 stylometric markers in input and output: function word

More from ClaudeAI

This guy can not be real

ClaudeAI·7

I built /graphify, 26 days, 450k+ downloads, ~40k stars. Here’s what I didn’t expect.

ClaudeAI·6

I dont even have a funny title for this one, I'm just confused

ClaudeAI·3

Whatever SVG/Python based Photo editing claude does is really Good.

ClaudeAI·3

It’s a Weird Time to Be Named Claude

ClaudeAI·3

Claude has a friend?

ClaudeAI·3