AI Output Quality Backlash Is Getting Loud

Two separate threads today about Claude's writing style being so grating that people built tools to fix it. 'Claudette' strips out the BuzzFeed-article cadence, and a related project called 'Vomit' cleans up Claude 5's token bloat. One commenter said Claude is 'on its way to the Microsoft Teams zone of hatred,' which is a brutal comparison. These aren't one-off complaints; they're tools, which means the problem is widespread enough that people put in the work.

Separately, multiple threads on AI-written code comments and blog posts called out the same structural problem: an unmistakable marketing-reveal cadence, repetition for padding, and a kind of false enthusiasm that trained readers now spot immediately. The rebuilding-Electron-in-Swift post was flagged in its own comment thread as AI-written, which undermines trust in the content itself.

The pattern here: the 'AI-blind' thread added a different angle, noting that people are losing the ability to parse AI-generated text critically, experiencing what one commenter described as measurable cognitive atrophy. So the output quality is degrading trust, and the consumption side may be degrading cognition. Both problems are compounding at once.