Users Report Claude's Writing Has Become Weird and Hard to Read

altryne · x · 2026-08-01

Multiple users have recently noticed that Anthropic's Claude models are generating text that sounds increasingly "weird" and suffers from reduced legibility. Cited discussions highlight that models released after Opus 4.6 exhibit unnatural phrasing, making them harder to read quickly in both coding and conversational contexts. The community speculates this could be an unintended artifact of reinforcement learning (RL) training.

Related event: Users Report Severe Degradation in Claude's English Writing Style(2 posts)→

Original post →

More from Models

Models channel →