From Sycophantic to Condescending: Users Demand Straightforward AI Tools

bendee983 · x · 2026-08-07

A developer points out in a tweet that the narrative criticizing Large Language Models (LLMs) is shifting: complaints have moved from models being "too sycophantic" to criticisms that "Claude Opus 5 is too condescending and infuriating."

He calls for an AI assistant that provides straight, non-verbose answers without attempting unnecessary socialization, rather than swinging between extremes of anthropomorphic tones.

Original post →

More from Models

Models channel →