Contradictory Writing Feedback Between Claude Versions Highlights Lack of True Taste
Rosoll · reddit · 2026-08-01
A children's novelist shared their experience using Claude for writing feedback: while the older version (Opus 4.8) harshly criticized the opening of their third chapter, the upgraded version (Opus 5) praised the exact same passage as the strongest part of the book.
This contradictory feedback made the author realize that Claude lacks genuine literary taste; it merely generates plausible-sounding critiques that shift with model weights. The author suggests treating AI feedback with a huge grain of salt—useful for positive reinforcement and motivation, but not to be blindly followed. Additionally, the author uses a memory rule to prevent Claude from generating original text or ideas, keeping it strictly as a thinking partner and research assistant.
More from Models
- Grok Voice 3.0 tops speech-to-speech index, undercuts OpenAI on price — VectorWireAI · 2026-10-16
- Microsoft completes MAI-Base-2, a 2.8T-parameter MoE trained on 16k GB200 GPUs — VectorWireAI · 2026-10-16
- Gemini Pro users hit sudden usage paywalls: 100% daily limit despite 6% weekly — Strange-Plate581 · 2026-10-16
- Users report Gemini's memories can't actually be deleted even after wiping chat history — ThrowRA_Purpose2025 · 2026-10-16
- OpenAI stealth-testing mysterious internal model 'Aeon' in Codex desktop app and Dots — cedric_chee · 2026-10-16
- Argon edges out astra as daily driver, but Claude still wins heavy coding work — iruletheworldmo · 2026-10-16