Users report fluctuating performance from large language models like Claude Opus, often seeing degradation during peak hours. Explore theories ranging from dynamic model routing to compute allocation, and discover user-shared strategies to mitigate these issues.
Discover why AI models frequently use em dashes in their writing, stemming from training data and auto-correction, and learn practical keyboard shortcuts to type them yourself.
Users are observing AI models like ChatGPT and Gemini displaying 'thoughts' in non-English languages. This discussion explores why this happens, linking it to multilingual training, internal token efficiency, and research findings that suppressing it can even reduce performance.