#prompting

A practical lesson: models confidently present wrong things as right. Here is a prompt you can paste directly, which forces it to weigh evidence for and against based on authoritative literature, news and public discussion, give sources, separate facts from speculation, and state what information is still missing. Usage note: many models do not actually browse the web, so check the sources yourself.

Beginner tutorial: why AI-written text often has a distinctive tone, and how to remove it. The core principle comes from Anthropic's internal guide on removing AI flavor: if you can say it directly, say it directly. At the end is a prompt that works in both Chinese and English and suits any text model; paste it as is.