I recommend that users of Gemini turn off this "conversation history" feature.

In new conversations, Gemini often forcibly combines historical conversations with the current one, making things up and sneaking in irrelevant information.

For example, if you're discussing how to construct an AI Agent, it suddenly says something like "Should I create an Agent for you based on the historical intent of 'Qin Shi Huang riding a polar bear'?" Anyone would be confused, right?

You might think that enabling this would help the AI understand you better.

But in reality, it pollutes the context of your current conversation, and the output results will be extremely poor!

The bottom line is that current AI technology hasn't yet achieved efficient memory management. Until then, it's best to disable this feature and treat each conversation with AI as a fresh start.