Key Facts

• The author initially set up Gemini's instructions to keep responses concise, use bullet points when appropriate, and avoid unnecessary fluff, which saved them from repeating the same requests at the start of every conversation.
• The author added specific details to their instructions, such as their typical protein and fiber goals for nutrition-related questions, and information about their balcony setup and plants for gardening questions, allowing Gemini to provide more personalized and relevant responses.
• By instructing Gemini to critically evaluate their ideas, point out weaknesses, risks, and alternative perspectives, the author found that the chatbot started providing more balanced and honest feedback, rather than simply agreeing with their ideas.
• The author also instructed Gemini to clearly indicate when it was not confident about an answer, distinguish between facts, assumptions, and speculation, and not present uncertain information as confirmed, which improved the transparency and reliability of its responses.
• The author found that by setting custom instructions, they could shape Gemini's behavior and decision-making process before the conversation started, leading to more accurate and practical responses that saved them time and effort in the long run.