How Do You Structure Your Prompts for Complex Tasks?

I've been including prompts in my project instructions to enhance clarity. I ask the AI to explain the basis of its responses, indicate its confidence level, provide citations, and respond critically to avoid biases. This has really helped improve the responses I get.

Remember, every complex task can be simplified into smaller subtasks, which is vital when working with AI. I often use AI to help break down these tasks into manageable parts and refine them with detailed requirements before tackling them in context.

I usually start by asking the AI for input and then asking it to iterate on its suggestions. It helps to break down the complexity into smaller steps, so I'm sure to emphasize step-by-step instructions. I also tell the AI to ask me if there's any uncertainty about a particular step, which helps clarify things as we go along.

I prefer to take my time with complex outputs, breaking them down into smaller, verifiable parts. I usually start with an overarching project statement, like building a learning model for differential equations. Then, I gather feedback to refine the plan before tackling specific components, making everything easier to test and transfer to new chats if needed.

One tool I’ve found super useful is connecting Linear to Augment Code for task management. It allows me to create tickets for complex tasks, keeping historical data handy and facilitating discussions about planning as I develop my prompts. Once I have everything organized into tickets, I can efficiently move forward without hitting prompt limits.

If you could hire a top-tier consultant for just one hour, what crucial information would they lack that could make a significant difference? Having that ready can really guide your prompts effectively.

Could you share some examples of complex tasks you’re working on? I’d love to help draft a few prompts and discuss their structure with you.