Loading request...
User suggests a feature where an LLM can help critique and iteratively improve prompts for other LLMs or agents, based on feedback and output review. Mentions using Copilot to write better prompts for Copilot Analyst.
LLMs and chatbots have come a long way. What makes for good prompting has as well. Here are some tips. First of all, don't think about it as writing one great prompt. You are having a conversation, so iterate your way to a good solution. Ask questions, provide more background as you go. Second, context is everything. Describe your situation, the task, the style you want it in, in as much detail as possible. Again, iterate if the answer isn't great yet. Third, work with the enhanced context features of chatbot apps. Upload documents (create projects for different tasks), use the memory feature (selectively, don't fall into an echo chamber). Finally, jump between sessions. If a back-and-forth session produced good material but doesn't progress. Copy the best bits to a fresh session as context and start again. And remember, the only truly intelligent party to the conversation sits in front of the screen. The more you contribute, the more you will get out of it. Sandra Peter and I have written up these tips (and more), in our latest The Conversation Australia + NZ piece. https://lnkd.in/g34jVviG #ai #genai #prompt #prompting #contextengineering #ChatGPT