Hi everyone, I’m new to the forum and had some thoughts on prompt engineering. I’ve been doing it for a few years and these are a couple things I’ve picked up along the way. I’d love feedback and your ideas as well.
Prompting Philosophy: Unlocking Generative AI
1. Problem-First Approach
• “When I sit down to prompt, I think about the problem but also my hypothesis for a solution.”
• Start with a clear problem and hypothesize how elements of the solution fit together. Generative AI helps you test and tweak these hypotheses quickly, like assembling a puzzle.
2. Iterative Refinement
• “Prompting is a back-and-forth between what I receive and how close I think I am to the solution.”
• Evaluate outputs iteratively:
• Adjust specific sections.
• Zoom out to learn or explore.
• Refine until the output aligns with your goal.
3. Zoom In and Zoom Out
• “I think a lot about zooming in and out—how often I need to direct AI and how often I let it explore.”
• Zoom In: Focus on granular details (e.g., improving code snippets or rewriting content).
• Zoom Out: Open AI to broader possibilities or conceptual insights (e.g., brainstorming approaches or exploring creative solutions).
4. First Principles Thinking
• “Generative AI doesn’t replace thinking—it sharpens it. As a copywriter, I understand sentence structure, logical flow, emotional appeal, and more. I take that skill and multiply my efforts with AI.”
• Leverage your foundational knowledge to guide AI, ensuring quality and relevance in its outputs.
5. Structure Meets Creativity
• “I use structure in thinking through problems but stay creative in how I explore solutions.”
• Use clear, logical prompts for problem-solving, but allow flexibility for ideation and brainstorming.
6. Breaking the Rules
• “Sometimes I start without the end in mind and just ask AI for its views upfront. This zooming out frees up potential routes to a solution.”
• Don’t always narrow AI—let it explore freely, especially for creative or high-level strategic tasks.
7. Validation and Responsibility
• “You have to be involved to win at prompt engineering. Thinking AI replaces skill is dangerous—it leads to complacency.”
• Always validate data and sources, particularly in technical or high-stakes outputs. Stay engaged in the process to maintain accuracy and ethical use.
8. The 3 Cs Framework
• Clarity: Be specific about what you need.
• Example: “Write a Python script to scrape headlines from a website.”
• Context: Provide enough background to guide AI.
• Example: “The site uses dynamic content. Use Selenium for scraping instead of BeautifulSoup.”
• Creativity: Use open-ended prompts to explore alternatives.
• Example: “What are three unconventional approaches to optimizing this algorithm?”