Tonight, I want to touch on an important topic - prompting a tool. The term “prompt engineering” has been a popular one for some time, and has recently been replaced with “context engineering”, which is a more comprehensive concept, where we feed not only a good prompt, but also session memory and RAG databases or other types of storage. I still feel though that nothing, NOTHING matters as much as providing the model with a good set of instructions designed to provide the answer you want. I call it “programming in English”. Of course, it COULD be “programming in your language of choice”, since LLMs are multi-lingual. Providing explicit instructions to the model will result in the best responses from the AI.
So how do I get started? How do I write the best prompts? What tools can I use to create them? Most important - what do I DO with the prompt to create the kind of system i want to build? Let’s dig into these questions by walking through developing my Idea Engine app. This will be web based but viewable on phones and other mobile devices. It will consume information from around the web (RSS feeds, etc) and allow me to act as the HITL (Human in the Loop) to decide how those pieces of news fit into my content strategy, then write content in my voice and publish it on the appropriate platforms. Sounds like a lot, right? Here’s the key - if I can describe it, I believe the AI can create it. This means that my primary roles in creating the end products are 1) having a vision, and 2) describing that vision vividly.
With that, I’m going to pause this episode. This is a play-along. Now that you’ve heard MY app idea, I want you to come up with your own. Here’s the assignment:
Give your app idea a short description - something like, “this app helps people calculate their daily caloric needs and macros.”
Think of ways to use AI tools to help you develop a detailed description of the app - describe the screens in the app, the shapes, sizes, images, and colors you want to use for branding, the specific functionality of the app, even the tech stack you prefer (Supabase vs Google Sheets, for example). You’re looking for a detailed description. Imagine yourself going into a developer’s store and asking him to build your app. How would you describe it? That’s the kind of description you want here.
What are the key data inputs, processing steps, and outputs? Knowing the way that information has to flow through your app is a huge win when it comes to design. You know what has to happen and what the user will be doing in the app, so now you just need to design the interface that lets them do it most effectively.
Find AI tools that will allow you to build the app. I’ve used or researched many of them (I’d say MOST, but the list grows every day) and can provide some good feedback to help you evaluate the creation tools. As an initial tip, look for tools that can be used for front-end and back-end development. Most specialize in one or the other - try to figure out the difference and bring your favorites.
If you’d like to share this step - email it to me at wayne@freebeer.ai
That’s it - that’s all you’re getting in today’s episode. Tomorrow, I’ll share my own prompting, and toolset with you.
Wayne