3 pointers: AI Prompts
Which prompt has helped you make an impact, and how?
I use our version of Chat-GPT, “ChatITP” mostly to quickly have a starting point for short error dialogs or app messages. I always end up fine-tuning it a bit, but certainly on days without inspiration or when a developer needs a response quickly it helps out tremendously. A recent example of this, that also was a big reminder for me, is this prompt:
Prompt
“Give me a title and a description for a message that states we need feedback from users for our beta.”
Response
“Invitation: Share Your Thoughts on Our Beta Release 🚀”
Description:
"We're excited to announce the launch of our beta version and we'd love to hear your feedback! Your insights are invaluable in helping us create the best possible product experience. Join our beta testing community, explore the new features firsthand, and share your thoughts with our team. Together, we can shape the future of [Product Name]. Sign up now and let's make it even better! 💡”
The big reminder for me was in how ChatITP made it feel personal, like the application is actually speaking directly to the user and how I and the team working on the application sometimes get lost in making sure the correct message with all the relevant information is present that we forget to make it delightful and are addressing a person.
I’ve got my starting point for my beta release message, but unexpectedly I also received the wake-up call to review all the copy in the application and see if we can make it more humane. The fact a machine reminded me of that really puts some things into perspective.
I love using the ‘/describe’ command in Midjourney.
One of the things that many people struggle with in Midjourney is writing good prompts that produce coherent images. And while prompting is still considered a rare skill, Midjourney just made it considerably easier for anyone to reproduce an image they really like.
Midjourney’s /describe command was released shortly after the launch of Midjourney v5.
One of the biggest benefits of the Midjourney describe command is that it offers a way to attempt to “reverse engineer” an image. This is particularly useful when you’ve found an image that you want to use as inspiration or it contains a certain style that you would like to replicate.
For example when you want to create new visuals to expand your illustration set in a specific style of illustrations you have already made. The Midjourney describe command will not only help you re-create something similar, but it can also help you discover new useful words to use in your prompts.
If you’re a developer, chances are you’ve struggled with setting up a CI/CD pipeline at some point. It’s one of those tasks that comes up often enough for developers to know the concepts, but not frequently enough to have the API memorized. Before artificial intelligence changed the game, this meant hours of poring over documentation, deciphering various options, and determining the best fit for the project at hand.
That all changed when I decided to test out ChatGPT. I needed to create a .gitlab-ci file for a yarn monorepo that would run ESLint and TypeScript, so I simply fed my requirements into ChatGPT. It quickly recognized I was asking for code and generated a CI/CD file for me. While the initial result was a bit basic, it provided a solid foundation for follow-up prompts like “Make the lint and compile jobs run in parallel” and “share the yarn install across jobs”. A useful bonus was the ability to ask ChatGPT specific questions about the code it generated, allowing me to better understand unfamiliar clauses, which it explained in context.
ChatGPT and Copilot are a game-changer for these kinds of situations, where you are already familiar with the concepts so you can quickly determine if the output is high quality or plain gibberish.The ability to iterate on AI-generated output through targeted follow-up prompts results in a far more productive workflow than sifting through documentation to find the right solutions.