AI is like an eager assistant who wants to help but has no clue what you actually want until you tell it. If you ask for a “good design,” you’ll get something generic and uninspired. If you ask for a “moody, futuristic UI layout with high contrast and deep blues, using this reference image and constrained to X design system with X elements in X hierarchy and in X style,” you’ll get something much closer to what you imagined. The difference? Specificity and intent. 

“A problem well stated is a problem half solved.” - Charles Kettering

I’m loving vibe coding, throwing ideas at AI, letting it generate, and seeing where it takes me. But I’ve also learned that if I don’t structure my thoughts before prompting, things get messy fast. Sure, AI can fill in gaps, but if you don’t steer it, it will hallucinate, contradict itself, or take creative liberties that completely miss the mark. When you’re generating images, that unpredictability is fun. When you’re writing code, it becomes a problem fast. No you don’t need to know how to code anymore (but yes I agree, it still helps a ton), but you DO need to know how to think. Understanding the foundational concepts that drive what to code is still really critical.

Focus is key. If you ask an AI to write a novel, you’ll get a wall of mediocre text. If you ask it to brainstorm compelling character backstories, refine a plot point, or punch up dialogue, suddenly, you’re getting meaningful results. The same principle applies to vibe coding and image generation. The best results come when you break things down step by step, using AI as an iterative tool rather than a one-shot answer machine.

One thing AI won’t replace is the need to communicate well. You can have the best idea in the world, but if you can’t express it clearly, AI won’t magically read your mind. It’s like working with a team. The better you can articulate your vision, break it into actionable pieces, and guide the process, the more valuable you become. Learning to take an abstract idea and describe it in a structured way takes practice, but it is a skill that translates across every AI tool you use.

“If you can’t explain it simply, you don’t understand it well enough.” - Albert Einstein

Vibe coding feels a lot like being the product owner of a very fast jr development team. A lot of the prompts I end up writing could just as easily be tasks handed off to real developers. The clearer I make my intent, the better and more reliably AI performs, just like a real team.

There’s also a huge parallel between prompting for different types of AI outputs. If someone is great at prompting for image generation, they’re often better at prompting an LLM to generate text, music, or even code. It’s not just about the tool, it’s about how well you can express what you want in a way AI can actually use. The more you work with a specific model, the more you “vibe” with it. Sounds weird, but it’s like working with the same person for a long time. Even if you’re a great communicator, every person, and every AI, has its quirks. The better you understand them, the better results you’ll get.

Some people talk about “Prompt Engineering” like it’s a career path. I don’t buy it. Marketing yourself as a “Prompt Engineer” feels a lot like calling yourself a “Google Search Engineer.” Yes, some people are bad at it, and yes, it is a skill, but I wouldn’t put it on a resume. I’d show it in the results I’m able to create vs a line item.

Knowing how to collaborate effectively, whether with people or AI, is a skill that has always been valuable. And as AI continues to evolve, it will only become more important, especially for those looking to move beyond individual contributor roles.