AI in the creative process and the curse of complicity
Looking for meaning in a world where I don’t have to learn anything, Part 2
Creating something used to be a hassle, but the real meaning and joy came out of that painful process.
It wasn’t enough to have an idea. You had to shape it into something you could actually build with the skills you had at the time, and that made you feel knowledgeable. You couldn’t just summon walls of text from a vague outline. You had to write, rewrite, review and get feedback, and that made you feel accomplished. Sometimes you had to branch off for a week or two to pick up a new skill. Boom, intellectual fulfillment.
And we thought these hurdles were what slowed us down. We didn’t realize they were where the real thinking happened, the kind that uncovers needs no one thought to ask for. That’s what made it meaningful: value for our users and viewers, and pride for us.
If only, we would say, if only I had more time, if I could learn React faster or if I knew how to design in Figma, I could build my project faster! And then, our prayers were answered. AI tools arrived. We were free of the required learning.
Yet, there aren’t any life-changing, mind-blowing products emerging every day. How come? You could think it's in AI's nature to generate slop, and you wouldn't be wrong to a certain point. But I don’t think it’s “because of” AI. It’s because of us as the human in the equation.
Many of us try things we wouldn’t have tried before generative AI. As inexperienced designers, developers and copywriters, we lean on Claude, Gemini or whichever LLM, to call the shots. We ask it to create “a user-friendly interface” without knowing what makes an interface friendly. The result looks good enough but we're left feeling empty and what we build feels disposable.
The way back starts with taking charge.
We’re still the adult in the room. An LLM is only a brilliant assistant. It needs to be given direction and feedback. To give direction, we need to have direction ourselves. We need to be informed, and that comes down to three things.
Know your niche. Every field has its own unwritten rules. In DevRel, for example, developers can smell a sales pitch in the first sentence. An LLM knows a little about every niche and a lot about none. When you know yours, you’ll notice the moment its output drifts towards the generic middle.
Know your audience. Your niche is the field; your audience is the people in it you’re actually talking to. A senior engineer and a bootcamp student both watch developer videos, but they don’t want the same video. An LLM can’t sit in on your user calls, read your comments or draw on years of talking to people face to face. That context is yours to bring.
Know the fundamentals. You don’t always need to be an expert, but you need enough of the basics to judge the result. What makes an interface easy to use? Why do viewers drop off in the first 30 seconds of a video? Without the fundamentals, every output looks fine, and “fine” is how we end up with slop.
An LLM can draft your next launch video: the script, the animations, the transitions, even the thumbnail. But does it know how your brand sounds to your audience? Does it know what kind of videos your audience connected with the most? Making the small decisions that add up to a good result is still our job.
I know there is a trend to delegate this work to LLMs too: asking ChatGPT for the 10 most important principles of user-friendly design and then asking Claude to design with them. It's just a faster way to arrive at the AI slop station if you ask me.
I don't mean stop using AI tools and go back to manual work. Instead, use AI to learn as you create. Ask it why it went a certain way, have it critique your ideas and try to understand what assumptions the decisions were based on. Both the results and you will be better for it.
But being informed is only the start. We need to lean in and be opinionated. Let the process inspire us. Make room for creative freedom. Which brings me to my next point...
