Essays on intelligence: artificial, human, and everything in between.

Once upon a time… My freshman college 2D Art Professor — let's call him 'Bob' — handed us canvas, a rudimentary bill of materials, and what I can only describe as a psychological endurance test disguised as an assignment: recreate a Rembrandt painting using the same materials and techniques the master would have used.

For far too long, I mixed pigments, studied brushwork, and tried to capture that signature interplay of light and shadow that makes Rembrandt, well, Rembrandt.

It was maddening. Not because it was hard — though it was — but because every brushstroke felt like I was painting someone else's vision while my own ideas sat in the corner, tapping their metaphorical feet and checking their watches.

I could appreciate what I was learning. The way oil paint behaves. How light creates weight on canvas. The deliberate choices behind every compositional element.

But the whole time, I kept thinking: When do I get to paint MY thing?

Then, a few weeks later, I was working on my own piece. Something original. Or so I thought.

Bob walked by, paused, and drawled: "I like that… Do you like that?"

"I guess," I said.

"Do you like DaVinci?"

"Sure," I said, wondering if this was a trap.

"DaVinci applied shading like that… Are you DaVinci, Mr. Maxwell?"

I muttered, "No."

"Then don't shade that way." Then he meandered on to critique a classmate's work.

The irony wasn't lost on me: Bob, who'd made me spend weeks copying a master was now calling me out for accidentally copying a different master. I remember sitting there thinking: Does any of this matter? If everything I do is just echoing someone who came before, what's the point?

Fair warning: this is going to take more than 90 seconds to read. If you're looking for "5 Quick AI Tips," this isn't it. But if you've ever wondered why all those prompt libraries aren't actually making you more creative — or if you've been collecting templates and still feel stuck — stick with me.

Pour some coffee. We're going somewhere.

Fast Forward to a Different Kind of Canvas

Last month, I watched my kids mess around with Sora, OpenAI's video generation tool. One son turned himself into a dancing ketchup packet set to an Irish reel. The other became a puppet playing stand-up bass in a rockabilly band for an audience of swing-dancing creatures.

They weren't consulting "The Ultimate Sora Prompt Library" or scrolling through "10 Video Prompts Every Creator Needs."

They were just… trying stuff. Describing what they wanted. Getting weird results. Adjusting. Trying again.

Meanwhile, my LinkedIn feed was drowning in posts about prompt libraries:

• "Copy these exact ChatGPT prompts for perfect marketing copy!"
• "The 5-Minute ChatGPT Hack for Perfect SEO Titles!"
• "Never Think Again: The Ultimate Template Library Has Arrived!"

And I felt that same art class frustration bubbling up. We're doing the Rembrandt assignment again. We're teaching people to copy brushstrokes instead of learning to see.

The Seductive Promise of Paint-By-Numbers

Look, I get the appeal of prompt libraries. They promise shortcuts. Proven formulas. Professional-looking results without the messy trial-and-error part. Input the right code, get the expected output. Like a vending machine for creativity. In fact, this is precisely the stage I was in with Rembrandt: learning the proven formulas and the technical 'code' to even begin.

And yes, prompts can teach you syntax. You can learn that ChatGPT responds well to "Act as a [role]" or that Midjourney likes specific aspect ratios and style references. That's useful information-the same way learning that oil paint dries slowly and needs medium to flow properly is useful.

But here's what prompt libraries don't teach you:

• How to think about what you're trying to make
• Why it needs to exist
• Whether the output actually serves your purpose
• How to recognize when something's not working
• What to do when the template doesn't fit your specific situation

You can create something that looks professional. That checks all the boxes. That matches the example in the library.

But you haven't learned to see. You haven't developed your own creative vocabulary. You're locked into someone else's way of thinking, painting someone else's painting, wondering why it doesn't quite feel like yours.

"Prompt libraries treat AI like a vending machine. Input the right code, get the expected output. But that's not how creative tools work. And it's not how creative thinking develops."

What Bob Was Actually Teaching (I Think)

The Rembrandt assignment wasn't really about Rembrandt. It was about understanding the fundamentals. How light works. How paint behaves. How composition creates visual weight. I needed that vocabulary before I could say something with it.

The problem is that most people — myself included, apparently — get stuck in step one. We keep copying because copying feels safe. We can look at a prompt that generated something impressive and think, "If I just use those exact words, I'll get impressive results too."

But the DaVinci moment was the real lesson. My professor wasn't saying "you suck" or "you aren't qualified." Bob was saying "you've learned the techniques, now stop just replicating and start synthesizing. Use what you've learned to say something that's yours."

That's what I tell myself today, anyway.

I like that, Bob.

I'd learned to shade like a dollar store DaVinci without meaning to. I had absorbed the technical vocabulary. But I hadn't yet made the leap from "how did they do it?" to "what do I want to do?" Bob didn't offer a better brushstroke; he forced me into the creative feedback loop: recognizing the gap between my fuzzy intent and the accidental result, and forcing me to refine my thinking.

That gap — between technical competence and creative intent — is where actual growth happens. And it's exactly where prompt libraries leave people stranded.

What My Kids Did Differently (Besides Creating Cursed Content)

My kids didn't start with technique. They started with vision: "I want to be a dancing ketchup packet." Not "What's the optimal prompt structure for anthropomorphized condiments?"

When Sora didn't deliver what they expected — and trust me, it often didn't — they didn't search for a better template. They asked themselves: "Why didn't that work? What did I say versus what did I mean?"

They hit constraints. Couldn't use trademarked characters. Had to describe things without pop-culture shorthand. This forced them to think more specifically. To articulate the essence of what they wanted, not just reference something that already existed.

They were doing something I struggled with in that art class: they were building creative judgment in real-time. The tool became a mirror showing them where their thinking was fuzzy.

And yeah, the final result was a video of them chasing turkeys through the woods to promote their Scout troop's Campsgiving Weekend. It was ridiculous. Their friends loved it. More importantly, they loved making it. Because it was theirs.

That's the opposite of the prompt library approach. Prompt libraries say: "Don't think, just copy this." My kids' approach was: "Think clearly, then describe clearly." One produces generic outputs. The other develops creative muscles you'll use forever.

"Prompt libraries say: 'Don't think, just copy this.' My kids' approach was: 'Think clearly, then describe clearly.' One produces generic outputs. The other develops creative muscles you'll use forever."

The Real Skill Has Nothing to Do With Prompting

Here's what actually matters when working with generative AI:

1. Knowing what you want to make (vision)
2. Understanding why it needs to exist (intent)
3. Describing it clearly (communication)
4. Recognizing when it's not working (judgment)
5. Adjusting based on the gap (iteration)

Prompt engineering is just #3. And you can't do #3 well without #1 and #2.

The prompt is just the translation of your thinking. If your thinking is fuzzy — if you don't actually know what you're trying to accomplish or why — no template will save you. You'll just produce an imitation Rembrandt, exactly like I did years ago.

Bob could teach me brush techniques. He could show me how Rembrandt built up layers of paint. But he couldn't teach me what to paint. Or why it mattered. Or how to know when it was working.

Those things only come from practice. From trying. From failing. From looking at the gap between what you meant and what you made and asking: "What do I need to understand better?"

A Different Approach (That's Actually Harder)

Ultimately prompt libraries are fine. They are a great place to start and gain understanding. Treat them like studying masters' techniques. Learn the syntax. Understand the patterns. See what's possible.

Then close the library.

Start with questions instead:

• What am I actually trying to communicate here?
• Who is this for, and what should they think/feel/do?
• What's the specific thing I need this tool to help me create?
• Why does this need to exist instead of the 10,000 other things like it?

Then prompt from that. Your first attempt will probably be bad. That's not a bug, that's the point. The gap between what you got and what you wanted reveals where your thinking needs work.

Build a practice. Keep a prompt journal like an artist keeps a sketchbook. Document what worked, what didn't. Notice your patterns. Where do you consistently get stuck? That's not a prompting problem — that's a thinking problem. And that's actually more valuable to discover.

The goal isn't "perfect prompts." The goal is clearer thinking, more precise communication, and stronger creative judgment. Better prompts are just a side effect.

"The goal isn't 'perfect prompts.' The goal is clearer thinking, more precise communication, and stronger creative judgment. Better prompts are just a side effect."

Did Any of It Matter?

I've been thinking about that art class a lot lately. Did the Rembrandt assignment matter?

I guess. But not because I can now paint like Rembrandt (I can't). It mattered because I learned to see how paint works. That vocabulary lives somewhere in my hands and eyes now, even if I never touch a canvas again.

Did the DaVinci criticism matter?

Sure. It showed me the difference between technique and voice. Between copying and synthesizing. Between having skills and having something to say.

I'm still learning that lesson. We all are.

The same is true with AI. Prompt libraries teach syntax, and that's fine for starters. But that's the first step on the journey, not the destination. The work that actually matters is knowing what you want to say. And being willing to fumble through saying it badly until you learn to say it clearly.

My kids already understand this instinctively. They didn't need the perfect prompt to create 'Chocolate Cereal Vampire: The Musical'. They needed a clear (if deranged) vision and the patience to iterate until Sora understood what they meant.

I've collected a fair share of prompt templates, but the consistent, daily practice of clear thinking is where I'd like to spend my time.

It's the only way forward. The alternative is just copying, and frankly, I'm not a very good accidental DaVinci.

Right, Bob?

Next time: If everything is built on what came before-if even my "original" art was unconsciously echoing DaVinci, if AI is trained on everything ever created-can anything truly be new? Or maybe that's the wrong question entirely.

Originally published at https://www.linkedin.com.