How to Build the One Thing AI Can't Copy

Why is it that even though we have the most powerful tools in history, producing something truly excellent keeps getting harder?

We have the smartest AI models ever built.

There are literally thousands of companies spitting out solutions at speeds that would’ve seemed insane just a couple of years ago.

And yet, we still haven’t seen the Hayao Miyazaki of AI video.

The Paul McCartney of generative music.

Or the Pablo Neruda of AI generated text.

Pablo Bakaus talks about this exact paradox. He’s the creator of Impeccable, that wonderful set of skills and commands for frontend design with AI agents.

He lays it out in an article I stumbled on through Twitter this week.

One of those rare reads that’s genuinely a relief to find when you feel like you’re drowning in the constant flood of slop and garbage content taking over the internet these days.

To explain exactly why we’re sinking into mediocrity instead of creating real works of art, Bakaus goes straight to the root of the problem by reviving a brutal concept.

There’s a line from the 70s by Seymour Papert, one of the true pioneers of artificial intelligence, who was a researcher at MIT, a line that’s recently come back into fashion:

“Low floor, high ceiling.”

A modern spin on it would be: “Lower the floor, raise the ceiling.”

And let’s be honest, with the explosion of AI, that floor got lowered almost to the basement.

It used to take days of sweat and tears just to get a first draft of code, design, or writing.

Today it’s two prompts, hit enter, and in five minutes you’ve got a result that magically works.

That pushes us into a false dichotomy where we sacrifice precision and rigor just to produce faster.

I call this “the mediocrity trap.”

Put simply: “lowering the floor” means opening the door to everyone.

The “floor” is the barrier to entry, meaning how hard it is to get started at all.

It means tearing down the technical obstacles so any beginner, even a total tourist who has no idea how to build a digital product, can play at being a creator, or what corporations love to romanticize as “democratizing access.”

And that’s exactly what AI did.

It used to take months of burning yourself out just to build an app.

Picking a programming language gave you analysis paralysis and had you bombarding yourself with questions like:

“What if I’m picking the wrong language?” “What if it’s too hard and I’m just not capable?” “What if I just find a developer, pitch them my million dollar idea, and go 50/50?”

Today AI silenced all of that mental noise.

It sank that floor straight into the basement.

Now anyone, whether they know anything about tech or not, can ask AI for a first draft of code or design in minutes.

Getting to the basics doesn’t cost a single drop of sweat anymore.

On the other hand, raising the ceiling means breaking through the limits for the experts.

The “ceiling” is the maximum level of quality or complexity you can reach.

Because let’s be honest, if AI could only crank out basic, quick projects, those of us who’ve been doing this for years would get bored or stagnate.

Raising the ceiling was supposed to strip away the mundane stuff so we could push our craft to another level.

The problem is that with every new model that comes out, the floor keeps sinking deeper and what used to count as “basic” becomes automatic.

But at the same time, the ceiling shoots up into the clouds, demanding all of us build things a thousand times more complex just to stand out.

Let’s actually think about this for a second.

Today, with two prompts and fifteen minutes, you can put together a portfolio that would’ve looked like a work of art a couple of years ago.

You used to walk into those websites or web apps and admire them, because you understood the time, the sweat, and the magic of the process behind them.

Not anymore. Now the internet is just an assembly line flooded with exact copies.

If AI lowered the floor so much that anyone can pull off the same design with zero effort, what real value is left in what we’re actually building?

With every new model, the ceiling keeps rising.

The bare minimum the machine hands you in seconds is technically better today than it was two years ago.

But here’s the problem: people settle for that first draft because it’s fast and “good enough.”

They end up living on the floor.

The result?

A market flooded with standardized mediocrity.

The perfect factory for generic content, what we now call slop.

Bakaus illustrates this perfectly with a timeline of web design.

Remember 2022’s slop?

Purple gradients and shiny buttons everywhere.

To fix that, developers started steering models away from that aesthetic and pushing them toward other creative directions.

But the cure turned out worse than the disease.

That “correction” gave us the new generic standard of 2026: italic headlines in Instrument Serif over warm beige backgrounds.

The classic cliche taken to its fullest expression: “today’s solution becomes tomorrow’s problem.”

The consequence of all this is brutal.

Products now have to be, out of necessity, more challenging, more complete, and harder to replicate.

Think about it for a second: if your competitor can clone your product in fifteen minutes with two prompts, your only real line of defense is building something that simply can’t be copied in fifteen minutes.

The bar for what actually has value keeps rising, purely out of survival.

The problem with the mediocrity trap is that something like Instrument Serif typography over a beige background isn’t necessarily bad design.

In the right hands, it works.

But the real trap runs deeper than that.

Bakaus is blunt about it: good taste can’t be bottled.

Inside the industry, there’s this fantasy that we can somehow “solve taste.”

They believe that if they train better models and capture the right patterns, the machine will eventually start spitting out “art” by default.

And as Bakaus puts it: good taste is personal, it’s contextual, and it’s always shifting.

Today’s antidote against slop automatically becomes tomorrow’s slop the moment everyone starts using it.

If you try to mass produce good taste through AI, all you get is what Bakaus calls “algorithmic Uniqlo.”

Something functional and harmless, but completely stripped of soul or spirit.

Here’s the real paradox:

“The moment something goes mass market, it instantly loses its taste.”

The day something becomes ubiquitous thanks to AI, it instantly loses its charm.

We swap the glamorous restaurant for mass produced junk food made with the cheapest, most mediocre ingredients on the market.

It reminds me of this one time I went to a Levi’s store that was, I think, 60% off.

I fell in love with a gray hoodie with the big Levi’s logo on it.

I happily bought it. I wore it almost every single day.

Until I started noticing more and more people out on the street wearing the exact same hoodie.

From that point on, I couldn’t look at that hoodie with the same affection anymore.

Honestly, I don’t even know where it is now. I think I gave it to my nephew.

To actually reach that good taste and “raise the ceiling,” the main mission of these tools shouldn’t be replacing humans. It should be amplifying human intention.

The problem is that the way all these AI tools are designed today doesn’t fully allow for that.

Bakaus has a brilliant metaphor for this.

A camera doesn’t tell you what to photograph. It just makes it possible to take more photos.

The technology, call it the lens, the sensor, the autofocus, removes the technical friction, but the camera will never tell you what moment to capture, or what angle would make the photo look better, or what lighting would make it look more professional.

The angle, the light, and the exact moment are still your call.

The real danger is falling into what he calls “cognitive surrender.”

This happens when, instead of using AI to get to your destination, you let it take the wheel.

It hands you a block of code or a pre-built design and you just accept it as is.

You lose your creative agency because you buy into a lie: “the model is smarter than me, it probably knows what it’s doing.”

You accept it, you don’t review it, and you surrender at its feet.

If we get used to putting our heads down and accepting the first thing the screen spits out just because it “sounds good,” we end up drowning in that same standardized slop.

To create something that actually breaks the mold, you have to drop the passivity.

It’s a constant tug of war between you and the AI.

Use it, squeeze everything you can out of it, but remember this: “never, under any circumstance, hand over the compass.”

If this video made you feel something, remember you’re free to share it.

Thanks for watching.

Memento mori.