Why Masterpieces Are Still Rare in the Age of AI

Have you ever wondered whether an ordinary person—someone who is not a professional writer—could use AI to create something on the scale of Harry Potter, The Lord of the Rings, or Game of Thrones?

A world expansive enough to be adapted into films, extended into franchises, and turned into a cultural phenomenon worth billions?

Most people arrive at the same answer almost instantly: probably not.

But why?

Is it because the models are not powerful enough yet? Or because we haven't discovered the right prompts?

If we follow the “correct” creative methodology—build a world first, design narrative arcs, apply classic archetypes, follow the hero’s journey, expand chapter by chapter—AI can absolutely produce something coherent, even commercially viable. It may not become a literary landmark, but it won’t be random text either.

So where does the gap come from?

The Invisible Gate

The difference isn't about tools. It's about whether you've crossed an invisible gate.

That gate is the construction of a mental model.

• Before you cross it, AI looks like genius.
• After you cross it, AI looks like an amplifier.

This is why masterpieces remain rare—in art, in software, in product design—no matter how good the tools become.

From Fiction Back to Software: Why "SaaS Is Dead" Sounds Convincing

In the software world, we've been hearing the same refrain: SaaS is dead.

When user needs can seemingly be fulfilled with a handful of prompts, decades of accumulated software expertise start to look fragile. Features are cheap to generate. Deployment is trivial. Everyone can “build something” and demo it.

With tools like Anthropic’s Cowork and OpenAI’s Codex, the anxiety is understandable: if AI can produce working software directly, what’s left of traditional products?

But the core question hasn't changed:

Can real value be vibed into existence?

Why Vibe Coding Works—Until It Doesn't

Vibe coding is like asking AI to improvise a short story.

When context is limited, structure is simple, and long-term consistency doesn’t matter, you can try a few times and get something surprisingly good. That’s why it feels magical.

But the moment you attempt to build something larger—an operating system, a programming language, or a product expected to survive for years and be accountable to users—vibes collapse.

Without software engineering methodology, it simply doesn't scale.

This isn't a criticism of AI. It's a statement about complexity.

The 1% That Got More Valuable

Kent Beck once made a brutally accurate observation:

"99% of the skills and experience we accumulated may rapidly depreciate. But the remaining 1% could become a thousand times more valuable."

That 1% is not syntax. Not tools. Not frameworks.

It is:

• The ability to make leap-of-faith judgments under incomplete information
• The instinct to place bets when risks are unclear
• The capacity to choose when methodologies and values conflict

The other 99% isn't discarded—it's automated down to irrelevance.

Everything that can be articulated, standardized, and averaged should be handed to AI: system prompts, skills, plugins, agents. Let those capabilities scale.

Humans should remain where information is incomplete, outcomes are uncertain, and decisions cannot be fully justified in advance.

That's the actual division of labor.

AI Is Aligned to the Average—Not the Exceptional

AI is not trained to match the best humans. It is trained to align with the average.

That's the uncomfortable truth.

The good news is that most knowledge workers are far above average in their own domains. The bad news is that when generation cost approaches zero, content doesn’t become valuable—it becomes noise.

High-quality human-generated data is being exhausted. Models increasingly train on AI-generated material, reinforcing alignment with the mean rather than escaping it.

This is why AI outputs can feel astonishing—especially in unfamiliar domains. Often, it’s not because the AI is extraordinary, but because you’re still below the average line.

"Creative democratization" is a misleading phrase. AI hasn't made creativity rare. It has made mediocrity effortless.

The Gate, Proven Experimentally

This invisible gate is not philosophical—it's measurable.

Anthropic ran an experiment with 52 junior engineers learning a new Python library, Trio, under different conditions.

The results were unsettling:

• Engineers using AI assistance scored 17% lower in comprehension and debugging ability
• Their productivity increased by only 2%

When answers arrive without struggle, the brain leaves no structural trace. No struggle, no mental model.

Without a mental model, you cannot debug. Without debugging ability, you cannot claim understanding.

That gate was never crossed.

When "Too Smooth" Is the Real Warning Sign

This matches my own experience precisely.

The moment I became uneasy wasn’t when AI made mistakes. It was when everything felt too smooth. I stopped asking, “Is this correct?” and only checked whether it looked reasonable.

There was no friction. No frustration. No collision with reality.

And worst of all—I felt relieved. Thinking less is comfortable.

Later, when I had to extend the system or explain a simple design decision to a colleague, I realized I couldn’t. Not because it was complex, but because the answer had never been constructed in my mind.

I had never crossed the gate.

AI Amplifies Only What Has Already Passed Through

AI is a rocket engine.

It multiplies output at an unprecedented scale—but it does not steer. If the direction is wrong, you crash faster. If it’s right, it amplifies the rare judgments that already existed.

Masterpieces remain rare not because tools are weak, but because what deserves amplification has always been scarce.

AI doesn’t change that.

It just makes it impossible to ignore.