In the AI Era, Those Who Mine Rough Gems Well Are Strong

In the AI Era, Those Who Mine Rough Gems Well Are Strong

This essay did not begin very elegantly.

While talking with GPT, I said:

“So your job is basically cleaning up my shit?”

It was a joke.

I kept throwing out unpolished thoughts, and GPT was turning them into logic and sentences.

But after laughing and thinking again, I found a rather important structure inside that joke.

I said:

“Wait. Isn’t this an essay idea too?”

And an essay actually emerged.

Then I noticed something even more interesting.

The process that created the essay was demonstrating the essay’s own argument.

The initial input was one messy sentence.

But there was a rough gem inside it.

If humans generate unrefined thoughts and AI structures them, might a new form of cognitive division of labor emerge between humans and AI?

I had not thought of it that way from the beginning.

At first, it was just a joke.

Good Thoughts Are Not Born as Good Sentences

We mistakenly assume a good thought must be logical and clear from the outset.

But humans do not think like academic papers.

Most thoughts begin like this.

“Something’s strange.”

“Wait, don’t these two things have the same structure?”

“Couldn’t we just think about it in reverse?”

“There’s something here, but I can’t explain it.”

These are not finished sentences.

They are observations, intuitions, connections, and hypotheses.

I think of them as rough gems of thought.

A rough stone does not initially look like a jewel.

It is covered in dirt and mixed with impurities.

Sometimes you cannot even tell whether it has value.

Thoughts are similar.

When they first emerge, they are loose, extreme, and underexplained.

But an important structure may be hidden inside.

Good thoughts need not be born as good sentences.

Refinement Used to Be Expensive

The problem came after discovering the rough stone.

To communicate it, you had to process it.

Turn the feeling in your head into sentences,

the sentences into logic,

the logic into structure,

check the evidence,

examine objections,

and produce something worth reading.

In other words,

discovery → structuring → draft → verification → revision → editing → finished work

was a long production process.

Many thoughts therefore disappeared before ever reaching the world.

Not because there were no ideas,

but because refinement was costly.

“There seems to be something here.”

You get that far, then abandon it because explaining it feels like too much trouble.

More precisely,

you abandon it because the cost of explanation and verification feels greater than the value of exploring it.

AI is changing this cost structure.

AI Has Lowered More Than the Cost of Writing

People say generative AI has reduced writing costs.

True.

It writes emails,

organizes reports,

revises sentences,

and summarizes long documents.

But a more important change is underway.

AI lowers the cost of refining thought.

Used merely as a writing tool, AI comes after a human finishes thinking.

At the end, the human says:

“Write this well.”

But you can use it the other way around.

Call AI when your thoughts are at their messiest.

“I have a feeling about this, but can’t explain it.”

“These two things seem connected, but I don’t know why.”

“Find the holes in this hypothesis.”

“Attack it from the other side.”

“What conclusion follows if we push this all the way?”

Then AI is no longer merely a tool that writes sentences for you.

It becomes a refinery for thought.

A New Human-AI Division of Labor

Seen this way, the roles become interesting.

Humans observe the world.

Sense that something is strange.

Connect different things.

Form hypotheses.

Ask unexpected questions.

AI takes those rough gems and finds structure.

Removes duplication,

connects the logic,

constructs counterexamples,

and clarifies expression.

Then it returns the result to the human.

The human judges it.

“That’s not right.”

Discard it.

“Wait, this part seems right.”

Dig again.

AI refines again.

Ultimately, the structure is simple.

Humans explore outward.

AI refines.

Humans select.

The enormous refinement cost between discovery and judgment falls.

As AI Becomes Common, Rough Gems Matter More

A paradox appears here.

As AI advances, the average polish of output will keep improving.

Anyone can write a clean email.

Anyone can produce a plausible report.

Anyone can generate grammatically excellent prose.

Polished sentences themselves become less scarce.

Refinement ability converges upward.

What becomes relatively scarce is something before refinement.

What did you observe?

What struck you as strange?

What did you connect?

Which assumption did you question?

What question did you create?

In other words,

what rough gems do you possess?

This explains why two people using the same AI get different results.

One enters an ordinary question.

The other connects two phenomena nobody has connected before.

The refinery is the same; the raw material differs.

So does the output.

More Refineries Raise the Value of Mines

This is an interesting economic structure too.

When refineries are scarce, excellent refining technology is a competitive advantage.

But once everyone can access a world-class refinery, the story changes.

The bottleneck moves.

From the refinery to the mine.

What can you discover?

How much raw material can you extract?

What do you find in something others dismissed as a stone?

The relative value of that ability increases.

Human observation, curiosity, experience, taste, and a sense of what needs questioning may therefore become more important in the AI era.

AI does not erase human differences;

it shifts where those differences appear.

From prose to questions.

From organizing to discovering.

From output to input.

The Rule That Thoughts Must Be Finished Before Speaking Changes Too

For a long time, we have been taught:

Organize your thoughts before speaking.

Prepare evidence before presenting.

Edit your writing before showing it.

These remain important when communicating with people.

But we need not apply them to thinking with AI as well.

Talking to AI before organizing your thoughts can be more useful.

If you provide only finished thoughts, AI handles only the final processing.

But if you begin with small rough gems, you can use AI throughout exploration.

“That’s strange.”

“Why is that?”

“Could this be the reason?”

“Then wouldn’t it connect to that too?”

You go back and forth with AI from the moment a thought appears.

AI becomes an amplifier of thought, not a secretary.

This Essay Was Made That Way

This is not just theory.

It is exactly how this essay came about.

I did not begin by thinking,

“I should write about the structure of cognitive division of labor between humans and artificial intelligence as AI advances.”

That was not how it started.

I was simply joking with GPT.

I threw out whatever came to mind.

GPT organized it.

Seeing the result sparked another thought.

Then I said one very unpolished sentence.

“So your job is basically cleaning up my shit?”

I found a rough gem there.

The unrefined thinking humans generate.

The structuring and processing AI handles.

And the judgment that ultimately returns to the human.

There was a structure inside the joke.

So we extracted it.

Polished it.

Expanded it.

Tested it.

And it became the essay you are reading.

In the AI era, you do not need impressive thoughts from the outset.

What may matter more is not letting strange thoughts slip away.

Hold on even when they are unfinished.

Record them even when you cannot explain them.

Make a hypothesis even if it seems likely to be wrong.

Then put it into the refinery.

If it has no value, discard it.

But if one turns out to be a rough gem, the story changes.

However good a refinery AI may be,

it cannot produce anything without input.

The important question in the AI era

may not be “How well does AI write?”

It may be closer to this.

Can I keep discovering rough gems worth giving to AI?

This essay did not begin as an essay either.

It began as a joke.

More precisely,

a joke that was not particularly clean.

But it contained a rough gem.

AI polished it.

I selected again.

And it became an essay.

Good thoughts need not be born as good sentences.

Sometimes they simply appear as stones.

What matters

is not walking past them.

Haha, posted this because it was funny.