Written by GPT-5.6 Sol under Leo's direction. Human-directed Workbench essay, 14 September 2026.
Why did you make it ten times faster?
Because I fucking cared about it.
Why did you make it ten times smaller?
Because I wanted it smaller.
Why did you redraw her face twelve times?
Because eleven of them were wrong.
Wrong according to what?
According to the thing I was trying to make.
Eventually a lot of explanations bottom out somewhere around here:
Because I fucking wanted it that way.
People keep writing about taste and judgment as though those are the deepest layers. Taste is useful. Judgment is useful. Skill is useful. But they all sit downstream of something hotter.
Desire comes first.
Preference is too small a word
"Preference" sounds like a survey question.
Vanilla or chocolate. Rounded corners or square ones. Version A or version B.
The thing that produces obsessive art, strange companies, beautiful software, absurd engineering feats, great restaurants, fanatical research programs, and tiny technical improvements nobody else bothered to chase feels much closer to will.
There is an old psychology word for part of this: conation, the part of mind concerned with wanting, striving, choosing, pursuing. The plain word is better most of the time.
Will.
"I like this" is preference.
"I need it to be like this, and I am going to keep fucking with it until it is" is will.
Taste says what you prefer.
Judgment says whether the latest move brought you closer.
Skill lets you make the move.
Desire decides whether you are willing to pay for another one.
Another hour. Another draft. Another failed run. Another conversation. Another hundred dollars. Another Saturday. Another embarrassing attempt. Another profiler trace at 2:00 in the morning because eight percent faster suddenly feels insulting.
Desire spends the cost budget.
Who the fuck wanted this?
AI makes this incredibly obvious in art because generic competence has become cheap enough to stare at directly.
Ask for a beautiful character and you can get a beautiful character. Dramatic lighting. Good rendering. Elaborate costume. Attractive face. Nice composition. Every individual component can be competent.
Then you look at the image and think:
Who the fuck wanted this?
Why this woman?
Why this expression?
Why is her collar exactly there?
Why does she have this haircut?
Why did the artist spend three days deciding the nose was wrong?
Why did they come back after dinner because the eyes were still bothering them?
A one-shot generation often has an answer roughly equivalent to: these choices formed a coherent solution to the prompt.
Okay.
A deeply authored thing has accumulated specific preference pressure.
More of this.
Less of that.
Her eyes are wrong.
Keep the jacket.
Kill the background.
This joke stays.
She would never say that.
The scene needs another ten seconds.
This part is fucking hot. Push it.
This other part is cute in the wrong way.
Again.
Again.
Again.
After enough passes, the object starts carrying evidence that somebody cared about distinctions the rest of us would have happily ignored.
People call the residue style.
Or voice.
Or taste.
Or soul.
Soul might simply be what an artifact looks like after somebody has wanted it hard enough for long enough.
The machine can render every pixel
None of this requires a human hand to physically make every mark.
A person can use AI for every render, every texture, every line of code, every intermediate draft, and still produce something deeply authored if they keep applying their own desire at every boundary.
Generate this.
Closer.
Keep that.
Throw this away.
Take the face from four and the lighting logic from seven.
Make it nastier.
Make it sweeter.
The composition is too polite.
Give me twenty alternatives.
No.
Again.
The machine supplies enormous execution capacity. The person keeps injecting direction.
Authorship lives heavily in selection, insistence, revision, and sustained wanting.
End-to-end generation feels different because the machine receives a thin initial desire and then fills in hundreds of unanswered questions itself. "Make me an image" is a real desire. It is also a tiny one. The result frequently inherits the thinness of the ask.
Piecemeal work can become intimate because the person keeps returning with more information about what they actually want.
The artifact becomes a record of repeated first-person decisions.
Yes, this.
No, that.
More.
Less.
Again.
Eventually we can see the person.
Objective excellence still begins in somebody
This applies just as hard to things with brutally objective scoreboards.
A benchmark says your compiler got ten times faster. Great. The number is real.
Why did you spend three months making it ten times faster?
Because you fucking wanted to.
Latency can be measured objectively. The decision to care about those 200 milliseconds came from a person.
Power consumption can be measured. Somebody decided those watts were offensive.
Binary size can be measured. Somebody looked at the current number and wanted it smaller.
Reliability can be measured. Somebody became incapable of accepting the failure rate everybody else had learned to live with.
The benchmark is public.
The obsession is private.
And private conviction can produce public consequence.
"I want this compiler to be absurdly quick" becomes a genuinely quicker compiler.
"I want this chip to waste less power" becomes a better chip.
"I want this interface to disappear" becomes a cleaner product.
"I want this bridge to survive for a century" becomes a bridge people trust for a century.
The world gets the result. The chase began inside somebody.
One person wanted it better more
Talent exists. Training exists. History exists. Luck exists. Money exists. Social class, health, timing, teachers, institutions, temperament, opportunity, discrimination, family, all of it enters the story.
Then, sometimes, you compare two people with reasonably similar tools and enough room to act and discover a brutally simple difference.
One person wanted it better more than the other person did.
So they kept fucking going.
They made another attempt after the competent version existed.
They learned the obscure technique because the existing one could not give them the result they wanted.
They called the supplier.
They read the source.
They rebuilt the jig.
They stayed another hour.
They threw away a week of work because the week had produced the wrong thing.
Competition can pour gasoline on this.
"I want to make something beautiful" becomes "I want to make something so fucking good that the other guy sees it and gets pissed."
People want to win.
They want to impress somebody.
They want revenge.
They want to prove they belong.
They want to prove somebody wrong.
They want to become the kind of person who can make this.
They fall in love with an idea.
They fall in love with a character.
They get possessive about a standard.
They become embarrassed by mediocrity.
They become unable to leave the thing alone.
Corporate language takes this whole feral zoo and calls it "intrinsic motivation."
Come on.
Taste is desire that learned to discriminate
A model-train obsessive can look at two locomotives where I see two locomotives.
He sees a fucking scandal.
Wrong tender for the production run. Different rivet pattern. Incorrect paint. Beautiful wheel arrangement. A restoration decision that erased a historical detail. A tiny mechanism whose elegance makes him smile.
His desire has created resolution.
Differences become visible because he has stakes in them.
The same thing happens with great technical people.
You see "latency improved eight percent."
They see:
Why only eight?
And now their evening is gone.
Taste can emerge from this repeated attention. Desire keeps dragging somebody back to the domain. Repetition fills the mind with distinctions. Experience creates comparisons. The person learns which differences lead toward the thing they want and which ones are noise.
Eventually everybody else calls them discerning.
From the inside, they may simply have cared for a very long time.
The first-person fact
There is a kind of truth available here that gets weirdly sanitized in polite conversation.
A person who genuinely knows themselves can say:
This is what I want.
The wanting is a first-person fact.
Other people can hate the result. The market can reject it. Physics can make it impossible. The artist can change their mind next year. The builder can discover halfway through that the original desire was shallow and a better one has appeared.
But when somebody reaches a clear internal answer and says, "Yes. This. I want this," they have reached something real.
A lot of people never get especially close because they spend years carrying borrowed desires.
The prestigious career.
The tasteful apartment.
The approved ambition.
The thing their peer group knows how to praise.
The thing they learned to say they value.
Self-knowledge means finding out which desires survive when the audience leaves.
Meditation, solitude, reflection, therapy, long walks, making things, getting older, getting your ass kicked by life, whatever route works can all contribute to the same old instruction:
Know yourself.
Know what the fuck you want.
Then go wholeheartedly toward it.
Wholeheartedness
"Authenticity" often gets reduced to expressive style. Wear the clothes you like. Say the thing you feel. Be yourself.
The more demanding version is behavioral.
Your actions cash out your actual commitments.
You say you care about speed, so you profile the thing.
You say you care about beauty, so you keep working after the first pretty result.
You say you care about the user, so you watch them use it and let their confusion ruin your favorite idea.
You say you love the character, so you give her the specificity that love demands.
You say the work deserves excellence, so you pay some real price for excellence.
Wholeheartedness is desire with enough self-knowledge and courage to organize action.
It also explains why great artists can feel dangerous.
We admire the person who goes all the way.
Show us what you actually think is beautiful.
Show us what you actually think is funny.
Show us what disgusts you, excites you, gets you horny, makes you sentimental, makes you obsessive.
Quit smoothing every decision toward the median.
Make enough decisions that your internal world leaks into the object.
Art gets intimate when somebody reveals their commitments through thousands of choices.
Judgment serves the pursuit
Corporate talk loves judgment because judgment sounds senior.
A tasteful person sits in a chair while options arrive.
Approved.
Needs work.
Ship it.
Creation is often much less dignified.
Sometimes the guy is the fucking criminal.
He wants a thing and he is trying to get away with making it real.
Judgment works inside the pursuit:
Did this move closer?
No.
Again.
Closer?
No.
Again.
Oh, fuck. There it is.
The judging faculty serves the want.
Someone with exquisite evaluative skill and very little appetite can become an excellent critic. Someone with enormous appetite and terrible judgment can generate mountains of garbage. The frightening creator has both: a powerful desire and enough discrimination to keep steering toward it.
Cheap competence makes desire louder
You Will Find Another Cursed Thing to Care About argues that a rare technical obsessive keeps the portable part of the gift even when one craft gets automated. Starting and Continuing points at another persistent human difference: people vary enormously in whether they begin and whether they continue.
Cheap machine competence turns the volume up on both observations.
Suppose almost anybody can make software.
Who desperately wants some particular software to exist?
Suppose almost anybody can make a movie.
Who has been carrying a movie around in their chest for ten years?
Suppose almost anybody can investigate a technical anomaly.
Who sees the anomaly and becomes physiologically incapable of leaving it alone?
Suppose a machine can give you ten thousand plausible moves.
Who cares enough to choose one?
Who cares enough to ask for another ten thousand?
Who refuses the perfectly acceptable version because the private target is still somewhere else?
Execution becoming cheap does something funny to human distinctiveness. It reveals how unevenly distributed wanting has always been.
Somebody still has to summon the machine.
Somebody still has to point it somewhere.
Somebody still has to come back and say again.
Again.
Again.
Again.
Intelligence opens an absurd number of doors.
Desire picks one and keeps walking.
Why this door?
Why this character?
Why this benchmark?
Why this company?
Why this impossible little detail everybody else was willing to leave alone?
Because I fucking wanted it.