The thing itself

The food is getting cold.

We left it on the table a long time ago.

Someone had said wait. The phone came out. The plate moved toward the light. A photograph was taken. Then another.

Eventually the phone went away.

And we ate.

That part matters because nothing in this essay requires us to hate the photograph.

The photograph can be beautiful.

It can preserve an evening we want to remember. Years later, someone may find it and remember who was sitting opposite, which city they were in, what happened afterward. The image can become part of the experience rather than its enemy.

The problem begins only when we forget that the photograph and the meal are different things.

One can travel without the other.

The photograph can reach ten thousand people who never tasted anything.

It can be improved, cropped, brightened, rearranged. It can make an ordinary dinner look extraordinary. It can make a lonely evening look social. It can make a carefully staged moment appear spontaneous.

And none of that changes the food.

The meal happened somewhere else.

Writing may now be entering the same separation.

An article can be excellent.

That point is worth repeating because otherwise the entire argument becomes too easy.

The AI-generated essay may be better than the human one.

It may contain a stronger structure, clearer language, better counterarguments and fewer factual mistakes. It may teach me something. I may enjoy reading it. I may even return to a sentence years later.

The artifact can have value independently of how it was produced.

That is not the question.

The question is what else we once believed the artifact told us.

When I read an essay under someone’s name, I am not merely receiving information.

I am encountering a claim of presence.

I have been here.

Not geographically.

Intellectually.

I have spent time with this question.

I know why these examples are here.

I understand what I am asking you to believe.

Something happened between this subject and me before these words reached you.

For a long time, writing itself offered weak but useful evidence that this encounter had taken place.

Now it does not necessarily do so.

That may be one of the quieter transformations artificial intelligence brings.

We have spent enormous energy asking whether machines can write like humans.

Perhaps we should have been asking what happens once they can.

Because once fluent writing is abundant, fluency stops telling us very much.

Once a sophisticated argument can be produced on demand, sophistication itself becomes a weaker signal.

Once uncertainty can be requested, hesitation becomes less reliable evidence of reflection.

Once a machine can manufacture the narrative of discovery, even the sentence I changed my mind no longer proves that anybody did.

The evidence remains.

Its relationship to the event has changed.

We have already lived through this elsewhere.

A photograph once provided unusually strong evidence that a camera had faced something.

Then images became editable.

Later they became generatable.

We did not stop looking at images.

We changed what we were willing to infer from them.

A recording once meant that a voice had spoken those words.

That inference is weakening too.

A video once meant that the person had stood in front of the camera.

That is no longer safe.

And perhaps writing is simply arriving at its turn.

The existence of the words will remain undeniable.

What becomes uncertain is everything we used to imagine behind them.

This does not mean the future belongs to suspicion.

I hope it does not.

A world in which every paragraph is treated as forensic evidence would be a miserable place to read and an even worse place to write.

Nor do I want writers submitting proof that they suffered sufficiently for every sentence.

Difficulty is not virtue.

AI should remove useless difficulty.

It should help people cross languages.

It should help knowledge escape bad prose.

It should allow someone who has never learned the conventions of an essay to communicate something worth knowing.

It should challenge weak arguments.

It should find contradictions.

It should make writers better.

I intend to keep using it.

But perhaps we need to become more precise about what we are using it to do.

There is a difference between asking a machine to help me say something and asking it to give me something to appear to have said.

There is a difference between extending a mind and renting one.

There is a difference between an article produced with artificial intelligence and an intellectual identity produced by artificial intelligence.

The technology cannot draw those lines for us.

A detector cannot recover intention from syntax.

A platform cannot look inside the moment before the prompt.

A label can tell us that AI was used without telling us whether it translated a lifetime of knowledge or invented Thursday’s opinion.

So perhaps the old shortcut has ended.

We can no longer look at the finished page and infer everything we want to know about the person behind it.

We will have to learn other signals.

History.

Continuity.

Responsibility.

Conversation.

The willingness to answer.

The willingness to correct.

The ability to say I don’t know when the article has ended.

The strange recurrence of certain questions across years.

The places where a writer changes.

The places where they do not.

Not proof.

Relationship.

And this may explain why the future of paid writing is not necessarily as bleak as infinite prose first makes it appear.

I would not pay someone merely because they can produce two thousand elegant words.

That ability is becoming inexpensive.

I might pay because I have read that person for years and want to know what happens when this particular question reaches that particular mind.

The distinction is subtle, but economically it may become enormous.

We may stop paying primarily for articles.

We may pay for continuity of mind.

That would give platforms such as Substack a choice.

They can become better machines for producing and circulating content.

Or they can become better places for readers to build durable relationships with people whose thinking they want to follow.

The first path leads toward a familiar destination.

More production.

More visibility.

More optimization.

More material competing for less attention.

The writer becomes a profile.

The article becomes feedstock.

The feed survives even when the individual work disappears.

We have seen versions of that world already.

The second path is harder because relationships do not scale as cleanly as content.

They require time.

A reader must return.

A writer must remain recognizable without becoming repetitive.

Trust has to survive disagreement.

The archive has to matter.

Silence has to remain possible.

A platform would have to accept that sometimes the healthiest thing a writer can publish this week is nothing.

That is an unusual proposition for a platform.

It may also become a valuable one.

Because when production becomes infinite, restraint acquires information.

The person who could generate something and does not may be telling us something.

I do not have enough yet.

That sentence may become strangely expensive.

I think again about the meal.

The photograph has been posted by now.

Perhaps people are liking it.

Someone asks where the restaurant is.

Someone else saves the location.

The image continues travelling.

At the table, none of that is happening.

There is only the food.

It is no longer arranged perfectly because we have begun eating it. The sauce has moved. Someone has taken the better piece. The conversation has gone somewhere the photograph could not have predicted.

Nobody outside the table can taste any of it.

That does not make the photograph false.

It simply means the photograph was never the meal.

Perhaps we will have to relearn the same distinction about writing.

The article is real.

The sentences are real.

The argument may even be good.

But the article and the thinking are not the same thing.

For centuries, they were expensive enough to remain close together that we rarely needed to notice the distance.

Artificial intelligence makes the distance visible.

And perhaps that is what platforms, schools, newspapers, publishers—and readers—will eventually have to learn to ask.

Not:

Was this written by a machine?

But something much harder:

Was there a mind here worth returning to?