Feels Like Writing Is Dead

This essay follows the feeling of anxiety about the writer’s place in the world to come. Throughout the piece i explore what differentiates us from the robots and why that differentiation seems necessary. more importantly, whether it is.

---

“The sun went down, and the world felt small.”

Claude wrote that sentence.

I don’t remember what I was looking for at the time. I do, actually. But saying I don’t remember sounds cooler. More mysterious. I was asking it about purple prose. I’d been writing something, and somewhere in the back-and-forth I asked what purple prose meant. I think I wanted it to demonstrate the difference between prose that is merely ornate and prose that actually does something.

Anyway, Claude gave me that sentence, and it blew me away.

I remember the feeling quite clearly. I searched to see if it was from a book I had missed, a movie, or something I had read years ago and forgotten. It wasn’t. As far as I could tell, it was an original from the machine that most writers would argue can do anything but write.

I probably couldn’t have written that sentence myself. I could have spent an afternoon trying to come up with something similar. But I didn’t. The machine did it in seconds, and it was beautiful.

It got me thinking.

Feels like writing is dead.

At least from the economically valuable standpoint.

Which means we have a problem.

The value of writing

By value, I mean the capacity of something to produce a gain or reduce the possibility of a loss. In an economic context, this usually means that someone is willing to pay for it, or that it helps someone achieve an outcome they care about.

Writing is no exception.

People write to make money. They write to gain recognition, influence, opportunities, status, or an audience. They write to persuade people to buy things, support causes, adopt ideas, or change their behaviour. Even when there is no immediate monetary reward, there can be some anticipated benefit.

I’m talking primarily about writing as a product: essays, fiction, articles, marketing copy, technical documents, and the countless other things we produce by arranging words. I don’t mean keeping a journal, writing to organise your thoughts, or composing a love letter to someone you care about. More specifically, I’m talking about the economic and reputational value attached to the ability to write well.

The ability to construct beautiful sentences. The ability to produce compelling prose, to arrange words in a way that captures attention, evokes emotion, communicates an idea, or makes a reader want to keep going.

The skill itself. The mechanics.

And that skill is losing its value. Will lose its value. Has lost its value.

Writing is difficult. Good writing is more difficult. It requires practice, familiarity with language, an understanding of how people think and respond to words, and the ability to translate an idea into something another person can understand and care about.

Not everyone can do it equally well. Those who can do it particularly well can distinguish themselves from everyone else.

That scarcity creates value.

Now AI is here.

And regardless of whether you believe we have achieved artificial general intelligence, whether these systems actually understand anything, or whether they are sophisticated statistical machines simulating human cognition, one thing is becoming increasingly difficult to deny: machines can already produce writing that is better than much of what humans produce, and the technology is likely to improve.

The very basis of the technology gives it an enormous advantage in this domain. Large language models are trained on vast quantities of human language and learn to generate new language from those patterns. They can imitate styles, reorganise information, adapt to audiences, revise their output, and generate different versions of the same idea almost instantly.

What once required a person hours or days can now take seconds.

Anyone with a computer and access to a capable language model can produce competent prose. With a little effort, they can produce excellent prose. The ability to construct beautiful sentences, organise an argument, adopt a particular voice, or make an idea accessible to a particular audience is no longer scarce in the way it once was.

The technology isn’t perfect. It doesn’t need to be. It only needs to be good enough, cheap enough, and available enough to destroy the scarcity that made the skill economically valuable in the first place.

And the technology is improving.

The value of a skill is not independent of how many people, or machines, can perform it. If something that once required years of practice can now be accomplished in seconds, the market has to adjust. The person who spent ten years mastering the craft doesn’t get to charge a premium simply because it took ten years.

The skill to write is not literally worthless. There are still costs involved in producing excellent work, and some forms of writing remain difficult (at least for now). But the ability to produce competent, polished prose is becoming less scarce. And competence alone is becoming a weaker basis for differentiation.

A novel that takes twelve years to write isn’t necessarily more valuable than one produced in a day. If the latter is more compelling, more original, or more useful to its intended reader, the additional labour invested in the former is irrelevant to that comparison.

This isn’t to say that all writing has become worthless. It’s to say that the technical ability to produce it is losing its value as a differentiator. The skill is becoming a commodity.

And the problem is compounded by the other side of the equation: demand.

Why bother reading?

Even if we could still produce good writing, what reason do we have to consume it?

Any page can be summarised in seconds. Any dense document can be explained in whatever format you prefer. Want a technical paper explained as though it were the third book in the Harry Potter series? Done. Want a difficult philosophical text translated into contemporary hip-hop lyrics? Give it a moment.

Why read a 300-page book when you can obtain a summary, ask questions, challenge the arguments, and have a conversation with an AI as though the author were sitting in front of you?

Of course, a summary doesn’t capture everything in a book. But the relevant question is whether you need everything in the book to achieve whatever you wanted from reading it in the first place.

If you want a particular insight, you might be better served by asking for it directly. If you want to understand a concept, you can have it explained at your level, tested through questions, and revisited until you understand it. If you want to compare several books, you can do so without spending weeks reading them all.

Perhaps the process of reading has intrinsic benefits. Perhaps struggling through a difficult text develops something that an easier route doesn’t.

But that argument often rests on an assumption that I find suspicious: that the difficulty of obtaining something somehow makes it better.

There is an almost religious attachment to effort. We tend to believe that if something required sacrifice, discipline, or considerable struggle to obtain, it must be more valuable than something that was easy to obtain.

Sometimes it is. But sometimes the difficulty is simply a cost imposed by the limitations of the available technology.

AI gives you a different journey. It can be easier, more interactive, and more rewarding. It can remove the tedious parts of learning while preserving, or even improving, the parts that matter.

Who’s to say that’s not superior?

No one argues that our ancestors lived better lives because they had to hunt for their food. We don’t insist on preserving every difficulty that technology has eliminated. We identify what we wanted from the activity and find more efficient ways of obtaining it.

If the destination can be reached more efficiently, and the journey provides no additional benefit that the person actually values, then the journey’s difficulty isn’t a virtue.

Why should reading be exempt?

There is also the question of attention. We’re already competing with an internet optimised to keep us scrolling. Short-form video, endless feeds, and algorithmically selected content make sustained attention increasingly difficult to justify. Now AI offers another shortcut: a way to generate exactly the information we want without having to seek it out.

And the supply of writing is exploding at the same time.

The internet is filling with articles, newsletters, guides, and posts produced in enormous quantities. Much of it is what we call AI slop: writing that is technically acceptable but recognisably generic, repetitive, and uninteresting.

Why bother reading any of it?

Why bother producing more of it?

But AI writing is slop

The obvious response is that AI writing is bad. It lacks originality. It repeats itself. It has no voice. It has no soul.

Some of that criticism is accurate about current outputs. But much of it is a strange way to evaluate the technology.

The internet is full of writing. There is more of it than any person could ever hope to read.

And a great deal of it is bad.

AI has made it easier to generate this stuff at scale, but let’s not pretend that human beings were producing masterpieces before ChatGPT arrived.

Don’t believe me? Spend some time going through fanfiction forums. Browse the self-published fiction market. Read random blog posts, product descriptions, online discussions, or the thousands of articles written because someone wanted to rank on Google.

The problems with AI writing are not mysterious. Models have recognisable tendencies, and those tendencies can often be corrected through better models, instructions, editing, and selection. The patterns we call slop today are not necessarily permanent features of the technology.

And the models are improving.

Which means that using the current weaknesses of AI writing as a permanent defence of human writing is a losing argument.

And even if today’s AI writing is often bad, it doesn’t follow that the underlying technology is incapable of producing good work. We are judging an evolving technology by its current failure modes while treating human writing as though its failures were exceptions rather than the norm.

Which is why the objection rarely stays there. It moves from quality to something harder to pin down.

What does it mean to say that writing has no soul?

People say AI writing has no soul. But what exactly is soul?

Sometimes it seems to mean that the writing is emotionally flat. Sometimes it means that the prose is predictable. Sometimes it means that the writer has no lived experience behind the words. And sometimes it seems to mean nothing more than that the reader has detected something they believe to be AI-generated and dislikes it.

These are different claims.

If soul means that writing evokes an emotional response, then we can examine whether it does. If it means originality, we can ask whether the ideas are original. If it means that a work reflects a particular person’s experiences, we can examine the relationship between the work and those experiences.

But if soul is an indefinable quality that we invoke whenever a machine produces something we don’t want to acknowledge as good, then it becomes a convenient way of protecting our assumptions from scrutiny.

I remember reading an essay on Aeon about the *illegible benefits of AI*whose author disclosed the use of AI in producing it. I thought the essay was good. It had an interesting idea, made a coherent argument, and was worth reading.

Yet much of the discussion surrounding it seemed to focus on the fact that AI had been involved. People were disappointed that the magazine had published AI-assisted writing. The disclosure itself became a reason to dismiss the work.

I found this strange. What did the authorship disclosure have to do with the quality of the argument?

If the argument was wrong, explain why it was wrong. If the writing was poor, identify the problem. If the article misrepresented something, challenge the claim.

But if the essay is good, and the only new information is that AI helped produce it, what exactly has changed about the work?

I suspect that some of the reaction had less to do with the work itself than with what the work represented. It challenged the assumption that good writing must be the product of a human writer’s effort. And if that assumption collapses, a great deal of the status and economic value attached to writing collapses with it.

There is an obvious incentive to resist that conclusion.

Writers have spent years developing their skills. Their identities, careers, and reputations depend on those skills. It is entirely understandable that they would be anxious about a technology that threatens to make them less valuable.

But anxiety isn’t an argument.

Nor is identifying AI-generated writing, announcing that you refuse to use it, or insisting that you can recognise its lack of soul.

Sometimes these positions become a form of in-group signalling. They allow people to demonstrate that they belong to a community that values human craft, that they possess the discernment to recognise machine-generated work, or that they remain committed to doing things the old-fashioned way.

There is also the satisfaction of being better than the people who use the technology: you, who espouse the virtues of human superiority, are nothing like those clanker lovers. Group identity. Tribal psychology. An in-group that defines itself partly by rejecting the practices of an out-group.

That said, it is important to separate the merits of an argument from the social incentives that encourage people to make it. The fact that a writer is anxious about being replaced doesn’t automatically make their objections invalid.

For example, there are the moral objections.

AI companies have trained models on enormous quantities of human-created material. Questions about consent, compensation, attribution, and the distribution of economic gains are legitimate. They cannot simply be dismissed because technological progress is exciting.

But neither does the existence of those questions establish that the resulting technology is inherently illegitimate, or that we should preserve an economic arrangement simply because some people benefited from it.

The moral questions deserve their own arguments. They should not become a substitute for evaluating what the technology can do.

If AI helps us discover new medicines, solve scientific problems, or produce things that materially improve human lives, those benefits matter. We still need to weigh them against the costs and the rights involved. But the mere fact that a technology threatens an established profession is not, on its own, a compelling reason to reject it.

Consider a hypothetical system that makes a major contribution to curing cancer. Suppose its development also displaces an enormous amount of existing intellectual labour. Would we conclude that the technology should never have been developed because some of the people whose work made it possible were not adequately compensated?

I find it difficult to accept that conclusion.

The benefits of scientific and technological progress can be so substantial that they justify profound disruption to existing economic arrangements. I am much closer to the position that progress can justify an extraordinary amount of disruption than to the position that a technology should be stopped because it threatens a particular profession.

Sometimes the ends do justify the means.

In an ideal world, there may be better ways to achieve the same results through ethical or less harmful means. The principle remains: protecting an existing distribution of economic value is not automatically more important than enabling technological progress.

And the idea that writing, or art more broadly, occupies some uniquely sacred position beyond technological advancement strikes me as a form of human exceptionalism that deserves examination.

Why should the humanities be exempt from the same technological pressures that have transformed every other field?

Why should we assume that there is some intangible quality in human writing that a sufficiently capable machine can never reproduce?

This is where our attachment to writing begins to reveal something about ourselves.

We want to believe that art occupies a special position. That it is somehow purer, more mystical, or more resistant to technological advancement than other activities.

We recognise that machines can outperform us in calculation. We readily accept that computers can process information more quickly than we can. We understand why a machine might be better at navigating traffic or optimising a supply chain.

But when the machine starts producing literature, writing poetry, or generating ideas, suddenly there must be something different about the activity. Something that transcends the ordinary rules of technological progress.

Why?

The Mona Lisa is a remarkable work of art. But if we are comparing things according to their economic or practical value to society, I don’t think it is self-evident that the Mona Lisa must rank above the intellectual machinery of a frontier AI model. I’ll even go further and say that, evaluated on the same scale with the same criteria, the weights of such a model aren’t in any way inferior in cultural significance or emotional beauty to the strokes of the Mona Lisa.

A painting can be culturally significant, emotionally powerful, and historically important. An AI model might contribute to advances in medicine, science, engineering, and countless other fields. The value of each depends on what we are evaluating and what we care about.

There is no reason to assume that the humanities must occupy some protected position above technology.

And if the purpose of human intellectual history was partly to accumulate knowledge, develop language, and build the conditions for systems that can extend those capacities beyond our biological limitations, then perhaps the emergence of such systems is not a betrayal of that history. Perhaps it is one of its consequences.

The problem with shifting the goalposts

Whenever a machine acquires a capability we thought was uniquely human, we look for another one.

Fine, AI can write, but it doesn’t have ideas.

Fine, it can generate ideas, but it doesn’t have taste.

Fine, it has taste, but it doesn’t have judgment.

Fine, it has judgment, but it doesn’t know which ideas matter.

And what happens when it does?

What happens when the next model, or the model after that, can produce ideas, evaluate them, exercise taste, make decisions, and pursue goals more effectively than we can?

There is no obvious reason to assume that the progression must stop at the precise boundary that makes humans feel comfortable.

A few years ago, the capabilities of these systems were substantially more limited than they are today. Now they can solve problems, generate useful code, assist with complex research, and produce work in domains that once seemed to require capabilities specific to human intelligence.

I don’t know where the ceiling is. Or if there is one. No one does.

But I find the repeated movement of the goalposts revealing.

Why must we keep finding a new capability that machines cannot possess?

The answer has something to do with our need to justify our place in the world.

We have built an enormous amount of our identity around what we can do. Our skills give us status, our professions give us purpose, and our contributions give us a way of measuring our worth. When a machine threatens one of those things, we don’t merely lose an economic advantage. We lose part of the story we tell ourselves about why we matter.

So we find another advantage.

But what happens when the machine crosses that goalpost, too?

Models that were considered incapable of particular tasks a few months ago can now perform them with striking competence. There is no guarantee that progress will continue at its current pace, but neither is there a good reason to assume that every capability we consider distinctly human will remain beyond reach.

If the technology eventually surpasses us in writing, reasoning, taste, judgment, and the ability to identify worthwhile problems, then what?

We can keep inventing new distinctions. We can keep insisting that the machine’s capabilities are somehow less authentic because they don’t arise from human biology. We can keep searching for a quality that can never be measured, tested, or replicated.

Or we can question the premise that our existence needs to be justified by our capabilities in the first place.

Do we need to be useful?

Suppose artificial superintelligence arrives. Suppose it can outperform the smartest humans in every intellectual domain that matters to us. Suppose it can develop scientific theories, solve engineering problems, write beautiful novels, compose music, make strategic decisions, and generate ideas that no human would ever have discovered alone.

What place does the human being occupy in that world?

One possible answer is that we find something else to do. We pursue leisure, relationships, hobbies, or whatever else we find meaningful.

But even that answer can become another attempt to justify ourselves.

We want to preserve a role. We want to remain useful in some way. We want to establish that even if we are no longer necessary for producing things, we are necessary for appreciating them, or choosing them, or deciding what should be done.

Perhaps.

But why must there be a point?

Why must we be useful? Why must we justify our existence through some contribution to the world?

We don’t ordinarily demand that a child demonstrate economic value before we consider their existence worthwhile. We don’t believe that a person has to outperform everyone else at a particular activity to deserve to participate in life.

At least, we don’t like to admit that we believe these things.

Yet when we consider a future in which machines can outperform humans at almost everything, we suddenly become desperate to find some activity that can preserve our relevance.

This reflects how deeply the idea of usefulness has become entangled with our sense of worth.

There are philosophical traditions that challenge this relationship between existence and predetermined purpose. Existentialism, absurdism, and related approaches differ considerably, but they offer ways to think about meaning without assuming that the universe must provide an externally guaranteed justification for our lives.

Our inability to identify a purpose does not automatically constitute a problem that must be solved.

There is no cosmic requirement that we remain relevant.

We can simply exist.

We can continue to write, paint, sing, play games, fall in love, argue about politics, go to church, and do all the other things human beings do without requiring those activities to establish our superiority over anything else.

If AI can write a better poem, that doesn’t mean I am forbidden from writing poetry.

If it can compose a better song, I can still enjoy composing music.

If it can generate better arguments, I can still have arguments with people.

If it can produce better art, I can still make art.

The existence of a better alternative does not eliminate the value of doing something yourself.

Horse riding is a good example. Cars have made horse riding largely irrelevant as a means of transportation, but that doesn’t mean people should stop riding. We ride for pleasure, competition, and countless other reasons.

Perhaps writing is heading in the same direction.

We will write because we enjoy it, because it helps us think, because we like expressing ourselves, or because the act of writing is meaningful to us.

And that will be enough.

But that doesn’t put food on the table

Telling a writer that their existence doesn’t need to be justified by their productivity is not a business model. It doesn’t tell them how to pay rent when nobody wants to buy their writing. If you want to make a living from writing, you still need people to value what you offer enough to pay for it.

There is a difference between the question of whether human beings need to remain useful and the question of how they should organise their lives in an economy that continues to reward useful things. We can reject the first premise while still needing to answer the second.

So what do you do when the technical skill that once gave you an advantage becomes widely available?

The answer requires us to stop looking exclusively at what a writer produces and start examining what people value about having that writer in their lives.

We have traditionally treated the relationship between a writer and a reader as a consequence of the work. The writer produces something good, the reader enjoys it, and a relationship develops around that exchange.

But what if we have the relationship backwards?

People follow particular writers not only because they want to read something good, but because they want to know what those writers think. They want to see how their views develop, how they respond to events, what questions they pursue, and how they make sense of the world.

They identify with them. They feel understood by them. They enjoy the experience of encountering another person’s mind over time.

Some of this is intellectual. Some of it is emotional. Some of it is social. And some of it is simply the pleasure of knowing and following a particular person.

These things are often treated as secondary to the work itself, but they don’t have to be.

In economics, there is a concept called relational goods: benefits that arise from relationships and interactions between people. There is also the related idea of relational labour, the ongoing work of building and maintaining relationships with an audience.

It’s not a new idea. Writers, musicians, performers, and other creators have always built relationships with their audiences.

What may be changing is the relative importance of those relationships when the technical production of the work becomes cheap.

If AI can produce an excellent essay, the essay alone may no longer be enough to distinguish its author. But an ongoing relationship with a particular writer is a different proposition. A reader may want to follow that person, participate in their community, discuss their ideas, or support their continued work.

A reader may want to know what you think, not because you have produced the best possible answer, but because they want to hear your answer. The human answer doesn’t have to be better. It just is.

The writing becomes one part of a larger offering.

And this is where we need to be shameless.

By shamelessness, I mean consciously doing what people have often done unconsciously. Stop pretending that the only legitimate reason to value a writer is the superiority of their prose or the profundity of their ideas. Stop treating personality, connection, identity, and the desire to encounter another human being as embarrassing substitutes for intellectual merit.

If people enjoy your company, value your perspective, or want to participate in the world you’re building, acknowledge that. Organise your work around it. Give people opportunities to engage with you, not just consume what you produce.

That might mean hosting live discussions, inviting readers into the development of your ideas, or building a community where readers can interact with one another and with you. It might mean making the process visible instead of presenting only the finished product, being candid about uncertainty, changing your mind in public, and taking responsibility for the consequences of your arguments.

The essay can become a conversation. The newsletter can become a relationship. The book can become an invitation into a body of work and a way of thinking that readers want to follow over time.

AI may eventually reproduce many forms of interaction, personalisation, and apparent intimacy so effectively that some of these preferences change. I am not arguing that human beings have discovered the one thing that technology can never touch. The point is that these are different sources of value from the technical ability to write, and we should investigate them as such.

The old game was to produce better writing than other writers. The new opportunity may be to stop playing that game exclusively.

Some of what people value was never in the words to begin with.

The things that cannot be reduced to the product

Consider a sermon.

You can read a transcript of a sermon. You can watch a recording. You can ask an AI to summarise it, explain its theology, or generate a sermon on the same subject.

But for many people, attending a service is not simply a way of obtaining the information contained in the sermon. It is an experience involving a congregation, a shared ritual, a particular place, and participation in a religious community.

A superintelligence might be able to produce a sermon more eloquent than any human preacher. It might understand the theology, anticipate objections, and generate language capable of moving an audience.

But whether it can praise God is a different question.

The same applies to other activities. A machine can generate the words of a love letter, but someone might still care that a particular person wrote it. It can compose a song, but a listener might want to attend a concert, see the performer, and share an experience with other people. It can produce a compelling account of grief, but someone might value hearing from a person they know who has actually experienced the loss.

And we don’t need to claim that AI is incapable of simulating the relevant experience to recognise this. Some people will be satisfied with the simulation. Others will not. Some will value the actual relationship, the provenance of the work, or the participation of a particular person.

The economic opportunity lies in understanding those differences.

This is not another attempt to identify the one human quality that AI can never replicate. It is an attempt to recognise that value depends on what people want, and that what people want is not always reducible to the technical quality of an output.

If we keep competing on the dimensions that machines are increasingly good at, we will keep finding ourselves defending a shrinking advantage.

So, is writing dead?

I don’t know if writing will disappear as an art form, and I don’t think it needs to.

People will continue to write because they enjoy it, because it helps them think, because they want to express themselves, and because they want to communicate with other people. There will be people who prefer human writing and people who don’t care who or what produced it. There will be people who value the process of creating a work as much as the finished product.

But none of that guarantees that writing will retain its current economic structure, or that technical proficiency will remain a reliable way to earn a living.

Perhaps writing will become like running: an activity people continue to do for pleasure, health, or personal satisfaction, even though it is no longer necessary as a means of transportation.

Perhaps it will become something else entirely.

What I find interesting is that we have difficulty accepting this possibility without first establishing that humans will retain some indispensable role.

We keep searching for the thing that makes us special. The idea, the judgment, the taste, the soul. We keep trying to prove that the machine cannot really do what it appears to be doing.

But what if it can?

What if the machine eventually writes better than we do, thinks better than we do, and makes decisions that produce better outcomes than we could have achieved ourselves?

We can acknowledge that possibility without concluding that human existence has become pointless. We can accept that some activities might no longer require us while continuing to pursue them because we want to.

There is a difference between being necessary to the world and being wanted by someone in it.

And if we still want to make a living from those activities, we can stop pretending that our only option is to outperform the machine. We can look instead at the relationships, experiences, and forms of participation that people value in their own right.

Maybe writing isn’t dead.

Maybe what is dying is the assumption that being able to write well is, by itself, a sufficient reason for anyone to care about what you write.

And perhaps that is the part we should be willing to let die.

Perhaps the sun went down, and the world felt small because, for a moment, I forgot to ask who had written the sentence.