Scribo, ergo sum

Why I am bullish on writing in an age where we seemingly do not need it anymore

In 2019 I was partway through a data science internship when I read about a breakthrough technology called GPT-2. With minimal human instruction it could produce sentences, and even paragraphs, that were both syntactically and semantically consistent. In other words, its outputs were grammatically correct and even had coherent meaning. When I reached the end of the article, I was informed that the text I had just read had itself been written by GPT-2, and my jaw dropped.

At the time of writing, in 2026, hundreds of millions of people are using its descendant, GPT-5.5, every month: mostly through the interface known as ChatGPT. Large Language Models, as we have come to know them, have transformed the way we approach our work, our academics, and even our close relationships. Where before we had to write to communicate, we can now generate. And it is a global phenomenon, with various Western and Eastern model providers bringing this technology to more and more of humanity.

It is entirely possible that we are witnessing a new form of abstraction for communication come into being. Short generative prompts, which are a distillation of our thoughts and intent, will be decoded into whichever length, format, and tone best suit our audience and medium. Even when the message is short, rather than type (or say) anything directly to our recipient, we may instruct our language-model as to what we want said. You can imagine one RSVP'ing to an invite by instructing "tell them I'll be there" rather than writing "I'll be there," because the message itself might require different length and phrasing depending on the relationships of the parties involved.

This is all incredibly efficient, and might even lead to further abstractions as technology improves: we might just think and have our intent dutifully communicated to another, who might even respond in the exact same way. Ironically, the guttural grunts and gestures of our ancestors may once again become the norm for communication, but with the ability to decode thought at higher resolutions than ever before.

Why strain ourselves to speak, or at distances, write, when technology removes the requirement?

Well, I insist we write. Or at least, I insist those who prize intellect, legacy, connection, and their own humanity, to write. It is a locally inefficient but, at least for what I value, maximally useful way to go about life.

Intellect

Simply put: writing is good for the brain. This is well documented, and part of its value is that it is difficult; especially at the beginning, where it can feel terrifying. To quote Micah Nathan, who is a lecturer in writing at MIT: "The act of confronting that terror is, itself, an education for the writer, because writing is both vehicle and vessel for thinking - abstract made concrete, feelings translated into words...writing isn't just the production of sentences - it's the training of endurance by way of sustained attention."

Once you sharpen your mind, you can go places. Decades of American tech innovation has been led by what we could call "articulate engineers," and I would wager that the clarity of thought required to build and lead companies at that level is a hard-won skill in which writing would have played a role. Here's the thing: the world has no shortage of ideas, but it lacks people with vision. A vision necessarily has to be articulated, otherwise it is a fuzzy amalgam of ideas, loosely tied or fruitlessly lassoed by a thread of ambition. You cannot tie a rope around a cloud, but you can around a block of ice.

So, what is left when we write less and instruct more? The empirical evidence is trickling in (a preliminary MIT Media Lab study found that participants who used ChatGPT to write essays showed lower neural connectivity than those without assistance), but my own take is that we will become more reliant on abstractions that eventually constrain our ability to self-express.

In 1984, George Orwell invented "Newspeak," a fictional language spoken across totalitarian Oceania. It was composed specifically to facilitate the communication of facts, not abstract thought, and thus comprised words such as good, plusgood (very good, great), doubleplusgood (excellent, magnificent, superb), ungood (bad), and uncold (warm). This simplified vocabulary was efficient, but reductive; and if instructing LLMs becomes as easy as a disjointed list of facts and Newspeak-eqsue tokens, which then unfurls into beautifully generated prose, I think our individual minds are in for a very ungood time. Especially since we still need to communicate with each other over the phone or in-person. One might believe that despite their dependency on Newspeak for anything written, they will rise to the occasion when it matters; but in my experience most people do not rise to the occasion, we regress to the mean.

Legacy

I have a complicated relationship with writing for posterity, and have never been able to commit to a journaling routine, despite recognizing the inherent, aforementioned, brain-enhancing value of writing. The apparent futility of it always caught up with me. Would I be able to get all of my thoughts, or even a reasonable fraction of them, down on paper? Would I lose my notes, or would they become buried in some folder on my computer? And once I'm dead, would anyone care to read what remains?

A beautiful thing about words is that they outlive their writers, or rather, immortalize them. We still quote and read Plato, Shakespeare, and Austen today, and I suspect future generations will pore over the writings of Paul Graham, Haruki Murakami, and Margaret Atwood long after they have left us. Many subscribe to the "great man" theory of history, the idea that several people have lived such momentous lives that they changed the course of our civilization; but I prefer the "great idea" theory, which suggests that history has been marked by radical ideas which infected a group of people, often through some influential vector: and often after sitting dormant long after its originator had died. All this is to say, if you write well, you can trigger pandemics of thought that ripple throughout history.

For that reason, there is an asymmetric upside to writing. Penning your thoughts, observations, and experiences costs very little and has inherent benefits. The same goes for sharing what you write. But, there is also a chance you will be long-remembered for what you have written, and that success may even happen in your lifetime.

So why then, if the artifacts of writing might truly become artifacts studied generations into the future, does it matter that you wrote it; rather than generated it with artificial intelligence? My answer is that the best writing, the stuff that gets remembered, will still be written by humans.

Shakespeare did not just demonstrate masterful command of the English language (which LLMs can clearly do too), he experimented with novel uses of it while also inventing hundreds of new words for it. Writers like him also changed their minds, made mistakes, went through identity crises: all things you might go through too. In contrast, language models are fundamentally probabilistic and select the most appropriate words from within their distribution of choices. They are primed for reliability and cannot produce things they have not seen before.

And people have come to notice that: there was a time when I could not spot AI as hard as I tried, and now I have a very good filter for it. Not because it has gotten worse, but because it has become reliably perfect. An android, with its uncannily perfect features and proportions, could in isolation pass the Turing Test and be indistinguishable in a crowd, but if you walked past fifty of them a day it would become hard not to know what to look out for. [1]

Good writing is spiky, irreverent, and iconoclastic. Copernican and Darwinian ideas that we treat as basic science today were once absolutely insulting, and one thing language models are supposed to avoid is offense. So as we go about life reading more content generated by AI, we will become more appreciative of what is evidently written by humans, because it will make us pause and may even offend us. And for that it will become memorable, and leave a legacy.

Connection

Earlier I suggested that one day people may be able to communicate with one another by thought alone. Even if that world comes to pass in our lifetimes, and even if that method of communication preserves all the nuance that I have argued only becomes possible when we commit to the act of writing - I still believe that our ability to communicate, sharpened by writing, will be of huge import. This is because we live in a global and intergenerational society.

To elaborate on the first, technology diffuses slowly. The internet has been around for decades, yet most business is still offline. Commercially viable electricity is almost 150 years old, yet some communities remain unelectrified. Today's shortage of compute means that AI will largely be accessible by people whose governments, and customers whose companies, can afford to be first in line for new supply. So if you sit in America, and believe in connecting with people from around the globe, for business or just for connection's sake, you should be prepared to communicate multi-modally: meaning you cannot yet let your written or spoken communication slip away.

Even in countries where the technology becomes widely available, some might just prefer not to use it, and this is where my intergenerational point comes in. As sure as death and taxes, technophobia is an aspect of the human condition. Older generations will find prompt-based communication frightening and, paradoxically, inefficient, because it will burn more calories adjusting to a new interface than the time it saves once used. Whether you value meaningful time with your grandparents, or just want to convince capital to support you (bear in mind that in the U.S. 75% of wealth is owned by people over the age of 50), writing will unlock multi-modal capabilities that will allow you to connect with these people.

I really cannot stress enough the importance of inefficiency in relationship building. It is something that was intuitive when hanging out with friends growing up in South Africa, but that as I grew older I had to relearn. An agenda is a useful way to ensure that critical information is transmitted, but it is the interactions outside of a meeting where trust is built: I learnt this while out drinking with clients in China where I fumbled Chinese greetings, and during airport layovers on MBA treks while adjusting to life in America. The ability to meet the needs of the situation, and risk awkwardness and miscommunication in attempting to do so unassisted by technology, across social lines and age groups, can be an awesome advantage. Writing can help you form a first draft of your personality for these situations, and serve as training in between, and people will reward you for trying: because they get exhausted by screens, perfected prose, and translators; or might not even have those alternatives available to them.

It is perfectly okay to exist as part of a subculture which adopts this technology to its extreme, and broadly I think younger people in developed countries will become this subculture as a matter of course. They will communicate in modes that alienate themselves from the rest of the world, because it is internally efficient (after all, this is how slang develops), and for the most part will lead functional lives. But my feeling is that anyone who wants to lead an ambitious life will have to surrender themselves to the inefficient, multi-modal reality of working with people older and/or poorer than them, or just different from them. [2]

Humanity

As we approach a world where different AIs will be vastly better than us across various physical and cognitive tasks, perhaps even in concert, there is a question of whether there is any point at all in trying to get good at something, like writing. I have expounded the functional benefits, but what about a post-work world where ambition is not strictly necessary anymore, how then do we spend our time, and why?

Writing is not just a means to various ends, it is a pleasure in itself. Not every great fiction writer used their story as a vehicle to deliver some layered message; many just wanted to tell the story, and relish in the challenge of it. The same can be said for any hobby, it is not about being the best at something or making a career of it, but about an individual sense of novelty, progression, and self-expression. We are not deterred from weightlifting contests because gorillas have always been stronger than us or because machines have been capable of lifting huge stones for thousands of years. Nor should we be discouraged now that another entity can outclass us in tasks thought to be mostly cognitive.

Speaking again of efficiency, now that we have to do fewer things ourselves, we should pick the set of activities that we want to spend more time doing, even if we are not the best at them. Use the time we have been given back to waste more time, as ridiculous as that sounds. For me, writing is one of those things. So is cooking (but not cleaning), so is learning languages. I will do these things even beyond the point where technology trivializes them, simply because they enrich my existence.

To sprinkle in some caution, then, I worry that if we never really have to write for formal purposes (academic, professional), many will never discover the spiritual power of the written word: its ability to conjure entire worlds and, bypassing so many defenses, write them to other people's brains. To evoke joy, sadness, and a pantheon of emotions in both the writer and the reader. Society, then, should never deprecate tasks which could be hobbies just because the machines have become adept at them.

To pause before the end, I feel I must clarify that I do admire AI. I use it every day. It has helped me perform previously intractable research, it got me to fall in love with building software again, in many small ways it helped me get into MIT and navigate moving to America; even the site you are reading this on was built with AI, and I feel very little shame in that. But, I do take issue with it becoming a crutch for written communication: a skill which, once lost, can be very hard to re-acquire. More generally, I believe that thoughtful refrain can help us develop skills that will make us better at using AI.

I think, therefore I am? Thinking is no longer something uniquely human, or proof that you exist in any meaningful capacity. Thought is happening right now, as we speak, in datacenters. Sand is thinking. But writing, and I am speaking about a specific configuration of thought; that is something that can be very human. By hunching over a table, with a pen or a keyboard, grappling with my mind as I commit to paper or PDF only a fraction of what I have thought or witnessed in a day, as my life passes me by, as I have no clue who or what will process this information, whether it will be deemed useful or novel, whether it will be preserved in or beyond my lifetime: that, makes me remarkable. It is not efficient, but I am not a drone, I am not meant to live a good or doupleplusgood life. I am meant to live an excellent, whimsical, well-meaning but hopelessly inefficient life.

I will close with the words of others, that could have only been communicated after a lifetime of writing - that struggle with the mind - and with the wisdom that it brought:

"We write to taste life twice, in the moment and in retrospect" - Anais Nin

"The first draft of anything is shit" - Ernest Hemingway

"The discipline of the written word punishes both stupidity and dishonesty" - John Steinbeck

"You must write, and read, as if your life depended on it" - Adrienne Rich

A very big thank you to Nicholas Makins, Anton Hantel, Nikhil Baddam, Di & Tony Wilson, and Charlotte Otter for reading drafts of this essay.

Notes

[1] A common argument is that with the right level of prompting, the right context and parameters for generation, people will not only be able to produce good, indistinguishable-from-human content, but at a scale previous writers never could have achieved in their lifetimes. My counterargument is that the level of prompting required to get there will be tantamount to writing, and few will try it. And the assumption that at the outset we know exactly what we want to write, is one I would firmly challenge. As we write, we produce micro-innovations, and possibly come up with the ideas that our product will become remembered for. More than half of the ideas in this essay were not in my head at the time I committed to writing it, and there is a good chance that whatever has resonated with you would not be here if I had dumped my ideas into a prompt, then edited a draft.

[2] There is one more reason to be excited about writing as a means of connection. Even within your subculture, for the reasons I have listed above (writing makes you smarter, and good writing makes you stand out), I think you will find peers, and peers will find you, through your writing. I read a blog post once titled, "a blog post is a very long and complex search query to find fascinating people and make them route interesting stuff to your inbox," and I think that title speaks for itself.