Stephen King, by his own account, writes about two thousand words a day. That’s roughly the length of this post. He has done this nearly every day for fifty years, and the result is one of the most prolific output records in modern publishing.

The average adult reads at about two hundred words a minute. That makes King’s daily output ten minutes of reading material.

One of the most productive writers in history produces, per day, about one-sixtieth of what his reader could consume. That ratio - call it sixty to one, give or take - has always been the underlying economics of text. A writer’s daily output is always tiny compared to a reader’s daily intake. The asymmetry between writing-time and reading-time isn’t a recent development; it is the oldest fact about written language.

This is worth saying out loud, because the current AI moment has a tendency to sound like the asymmetry is new. It isn’t. What’s new is that the writing side just collapsed in cost.

What collapsed

For five thousand years, a sentence cost something to produce. Maybe a few seconds of King-grade fluency, maybe several minutes of careful craft, maybe an hour for a difficult paragraph. Whatever the rate, it was bounded by the human producing it. The natural throttle on text supply was that text required a person.

A large language model produces King’s daily output in about thirty seconds. The marginal cost is a fraction of a cent of electricity. The throttle is gone.

So the asymmetry didn’t appear; it went from sixty to one to roughly infinity to one. Anyone who wants the reader’s time now has unlimited material with which to ask for it.

This sounds like a problem of volume, and at the surface it is. Feeds, inboxes, recommendation engines, anywhere a paragraph fits is being filled with paragraphs. The supply curve has gone vertical.

But the deeper consequence is what happens to the meaning of the act of reading. When a human writes a sentence, there is a working assumption that the sentence reflects some thinking, however bad. When an AI writes a sentence, there isn’t. Reading used to be the act of receiving thought, with text as the medium. Now it is the act of evaluating text, and the question of whether thought is behind any given sentence is open every single time.

This is the kind of structural shift that doesn’t have a name yet, doesn’t have a movement around it, and is busily reshaping every reading habit you have whether you’ve noticed or not.

The pause

There is a thing reading does that writing does not, and it is the thing AI text-generation hasn’t touched yet. The reader stops between sentences. The reader looks away from the page, takes a breath, and generates - sometimes a rewording, sometimes a connection to something else, sometimes a disagreement, sometimes a question the writer didn’t anticipate. The reading is not happening on the page. It is happening in the space between the page and the rest of the reader’s life.

This is what writing is for. Not transmission of information; generation of new thought in another mind. The text is the catalyst. The pause is the payload.

The reason the sixty-to-one ratio made sense for five thousand years is that the writer’s investment was the price of admission for the reader’s investment. A writer who spent a King-level day producing two thousand words was paying for those ten minutes of the reader’s attention with hours of their own time. The asymmetry was steep but it had a shape. The writer’s hours signed the work. The reader trusted the signature.

The signature is now gone. The same two thousand words can now be generated with thirty seconds of GPU time. The visible artifact looks the same. The writer’s hours, which were the implicit guarantee that the writing was worth reading, no longer reliably exist behind it.

What this means for the reader is uncomfortable. The defense most readers reach for - filter harder, skim faster - runs in the wrong direction. Skimming faster now means processing more low-yield material per minute, which is the opposite of what a finite reader should be optimizing for. The defense that scales is reading less, more carefully.

What it looks like to write for the pause

A piece written for the pause has a different shape than a piece written for the click. It carries one idea, sharp enough to make the reader stop, far enough developed that stopping is worth the time. It treats the reader as a thinking partner who can be trusted to extend the argument without having every conclusion spelled out. It does not pretend to resolve what cannot be resolved. It leaves a question worth carrying away.

The shape is not new. Essays have done this for centuries. Long-form journalism has done it. The science-fiction tradition this publication takes its tonal cues from - The Expanse, Bobiverse, Children of Time - does it across thousands of pages. What is new is that the alternative has gotten cheap enough to drown the original. Pause-earning writing now competes for attention with infinite supply of skim-friendly summary. The skim-friendly summary always wins on volume. The pause-earning writing has to win on something else.

That something else is honesty about the argument’s limits. A piece that admits where its claim stops being supportable is one a reader can engage with. A piece that papers over the gaps is one a reader bounces off, or worse, files away as something they already understood. The pause requires the reader to feel like the next move belongs to them. Closing the argument too cleanly takes that move away.

So the bet here is that there is a class of reader who is now actively allergic to the closed, the polished, the resolved. Not because they don’t like good prose, but because they have noticed how cheap the appearance of resolution has become. That kind of reader is looking for writing that does the work and stops. Not writing that does the work and then tells them what to think about the work.

Who this is for

You probably already know: Someone who reads science fiction that takes itself seriously. They finish Bobiverse and wish there were more of that kind of attention paid to the actual present, because the actual present is starting to look uncomfortably like the setting for that kind of writing. Is that you?

The pieces here are about long-arc futures. AI and what it does to the institutions that adopt it. Civilizational coordination problems. What happens to politics and cooperation when the constraints change. Some pieces are about reading, like this one. Some are about the very old logic of empathy and cooperation, scaled forward into worlds that don’t exist yet. None are about zombies.

The cadence is weekly, deliberately. Every piece has to be good enough that publishing nothing is preferable to shipping something that misses. That filter is the publication. It’s the only product.

The asymmetry, again

The most important consequence of cheap writing is not the volume. It is that the writer’s time, which used to be the implicit signal of thinking, no longer reliably tracks anything. The reader’s old strategy - assume the writer spent something to write this, so spend something to read it - is now miscalibrated. The new strategy has to be built from scratch, by readers who have noticed the calibration has slipped.

For a writer, this means the work has to do something that thirty seconds of generation cannot do. It is not enough to be readable, or correct, or even pleasant. The piece has to hold multiple threads together over a sustained argument, in a way that an LLM optimizing one paragraph at a time still does poorly. The depth has to be structural, not surface. And the reader has to be able to feel that structure landing, not just be told it’s there.

The sixty-to-one asymmetry of King’s era was an honest deal. Ten minutes of reading, paid for with King’s ten-hour day. The infinity-to-one asymmetry of now is not a deal at all. The reader has to decide, sentence by sentence, whether anyone bought in on the other side.

What reading becomes when its purpose has to be rebuilt from scratch, by readers and writers who both know what changed, is the open question. This publication is one attempt at an answer.

Further Reading

  • The Shallows (Nicholas Carr) - what the medium does to the reader.
  • On Writing (Stephen King) - source of the two-thousand-words-a-day claim.
  • Reclaiming Conversation (Sherry Turkle) - what disappears when the pause goes.
  • The Silence Budget - the economic version of this argument.
  • Governed by Gradient Descent - the structural argument about what happens when optimization runs on systems that decide things.