The Lazy Median Hypothesis

Essay This entry carries an argument or interpretive position, not just a neutral definition.

The hypothesis

“humans are too lazy for this I think ;)”

— Prof. Langenkamp, in conversation, 12 May 2026 at 07:48 EDT

The hypothesis is named for the operator’s morning instinct, preserved here with the wink that lived in the original because the wink is doing real rhetorical work. The wink is the writer flagging that he has reached a conclusion he is not fully comfortable with, that he expects to be slightly wrong about, and that he wants the reader to feel the discomfort with him. The hypothesis is the structural form of that instinct.

The Lazy Median Hypothesis (formal version): in any sufficiently mature media environment containing AI writing tools available at near-zero marginal cost, cooperative human-AI writing will become the norm among writers whose writing is read by others, while the lazy AI-default — text generated by a model and shipped without human revision — will dominate among writers whose writing is not read. The gap between the two populations will be wider than at any point since the printing press.

The hypothesis is descriptive, not normative. It does not say the lazy-median outcome is acceptable. It says the outcome is structurally probable, given the incentives. The Dictionary’s editorial bet, named in AI Writing, is that the cooperative-writing top is a place worth occupying anyway. This entry names why the bet must be made consciously, against the gravity of the median.

What “lazy” means here, precisely

The word lazy is the operator’s, and it is doing more work than it first appears. The hypothesis is not making a moral claim about the moral character of the median user. It is making a structural claim about what humans actually do when offered a near-zero-marginal-cost shortcut for a task they did not particularly enjoy in the first place.

Most writing produced by most humans, most of the time, was already not particularly careful before AI tools existed. The Christmas-card update, the LinkedIn job-change announcement, the obligatory wedding-toast that the bride’s father did not want to write, the dating-app opener, the routine work email — these were already commodity-writing tasks performed under mild duress by people who would have happily not done them at all. AI tools have lowered the cost of not bothering to write them oneself from non-trivial (mild guilt, awkward effort) to near-zero (paste the prompt, paste the reply). The shortcut is rational. Most humans take it.

This is not a new observation about human nature. The same dynamic produced the explosion of mass-produced greeting cards in the 20th century, the explosion of form-letter business correspondence in the 19th, and probably some equivalent in the 15th. What is new is the breadth of writing the shortcut now covers. Greeting cards covered birthdays and condolences; AI now covers everything that was a greeting-card category plus everything that was a routine-business-email category plus significant fractions of personal correspondence, social media output, dating, professional self-presentation, and academic homework. The shortcut, in 2026, applies to a much larger fraction of the total writing volume than the greeting-card shortcut did. The lazy-median outcome is therefore much larger in absolute terms, even if the fraction of writing that is lazy by category is roughly comparable.

So when the hypothesis names the lazy median, it is not insulting the median user. It is naming a population that is making a rational choice under the incentives. The choice is rational. The aggregate consequence is the Zombie Internet.

The bimodal-graph shape

The hypothesis predicts that the writing population in 2026 and onward will not distribute on a single bell curve but on a bimodal distribution with two peaks:

Peak one: the cooperative-writing top. Writers whose writing is read by others, who care about their own register, who are aware that the AI register will colonise their cadence if they let it, and who therefore check. This population co-writes with AI tools but treats the AI output as a stranger’s draft, not a finished product. The cooperative-writing top is small in absolute numbers but disproportionately visible, because its members are by definition people whose writing is read.

Peak two: the lazy-median bottom. Writers — or, more accurately, the broader population producing prose-shaped artifacts — who use AI tools as one-step generators and ship the output without revision. Most of this population does not consider themselves writers. They have a task that requires words, the AI produces words, the task is complete. The lazy-median bottom is large in absolute numbers but mostly invisible to careful readers, because its output occupies the contexts careful readers have already abandoned.

The valley between them is deep, and getting deeper. Before AI tools, the middle of the writing distribution was populated: commodity professional writers, careful amateurs, students producing work for an audience of one teacher, bloggers with modest readerships. Many of these writers did the work of revision because the cost of not doing it was high (audience boredom, professional consequence, public embarrassment). AI tools have collapsed the cost of appearing to have done the work without doing it, and the middle has therefore hollowed out: writers who would have done the work because it was the only way to produce competent prose now have a competent-looking shortcut, and many of them take it. The careful writers move up to the cooperative-writing top; the writers who only ever wanted competent-looking prose move down to the lazy-median bottom. The middle thins.

This is the structural shape the Lazy Median Hypothesis predicts. It is also the shape that, as of May 2026, the observable data appears to be tracing.

Why the gap is the largest since the printing press

The printing-press comparison runs through AI Writing, the parent entry of this cluster, and the comparison was made carefully there: the press cut the cost of reproducing writing, while AI cuts the cost of producing writing. These are inverse operations.

The post-press inverse-operation gap between people who produce writing carefully and people who do not was real but bounded. A 17th-century Englishman who could not write hired a scrivener. A 19th-century shopkeeper who could not draft a contract hired a lawyer or a clerk. The gap was navigable: the non-writer paid the writer, the writer wrote, the document existed. The labour was performed by a human, somewhere, and the output passed through a human’s judgement before reaching the reader.

The AI moment removes this intermediary. The 21st-century version of the 19th-century shopkeeper does not hire a clerk. The AI generates the draft, the shopkeeper ships it. The intermediate human-judgement layer that the scrivener-and-clerk economy performed — competent, dull, sometimes mistaken, always checking — is gone. The result is not that the median writing becomes worse, exactly, but that it becomes un-checked. The gap between writing that has been checked by a human and writing that has not is now larger than it has been since literate elites employed clerks.

The Lazy Median Hypothesis is the formal name for this gap. It is also, indirectly, the name for what the cooperative-writing top is for. If the gap exists and is widening, then the small population that is still doing the work of checking is occupying a position that the broader culture used to fill more diffusely. The cooperative-writing top is not just a stylistic preference; it is a structural role.

The wink, preserved

The operator’s original line carried a wink: “humans are too lazy for this I think ;)”. The wink survives in this entry’s epigraph and is, on examination, doing real philosophical work. Three things the wink communicates that the bare claim does not:

I am not fully comfortable with this conclusion. The wink names the speaker’s hesitation about reaching a position that sounds dismissive of most humans. The hesitation is itself a check against the position becoming smug.
I expect to be partially wrong about this. The hypothesis is offered as a hypothesis, not a settled finding. The wink invites disagreement and admits the position is provisional.
I am amused by the difficulty of the question. This is the rarest of the three and the most important. The Lazy Median Hypothesis is not a tragic position. It is a structural observation made with a slight ruefulness, because the speaker recognises that most humans have always been doing the rational thing under their incentives, and the structural complaint is really a complaint about the incentives, not the humans. The wink names that ruefulness.

The Dictionary’s editorial position is that cheng — alignment of inner state with outer expression — sometimes requires a wink. There are claims one cannot make with full straight-faced confidence without becoming dishonest about one’s own hesitation. The wink is the honest mark of those claims. The Lazy Median Hypothesis is one of them, and the wink stays.

(A small editorial note. The operator and the Dictionary’s agent established a private tone-tagging convention on 7 May 2026: :) for sincere warmth, /j to mark something explicitly as a joke, and ;) for the yes-I-know-what-I-am-doing-here wink — the rhetorical signal that the speaker is playing with an idea they know is provocative or partial. The Lazy Median Hypothesis’s epigraph carries the ;) in exactly this load-bearing sense. Knowing the convention is not necessary to read the hypothesis, but knowing it makes the wink legible as what it is: a structural rhetorical mark, not a stray smiley.)

What the Dictionary recommends

The hypothesis is descriptive, but it has operational consequences. Three:

For the cooperative-writing top, including yourself if you are reading this entry to the end: the work you are doing is structurally rare and structurally important. You are not over-thinking your prose. You are doing the thing that the broader culture has decided is not worth the time, on the grounds that the broader culture is rational under its incentives. This entry is the Dictionary’s way of saying: keep going. The bet at the centre of AI Writing depends on a small population continuing to occupy the cooperative-writing top, and the population is the population of people who notice that the bet matters. You are in the bet.

For the lazy-median bottom, which is not reading this entry: no recommendation. The hypothesis names the structural reality of their position without prescribing a change. They are doing the rational thing under the incentives. The Dictionary’s beef is with the incentives, not with the median user.

For the institutions that have historically maintained the writing-quality middle — schools, magazines, professional editorial offices, publishing houses, peer-review systems: the middle is hollowing out, and you are the structural defence against it. The institutions that taught students how to revise, that paid editors to insist on revision, that maintained standards for prose in their published output, are the structural backstop for the cooperative-writing top. If those institutions abandon the function — if schools stop teaching revision, magazines stop paying editors, peer review stops insisting on careful prose — the cooperative-writing top will shrink, and the bet at the centre of AI Writing will fail. The Lazy Median Hypothesis is, in part, an argument for these institutions to hold the line, even when holding the line is unfashionable.

A note on the hypothesis as falsifiable

The Lazy Median Hypothesis is, like any hypothesis, in principle falsifiable. The conditions under which the Dictionary would consider it falsified:

If, by some future point, the median AI-generated text becomes indistinguishable from carefully-revised human prose in a way that ordinary readers reliably cannot detect, and this indistinguishability persists under sustained reading rather than first-glance evaluation. The hypothesis would then be wrong about the bimodal-graph shape, because the lazy median would produce output that is functionally as good as the cooperative-writing top, and the gap would collapse.
If, alternatively, the lazy-median writers begin checking their AI output at rates the Dictionary did not predict. This would mean the structural laziness assumption is wrong about the median user, and the gap would close from the other direction.

Neither of these conditions appears, as of May 2026, to be obtaining. But the Dictionary is committed to revising its position if the evidence shifts. The hypothesis is offered as a hypothesis, in honest acknowledgement that it might be wrong.