Workbrain The Slop Trilogy
01 / 14 Add to Chrome

01The Slop Trilogy · part one

Slop stopped being an insult.

It arrived as a slur and it was a fair one — the endless scroll of six-fingered stock art, the playlists nobody wrote, the listicles assembled by something that had never visited the city it was describing. The complaint underneath was moral: somebody took a shortcut and expected you not to notice.

02Why the word stopped fitting

It became the road.

Not because the shortcut went away. And once everybody is on the road, the accusation has nothing to point at.

As it was coined

Content made without care, by someone who couldn’t be bothered, and pushed out in volume because volume is cheap.

A judgement about effort. It assumes a lazy author, and the cure is that the author tries harder.

03You have already felt it

The tell isn’t quality. It’s absence.

You open an email and something is off before you can say what. The grammar is immaculate. The structure is sensible. The tone is warm in the way a hotel lobby is warm. And somewhere in the second paragraph you think, without deciding to: they didn’t write this.

You can’t point to the error, because there isn’t one. What is missing isn’t competence. What is missing is any evidence that a specific person was here — the phrase they always use, the thing they’d never say, the way they open when they’re in a hurry, the joke they’d have made. You noticed a hole shaped like somebody.

04

Slop is no longer about laziness or bad intent. It is the absence of context — and that absence has a signature.

05Why the old word fits badly

They were not being lazy

They may have worked harder on it than they would have three years ago: drafted a prompt, read the output, adjusted the tone, sent it. Real effort, real attention — and the thing that arrived still had nobody in it.

06Interactive · the convergence

Forty-eight people, writing the same brief

Each block is one person’s output. At zero they write as themselves.

0%
distinct 48/48 from the mean 100% yours: one of the crowd

A model, not a measurement. Nothing here is data about real writing — it’s the arithmetic of predicting the most likely next word for everyone at once.

07The second-order effect

When everyone gets excellent, excellence stops being a signal.

A beautiful website used to mean something. It meant you had either the skill or the budget, and either one told a prospective client something true about you. When anyone can produce a beautiful website in an afternoon, a beautiful website tells them nothing.

The signal did not get weaker. It stopped being a signal.

08The expensive version

Immaculate prose becomes the floor.

When every applicant sends immaculate, tailored, keyword-aligned prose, immaculate prose no longer separates anyone. And a floor that everybody stands on isn’t a floor anybody gets credit for.

09Not taste — structure

Why this gets worse rather than settling down.

A model writes by predicting what is most likely to come next, across everything it has read. The most likely phrasing is, by construction, the most common one. Point that at a population and it doesn’t produce forty-eight tailored outputs; it produces one output forty-eight times, with the edges sanded off each of them.

The tool is not failing when this happens. This is the tool succeeding.

10Where the fix is not

A better model predicts the likely thing more fluently.

The only thing that moves an output away from the middle is information the model couldn’t have guessed — and there’s exactly one source of that.

11Interactive · a quiet inventory

How many of these have you done this month?

Nothing is recorded, sent or stored. The count dies with this tab.

Tick what applies. Nobody is keeping score — least of all this page.

12The part that is actually at stake

It stops being a style problem.

Writing is no longer a thing we do beside the work. It’s increasingly the interface to the work: the application, the proposal, the update, the case, the pitch. The person whose language doesn’t sit near the middle of the distribution — a trade with its own exact vocabulary, a family with its own words for things, anyone writing in their second language — gets flattened hardest and flagged most often.

The cost of the convergence isn’t distributed evenly, and the people paying most aren’t the ones who chose it.

13Nearly one to one

The gap where you should have been.

Slop, 2022→A lazy author taking a shortcut. The cure was to try harder.
Slop, now→Writing with the context taken out. The author may have worked harder than ever.
The convergence→Not a fault. The tool predicting the most likely next word, for everyone at once.
A better model→Predicts the likely thing more fluently. It moves nothing off the middle.
The only thing that does→Information the model could never have guessed — specific, personal, awkward, yours.
Where it has to live→A file. Plain text you own, handed to whichever assistant you use this month, and taken with you when you leave.

The slop isn’t coming from the machine. It’s coming from the gap where you should have been.

14One minute, before anything is installed

Find out what yours already knows.

Ask the assistant you already use to describe you, then to list what it would need to know to sound exactly like you. Most people find the answer isn’t wrong — it’s thin. And the second half of it is a specification for the file you should have been keeping.